Electoral Studies 83 (2023) 102615 Contents lists available at ScienceDirect Electoral Studies journal homepage: www.elsevier.com/locate/electstud Ideological identity, issue-based ideology and bipolar affective polarization in multiparty systems: The cases of Argentina, Chile, Italy, Portugal and Spain Josep M. Comellas a, b, *, Mariano Torcal b a Universitat Autònoma de Barcelona, Department of Political Science and Public Law, Edifici B, Campus de Bellaterra, 08193, Cerdanyola del Vallès, Barcelona, Spain b Universitat Pompeu Fabra, Department of Political and Social Sciences, Ramon Trias Fargas 25-27, 08005, Barcelona, Spain A R T I C L E I N F O A B S T R A C T Keywords: In multiparty contexts, we know that affective polarization tends to cluster in ideological blocs, although the Affective polarization factors driving this process are still quite unexplored. In this paper, we contribute to filling this gap in the Ideological identity literature by exploring the capacity of ideological identity vis-à-vis issue-based ideology to polarize sentiments Issue-based ideology towards party voters into two opposing left-right blocs. Specifically, we provide empirical evidence that affective Southern Europe Latin America attachments to ideological labels increase the affective distance between ideological blocs to a greater extent than issue extremity and issue consistency. These bipolarizing effects of ideological identity persist even when the identity is inconsistent with issue-based ideology. Additionally, we show that bipolar affective polarization exerts little reverse influence on ideological identity. We support these arguments using an original survey from the TRI-POL project carried out in five multiparty systems: Argentina, Chile, Italy, Portugal and Spain. 1. Introduction ideological and social cleavages (e.g. Huddy et al., 2018; Kekkonen and Ylä-Anttila 2021; Kekkonen et al., 2022; Reiljan and Ryan 2021). Affective polarization broadly refers to the tendency to view However, our understanding of the factors and dynamics that drive this opposing partisans negatively and co-partisans positively (Iyengar et al., tendency of affective polarization to cluster in ideological blocs in 2012; Mason, 2016). This type of polarization has mainly been studied multiparty settings is still scarce, missing the opportunity to disentangle in the Unites States (US) context (Iyengar et al., 2019), characterized by the relationship between affective polarization and ideology. the existence of a two-party system with two ideologically diverging In this paper, we contribute to the study of affective polarization in parties. Most of the literature exploring the nature of the antipathy be­ multiparty systems by focusing on the role of ideology in polarizing tween Democrats and Republicans focuses on determining whether this sentiments towards party voters into two opposing left-right blocs, but type of polarization is mainly caused by the increasing salience of distinguishing between ideology understood as a belief system (i.e., partisan social identities (e.g. Dias and Lelkes 2022; Iyengar and West­ issue-based ideology) and ideology understood as an affective attach­ wood 2015; Mason, 2018a) or the increasingly stark policy and/or issue ment to ideological labels (i.e., identity-based ideology). The literature positioning disagreements (e.g. Bougher 2017; Lelkes 2021; Webster shows that ideological identity contributes to accentuate perceptions of and Abramowitz 2017). In recent years, a growing number of studies elite ideological polarization (Vegetti and Širinić 2019), fuels political have explored affective polarization in multiparty systems around the engagement (Oshri et al., 2021) and, in the US context, is more strongly globe (e.g. Boxell et al., 2022; Garzia, Ferreira da Silva and Maye 2023; associated with the affective distance between liberals and conservatives Gidron, Adams & Horne 2020; Guedes-Neto 2022; Reiljan 2020; Torcal than issue-based ideology (Mason 2018a, 2018b). However, the litera­ and Comellas 2022; Wagner 2021). Some of the studies on this topic ture exploring affective polarization in multiparty systems has not show that, in multiparty contexts, citizens usually display positive distinguished between these two different types of ideology. While in feelings towards several parties and animosity towards another set of the US two-party system partisan and ideological identities largely parties, conforming different affective blocs that tend to be related to overlap, in multiparty contexts the same ideological identity is usually * Corresponding author. Universitat Autònoma de Barcelona, Department of Political Science and Public Law, Edifici B, Campus de Bellaterra, 08193, Cerdanyola del Vallès, Barcelona, Spain. E-mail addresses: josepmaria.comellas@uab.cat (J.M. Comellas), mariano.torcal@upf.edu (M. Torcal). https://doi.org/10.1016/j.electstud.2023.102615 Received 15 November 2022; Received in revised form 3 March 2023; Accepted 26 March 2023 Available online 5 April 2023 0261-3794/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). J.M. Comellas and M. Torcal Electoral Studies 83 (2023) 102615 shared by different political parties (Garry 2007; Wagner 2021), so that which individuals are ideologically consistent in a battery of different ideology constitutes a supra-partisan identity with the potential to exert issues (issue consistency). its own particular effect on perceptions and affective evaluations of However, ideology can also be understood as a social and political political objects, including party supporters. identity, not necessarily rooted in a coherent set of opinions on different In the following pages, we argue that the left-right ideological labels concrete issues (Conover and Feldman 1981; Levitin and Miller 1979). work in many multi-party systems as group identifiers, so that citizens This identity-based type of ideology is defined as individuals’ affective tend to perceive parties and partisans of their own ideological bloc as in- attachments to ideological labels and to the people and symbols asso­ groups and those of the other ideological bloc as out-groups while, at the ciated with those labels. It has been primarily attributed to the influence same time, emphasizing the differences between partisans belonging to of social groups (e.g. Popp and Rudolph 2011). Symbolic attachments to these two ideological blocs (Vegetti and Širinić 2019). As a result, any of the ideological groups significantly influences policy preferences ideological identity leads to the development of a bipolar type of af­ and political behavior, although their effects depend on the existence of fective polarization characterized by a strong animosity between the some social psychological and emotional ties with one of the ideological party supporters who, respectively, belong to the left and right blocs. By groups’ members (e.g. Devine 2014; Malka and Lelkes 2010; Popp and contrast, issue-based ideology is expected to contribute significantly less Rudolph 2011). This type of ideology has received different names in the to this bipolar affective polarization. In addition, identity-based ideol­ literature, such as ‘symbolic ideology’ (Ellis and Stimson 2009), ‘ideo­ ogy is not necessarily the product of a coherent set of political beliefs, so logical social identity’ (Devine 2014), ‘ideological identity’ (Malka and its bipolarizing effects may be present even among citizens with centrist Lelkes 2010), or ‘identity-based ideology’ (Mason 2018a, 2018b). positions on salient policy-issues or those with ideologically inconsistent Sometimes it is measured using the traditional left-right self-placement beliefs on different issues. Finally, we also explore whether bipolar af­ scale, although it is preferable to use survey items that ask about the fective polarization exerts significant reverse effects on identity-based degree of identification with specific ideological labels (Mason 2018a). ideology. We test these expectations in the multiparty systems of Argentina, 2.2. Towards bipolar affective polarization Chile, Italy, Portugal and Spain by using an original survey from the TRI- POL project (Torcal et al., 2023). The dataset includes, for each selected Social identity theory (Tajfel and Turner, 1979), on which most of country, feelings scales towards voters of the main parties, identification the affective polarization literature is based, constitutes a comprehen­ scales towards ideological labels and policy-issue positions, as well as sive and integrated analysis of the relationship between self-concept and some relevant control variables. The selection of countries allows us to group and intergroup behavior. According to this theory, the origins of explore the relationship between ideological identity, issues and affec­ social identity are driven by both motivational and cognitive factors tive polarization in some rather different institutional and political (Huddy 2001: 132). More concretely, this theory defends that people contexts. have a natural inclination to divide the world into in-groups and This paper contributes to the literature that studies affective polari­ out-groups and are motivated to positively distinguish their group from zation in multiparty systems by showing that ideological identity shapes others, leading to the development of in-group bias, even in the most affective polarization into two opposing ideological poles and that this basic definition of a group (Tajfel et al., 1971). effect is not conditioned by issue-based ideology. This constitutes a By contrast, the so-called self-categorization theory (Turner et al., pernicious form of polarization that, despite the presence of multiple 1987), which derives from the classic social identity theory (Tajfel and parties, divides societies into ‘Us vs. Them’ camps resulting in a single Turner 1979), is basically focused on the cognitive roots of social dimension that overshadows all others (McCoy and Somer 2019). identity. Categorization is the cognitive process by which people classify Furthermore, the results show that bipolar affective polarization exerts objects into groups to make sense of the world (e.g. Allport 1954) little and inconsistent reverse effects on ideological identity, reinforcing through discrimination and generalization (Wedell et al., 2007). the idea that identity-based ideology leads to, rather than follows, the Discrimination facilitates treating objects from different categories affective distance between ideological blocs. differently, whereas generalization allows predicting information of an individual object based on other objects sharing a category. This leads, 2. Theoretical framework thus, to two possible effects of categorization on perceptions and judgements of objects: differences between objects that belong in 2.1. Issue-based vs. identity-based ideology different categories tend to be accentuated (inter-category effect), while differences between objects that belong to the same category tend to be Ideology has been traditionally conceptualized as a ‘belief system’: ‘a overlooked (intra-category effect) (e.g. Goldstone and Hendrickson configuration of ideas and attitudes in which the elements are bound 2010). These two categorization effects are not always found together, together by some form of constraint or functional interdependence’ and in most cases one effect predominates over the other (e.g. McGarty (Converse 2006: 3). This type of ideology is based on principled beliefs and Tuner 1992). and convictions about the proper role of government, constitutes a guide The principles driving the categorization of objects can be extended to interpret and respond to aspects of the political and social environ­ to explain the categorization of people, including oneself, into social ments, and has been shown to be a good predictor of individuals’ issue grouping (e.g. Huddy 2001). Given a certain social categorization, positioning (Popp and Rudolph 2011). Generally, people on the left (or people would tend to accentuate the perceived differences between liberals, in the US) support redistributive social policies and adopt themselves and members of the out-group as well as to increase the progressive stances in cultural issues, while people on the right (or perceived similarity with members of the in-group. The extent to which conservatives) are more reluctant to endorse government-based solu­ a specific social categorization (and associated identity) serves as the tions to socio-economic problems and hold more traditional cultural basis for self-conception and for the perception of others is determined views. This ideology type has been referred in the literature mainly as by the salience of the category in a given context (Turner et al., 1987), ‘operational ideology’ (Free and Cantril 1967) or ‘issue-based ideology’ which, in turn, rests on the notions of accessibility and fit (e.g. (Hogg (Mason 2018a, 2018b). The operationalization of this type of ideology and Smith 2007). That is, people rely on categorizations that are valued, requires measures of opinions and stances towards different salient important and frequently employed and, consequently, use them to policy-issues. It can be conceptualized and measured in two different investigate how well the categorization fits the observed similarities and ways. On the one hand, issue-based ideology is usually understood as the differences among people. extent to which citizens hold extreme opinions on a given set of issues There are two main political categories to which people develop (issue extremity). On the other hand, it is also measured as the extent to affective attachments helping them to organize and understand their 2 J.M. Comellas and M. Torcal Electoral Studies 83 (2023) 102615 political environment: partisanship and ideology (Vegetti and Širinić empirically distinguishable concepts, with the positive correlation be­ 2019). In the context of the US two-party system, partisan and ideo­ tween the two measures being quite weak (see, for example, Mason logical identities overlap to a great extent: being a Democrat is practi­ 2018b for the US, and Oshri et al., 2021 for Israel). There is, accordingly, cally synonymous with being a liberal, while being a Republican is the existence of ‘ideologues without issues’: people who identify almost equivalent to being a conservative (Levendusky 2009). This way, strongly with an ideological label but hold moderate stances on salient partisan identity tends to be the main factor influencing political per­ issues and/or ideologically incongruent positions on some of these is­ ceptions and behaviors (Heit and Nicholson 2010). However, in multi­ sues. Interestingly, ideological identity tends to generate political atti­ party systems, the same ideology usually encompasses different partisan tudes and behaviors purely on the basis of group reasoning. In the categories and some people may not identify as much with a particular bipartisan US context, Mason (2018b) empirically showed that party as with an ideological label (e.g. Garry 2007), constituting a identity-based ideology is associated with affective polarization be­ supra-partisan identity. In fact, it has been traditionally argued that tween liberals and conservatives even when issue-based ideology is ideology is the main driver of party voters’ preferences and behavior in weak or conflicting with ideological identity; that is, ideological iden­ many contemporary democracies, leading to a bipolarization of the tities drive polarization even among those individuals who identify as political conflict in those left-right ideological terms (Thomassen 2005; conservatives but are left-leaning in their policy-issue positions – and Oshri et al., 2021). vice versa. We expect to find the same result for multiparty systems, so Individuals identified with the ‘left’ or ‘right’ labels may tend to the third hypothesis is the following: categorize people as in-groups and out-groups based on their ideological H3. Ideological identity is associated with the affective distance be­ adscription. Moreover, they may accentuate similarities between voters tween ideological blocs even when it is inconsistent with issue-based belonging to the same ideological bloc and differences between sup­ ideology. porters of different ones. As Vegetti and Širinić (2019) have recently showed, when the left-right distinction has a strong identity component 2.3. Reverse effects in a given political environment, individuals are better equipped to discriminate between parties belonging to different ideological blocs So far, we have assumed that ideological identity fuels a bipolar type and tend to perceive parties belonging to the opposite ideological group of affective polarization between left and right blocs. This is congruent as more distant from themselves than they actually are. By contrast, with social identity theory: the mere identification with an in-group citizens do not tend to perceive parties within ideological blocs as more leads individuals to positively distinguish their group from others; uniform. Following these previous results, we expect that individuals and, the greater the partisan identity, the stronger the antipathy towards with high levels of ideological identity tend to express much more the members of the out-group(s) (Huddy 2001; Tajfel and Turner 1979). polarized sentiments towards voters of left-wing parties and supporters However, the affective distance between ideological blocs may also of right-wing ones than those with low levels of ideological identity (or exert reverse effects on identity-based ideology, such that ideological without any); that is, identity-based ideology increases the affective identity and bipolar affective polarization reinforce each other over distance between the supporters of both ideological blocs, leading to the time. This is a relevant question that has been insufficiently explored in conformation of a bipolar type of affective polarization. The first hy­ the literature and even then, as far as we know, only for issue-based pothesis, hence, is as follows: ideology. Specifically, in the US context, Lelkes (2018) explored the H1. Ideological identity is positively associated with the affective possible two-way relationship between issue consistency and affective distance between ideological blocs polarization, showing that there is a reciprocal, albeit weak, relation­ ship. Since identity-based ideology is grounded in the notion that in­ Taking extreme positions on salient policy-issues, and/or holding dividuals develop affective ties to ideological labels, we expect that the ideologically consistent positions on a range of different issues, may also reverse effects of bipolar affective polarization on ideological identity be associated with bipolar affective polarization. We know that extreme are significant and substantive. Therefore, the last hypothesis is the attitudes towards salient policy issues and/or extreme positions on the following: left-right scale are positively associated with animosity between ideo­ logical blocs in multiparty systems (e.g. Huddy et al., 2018; Reiljan and H4. The affective distance between ideological blocs exerts positive Ryan 2021; Wagner 2021). However, we expect this relationship to be reverse effects on ideological identity. significantly weaker than with ideological identity, because issue-based ideology lacks the component of psychological and emotional attach­ 3. Case selection, data and methods ments that allows citizens to classify different objects into clearly defined groups and accentuate differences between them. Recent studies 3.1. Selected countries and parties in fragmented party systems show that citizens’ evaluation of out-parties and their supporters become gradually cooler as the ideological distance We test the different expectations in five multiparty systems of to a partisan group grows, suggesting that ideology understood as a Southern Europe and Latin America with quite different institutional belief system tends to push citizens to evaluate partisans not so much settings: Argentina, Chile, Italy, Portugal and Spain. Previous studies based on ideological blocs, but rather on the extent to which they agree have shown that Southern Europe is a region that contains some of the on specific issues (Harteveld 2021; Van Erkel and Turkenburg 2022). most polarized Western countries, both in ideological and in affective Following this argument, for example, a Social Democrat would express terms. In a recent study, Dalton (2021) empirically showed that Italy, high levels of disgust towards supporters of the radical right, while also Spain and Portugal have, respectively, the 3rd, 4th and 7th most exhibiting mildly cold sentiments towards voters on the radical left and polarized party systems along the left-right ideological divide out of a center-right. Only if this Social Democrat had a strong affective total of 21 Western countries. Congruent with this classification, Gidron attachment to the left, would he express warmer feelings towards et al. (2020) found that Southern Europe is the most affectively polar­ radical-left voters than towards center-right ones, and the affective ized region in the Western world, and Reiljan (2020) showed that it is distance between blocs would increase. Therefore, the second hypoth­ the region presenting the highest levels of partisan antipathy in Europe esis is as follows: along with Central and Eastern Europe. Regarding Latin America, most studies argued that parties tend not H2. Ideological identity is more strongly associated with the affective to differentiate themselves from their rivals in ideological and pro­ distance between ideological blocs than issue-based ideology. grammatic terms, so that voters choose parties and candidates more on Identity and issue-based ideology are two conceptually different and the basis of their personal characteristics than on ideology and issues (e. 3 J.M. Comellas and M. Torcal Electoral Studies 83 (2023) 102615 g. Mainwaring and Torcal 2006; Roberts 2002). However, the levels of right Partido Popular (PP), to the current multiparty system with the party system ideological polarization, and the extent to which voters’ presence of the radical-left party Podemos, the centre-right party Ciu­ ideological positioning predicts their electoral choices, differ across dadanos (Cs) and the radical-right Vox (Rama et al. 2021). This trans­ Latin American countries and within them over time (Moraes 2015; formation has resulted in an inter-bloc ideological confrontation Singer 2016). These analyses reveal that, during the period 1995–2010, between two major groups at the national level: the left, formed by Chile was among the group of countries in the region with Podemos and PSOE, and the right, formed by Cs, PP and Vox (Orriols middle-to-high levels of both elite ideological polarization and left-right and León 2020; Simón 2020). voting, while Argentina was among countries with middle-to-low levels Finally, Portugal has experienced more modest changes in its party in both measures. Some studies show that, congruent with the low levels system (Lisi et al., 2020). The two major traditional parties, the of elite ideological polarization, the meaning attributed to the left and centre-left Partido Socialista (PS) and the centre-right Patido Social right labels by people in Argentina is quite heterogeneous, poorly linked Democrata (PSD), continue to gather the majority of support. To the left to the main parties and not very politically relevant (e.g. Brussino et al., of the PS are two main parties: the communist Coligação Democrática 2016; Zechmeister 2006; Zechmeister and Corral 2013), which suggests Unitária (CDU) and the radical left Bloco de Esquerda (BE). Two small that the saliency of left and right categories (i.e. the extent to which this progressive green parties are also considered in our analysis as categorization serves as the basis for self-conception and for the belonging to the left bloc: Pessoas-Animais-Natureza (PAN) and Livre. To perception of parties and their voters) may be particularly weak in this the right of the PSD, we find its traditional conservative partner, the country compared to the others selected (Hogg and Smith 2007). Centro Democrático e Social-Partido Popular (CDS-PP). Moreover, two In each country, we have selected the main political parties for our small right-wing parties surged in the 2019 legislative elections and empirical analysis. They are classified as being in the ‘left’ or ‘right’ increased their support in the recent 2022 elections: the libertarian ideological blocs according to their average ideological position given Iniciativa Liberal (IL) and the radical-right Chega (Lopes 2023). by the respondents of the survey used in this paper (see Figure A1-A5 in the online Appendix), as well as according to specialized literature cited 3.2. Data and operationalization in the following paragraphs. Table A1 in the online Appendix lists the selected political parties in each country and their location in the left The TRI-POL dataset comprises a three-wave online panel survey and right blocs. conducted in Argentina, Chile, Italy, Portugal and Spain among their The Argentinian party system is currently dominated by two large respective voting age population (Torcal et al., 2023). A non-probability electoral alliances that took most of the support in the last presidential quota sampling method was applied, ensuring that the sample reflects and legislative elections: Frente de Todos (FdT) and Juntos por el Cambio the characteristics of the general population in terms of region of resi­ (JxC). The former brings together different currents of Peronism and is dency, gender and age. To test H1, H2 and H3, we used the first wave of generally associated with a centre-left position, while the latter repre­ the panel, which is the one containing all respondents. In the case of sents the alliance between traditional radicalism and different liberal Argentina, however, the first wave does not include the feeling ther­ sectors and is located on the centre-right (Sendra and Ortiz 2022). To the mometer scale towards the voters of a relevant right-wing party that left of the FdT is the coalition of socialist and Trotskyist parties Frente de emerged very recently (La Libertad Avanza) and, hence, we chose to use Izquierda y de Trabajadores (FIT). At the opposite end of the ideological the third wave.12 To test H4, we take advantage of the panel structure of spectrum, we have the newly founded La Libertad Avanza (LA), a liber­ the data and used all three waves. The first wave was carried out be­ tarian and conservative party that some analysts place close to the tween late September and mid November 2021; the second between radical right (Cruz Olmeda and Soto Licea 2021). early December 2021 and early January 2022; and the third between In Chile, the party system was structured along four main poles in the late March and late April 2022.3 2021 presidential and legislative elections. First, there were the two To test the different hypotheses, we must first classify the re­ traditional coalitions: a centre-left bloc formed by the Partido Demócrata spondents into left and right ideological blocs. We did this using the Cristiano (PDC), the Partido Socialista (PS), the Partido por la Democracia classic eleven-point ideological self-placement scale, where 0 means (PPD) and the Partido Radical (PR); and a rightist bloc mainly composed extreme left and 10 means extreme right. In particular, those re­ by the Renovación Nacional (RN), the Unión Demócrata Independiente spondents who fall between points 0 and 4 were classified in the left bloc (UDI) and the Evolución Política (Evópoli) (e.g. Alemán et al., 2021; and those who fall between points 6 and 10 were classified in the right Torcal and Mainwaring 2003). To the left and to the right of these bloc. To limit the missing data, respondents who are located right in the traditional blocs were, first, a left-wing coalition formed by the parties center (i.e., in point 5) were classified in one of the blocs according to, and movements of the Frente Amplio (FA) and the Partido Comunista (PC), first, the ideological position of the party with which they identify and, among others; and, second, the new radical-right Partido Republicano second, the ideological affiliation of the party for which they express the (PLR) (Somma and Donoso 2022). intention to vote in the next general elections. The respondents who are Italy has suffered a continuous transformation of its party system located right in the ideological center and who do not express any during this last decade (Emanuele and Chiaramonte 2020). The major identification or intention to vote for one of the main parties selected in party on the left continues to be the centre-left Partito Democratico (PD). To the left of the PD, the most relevant party that gained representation in the 2018 legislative elections was the left-wing Liberi e Uguali (LeU). To the right of the political spectrum, the traditional centre-right party, Forza Italia (FI), was relegated to a secondary position in the 2018 elections by the radical-right Lega. Another radical-right party, Fratelli d’Italia (FdI), has progressively gained relevance until winning the 1 We do not use the second wave because it does not contain policy-issue recent 2022 elections. Two other relevant political parties present an positions. ambiguous ideological position: the populist catch-all party Movimento 5 2 The number of completed interviews in the first wave is of 1337 for Chile, Stelle (M5S) and the socio-liberal Italia Viva (IV) (Berlucchi 2021). Both 1231 for Italy, 1028 for Portugal and 1289 for Spain; in Argentina, 979 re­ parties have been classified in the left bloc following the average spondents completed the third wave. 3 perception of survey respondents. For more information about the dataset, see https://www.upf.edu/web/t Spain has also completed a deep transformation of its party system ri-pol/documentation-and-data. during the last decade, from an imperfect bipartisanship dominated by the centre-left Partido Socialista Obrero Español (PSOE) and the centre- 4 J.M. Comellas and M. Torcal Electoral Studies 83 (2023) 102615 the study are not considered in the present analysis.4 their bloc than with the label of the other bloc; zero signifies that re­ spondents do not identify with either ideological label, or identify 3.2.1. Affective distance between ideological blocs equally with both; and negative values mean that respondents are more The dependent variable, affective distance between ideological blocs attached to the ideological label of the opposed bloc than to that of their (ADBB), is built based on feeling thermometer scales towards the voters own bloc. of the main parties in each country, which range from 0 (negative sen­ Figures A11-A15 display the distribution, mean and standard devi­ timents) to 100 (positive sentiments). The use of feeling scales towards ation of our key independent variable in each selected country. Spain is, voters constitutes a relevant improvement from the most of the again, the country that presents the highest average level of ideological comparative literature that employs feelings towards parties. As has identity (1.60). By contrast, Argentina is the country with the weakest been shown, the polarization of sentiments towards parties tends to mean ideological identity (1.07), which seems congruent with previous overestimate the levels of antipathy between ordinary partisans (e.g. studies. It should be noted that the number of observations with negative Druckman and Levendusky 2019; Knudsen 2021), which is central to the values in this variable is very low, ranging between 2.14% (Chile) and definition of affective polarization. The ADBB was obtained by calcu­ 4.38% (Portugal) of the total number of respondents. lating, first, the mean sentiments towards voters of parties belonging to the left bloc and the right bloc, respectively. Then, we calculated the 3.2.3. Issue-based ideology distance between the mean sentiments towards the voters of their own Issue-based ideology is conceptualized and operationalized in two ideological bloc and the mean sentiments towards the voters of the other different ways: issue extremity and issue consistency (e.g. Mason 2018a, bloc. The resulting variable can range from − 100 to 100. The positive 2018b). We take advantage of the fact that the TRI-POL dataset includes values mean that respondents give better evaluations to the voters of opinion scales on six salient policy issues: state intervention in the parties belonging to their own ideological bloc than to the voters of economy, provision of public services, cultural assimilation of migrants, parties of the other bloc, while negative values mean that respondents desired immigration level, same-sex marriage and abortion rights (see have colder feelings towards supporters of their own bloc than towards exact wording in the online Appendix). Table A7 in the online Appendix supporters of the opposed one. Tables A2-A6 in the online Appendix shows bivariate correlations between each opinion scale and ideological provide actual examples of the ADBB caculcation in each selected self-placement, confirming that issues are related to the left-right country. dimension in the same direction (although with some differences Figures A6-A10 in the online Appendix display the distribution, regarding the strength of the relationship) across our selected countries: mean and standard deviation of our dependent variable in each selected being favorable to state intervention, public services, immigration and country. Spain is the country that presents the highest average ADBB multiculturalism, and defending abortion and same sex marriage is (30.94), closely followed by Argentina (29.45). By contrast, the country linked to the left side of the ideological spectrum, and vice-versa. that clearly presents the lowest mean levels of bipolar polarization is On the one hand, we constructed the issue extremity variable by Portugal (12.35). combining the six policy-issue scales into a composite scale that goes from the most left-wing average positions (0) to the most right-wing 3.2.2. Ideological identity average positions (10) for those respondents who belong to the right Ideological identity is sometimes simply captured using the tradi­ bloc, and from the most right-wing positions (0) to the most left-wing tional self-identification scale, although this measure is quite ambiguous (10) for those who were classified in the left bloc. That is, the higher because it may also partially contain issue positioning and other the values of this variable, the more extreme are the respondents’ instrumental components (Mason 2018a). Therefore, ideological iden­ average positions towards the direction of their bloc on the selected tity is best measured using survey scales that ask about the level of issues. attachment to specific ideological labels. In our case, we use two survey On the other hand, we created each country’s issue consistency questions that ask respondents their level of identification with the variable by, first, calculating the number of issue items that were ideological labels ‘left’ and ‘right’. Specifically, the question is as fol­ answered on the left-leaning end of the spectrum and those on the right- lows: ‘How much do you identify with the following political/­ leaning end. Then, we calculated the difference between the number of ideological labels [Left/Right]?‘. Respondents had four response left-leaning issues and right-leaning issues for those respondents classi­ options: ‘not at all’, ‘a little’, ‘somewhat’ and ‘very much’. We calculated fied according to each ideological bloc. The resulting variable can vary the difference between identification with the label of one’s own ideo­ from − 6 to 6; that is, from those who maintain inconsistent positions in logical bloc and identification with the label of the other bloc. The logic relation to their ideological bloc on all six issues (− 6) – e.g. an individual behind this operationalization is that the ideological identity strength of who self-places on the right but holds left-leaning positions on all six a respondent who uniquely identifies with her ideological label is issues – to those who express a position on all issues that is consistent stronger than the identity strength of a respondent who also exhibits with their ideological bloc (6) – e.g. an individual who self-places on the some degree of identification with the opposite label.5 right and holds right-leaning positions on all issues –, through those who The resulting variable, therefore, ranges from − 3 to 3, where positive hold a centrist position on all issues or the same number of left and right- values mean that respondents identify more with the ideological label of leaning issues (0). Figures A16-A25 in the online Appendix show the distribution, mean and standard deviation of issue extremity and issue consistency in each country. Spain is once again the country with the highest average levels 4 The percentage of respondents located in the center among those who of issue extremity/consistency (6.23/1.67), while Portugal registers the answered the ideological self-placement question is 33.8% in Argentina, 31.6% lowest levels (5.52/0.74). Finally, tables A8-A12 display, for each in Chile, 20.1% in Italy, 29.7% in Portugal and 23.7% in Spain. Once we classify selected country, a correlation matrix between ideological identity, these respondents in one of the blocs according to their party identification and issue extremity and issue consistency. The correlation coefficients be­ vote intention, the percentage of discarded respondents without ideology is tween ideological identity and the two issue-based variables range from substantially reduced: 15.2% in Argentina, 19.5% in Chile, 10.5% in Italy, weak to moderate and never exceed the value of 0.5, which seems to 12.8% in Portugal and 14.8% in Spain. 5 empirically confirm, in line with previous literature, the distinction For example, a respondent who places herself on the left and claims to identify ‘somewhat’ with the left and, at the same time, ‘a little’ with the right, between identity-based and issue-based ideology. The correlation be­ is considered to have a more blurred and weaker identification with her bloc tween issue extremity and issue consistency, by contrast, is very strong than another respondent who shows the same identification strength with the in all countries (between 0.92 and 0.95). left but without any attachment to the right. 5 J.M. Comellas and M. Torcal Electoral Studies 83 (2023) 102615 3.2.4. Control variables 4. Results We have also selected some basic control variables: partisan identi­ fication, gender, age, education level and subjective income (see the Fig. 2 displays the main results of the linear regression models with online Appendix for a detailed description of these variables). ADBB as the dependent variable and ideological identity and issue ex­ tremity as key explanatory factors (see linear regressions in Model 1 of 3.3. Methods Tables A13-A17 in the online Appendix). As can be appreciated in the graphs, H1 is confirmed in all five countries: ideological identity is To test H1, H2 and H3, we implemented two different analyses in positively and significantly associated with bipolar affective polariza­ each selected country: one with issue extremity as the variable tion, controlling for all the other factors. Specifically, ideological iden­ measuring issue-based ideology and the other with issue consistency. In tity increases by one standard deviation the affective distance between particular, we perform linear regression models with ADBB as the the own ideological bloc and the other by 0.44 standard deviations in dependent variable and, as independent variables, ideological identity, Spain, 0.42 in Portugal, 0.38 in Chile, 0.37 in Italy and 0.24 in issue extremity/issue consistency and the aforementioned control vari­ Argentina. ables. All variables are standardized except gender and education levels. Our second hypothesis (H2) is supported by the results in Chile, Italy, As indicated above, these linear regression models were performed in Portugal and Spain: although issue extremity is positively associated the first panel wave, with the exception of Argentina, where we used the with ADBB, its bipolarizing effects are significantly weaker than those third wave. exerted by ideological identity. The difference in the effect strength To test H4, we estimated a three-wave cross-lagged structural between both variables is especially large in Spain, where the effect of equation model to check reciprocal relationships between ADBB and issue-based ideology is the weakest in all five countries: increasing the ideological identity (Finkel 2008). This model specifies a lagged effect average position on the selected issues by one standard deviation to­ from each variable on itself over time (autoregressive effects), and wards the direction of respondents’ ideological bloc fuels the affective cross-lagged effects of bipolar affective polarization and identity-based distance between the own bloc and the other by only 0.12 standard ideology. Given that there are three panel waves, we can relax con­ deviations. The results in Argentina, however, do not support H2, given straints regarding error co-variances between the two analyzed vari­ that the effect of issue extremity is not significantly different from the ables in each wave. We have also included the lagged effects of each one exerted by identity-based ideology. This finding is congruent with variable at t-2 for two main reasons: first, controlling for any possible the low levels of elite ideological polarization found in Argentina ac­ lingering effects and, second, obtaining an acceptable model fit. This cording to Singer (2016: 181), together with the fact that the popular could make it more difficult to find significant and substantive meaning of left-right label has been found to be quite incongruent and cross-lagged effects between the second and the third waves than be­ not often linked to party labels (e.g. Zechmeister 2006). It is also tween the first and the second, since lingering effects are controlled for. congruent with a recent study showing that the main drivers of affective Finally, we also added in the models issue extremity/issue consistency polarization in Argentina are the feelings towards “justicialismo” and the rest of control variables measured in the first wave. We have responding more to the populist/anti-populists conflict (Torcal and standardized the variables. For a graphical representation of the model, Carty 2023). see Fig. 1. This model is estimated in Chile, Italy, Portugal and Spain. Concerning the control variables, partisan identification is positively Unfortunately, we cannot perform this analysis in Argentina due to the associated with bipolar affective polarization in all countries, its effect data limitations described above. being clearly weaker than that exerted by ideological identity (with the Cross-lagged models are estimated as in any structural equation exception of Argentina, where the effect strength of the two variables is model, with the variances and co-variances between observed variables not significantly different). Women tend to be less polarized by ideo­ expressed in terms of the unknown parameters. Under the assumption of logical blocs than men in Portugal and Spain; age is positively associated multivariate normality of the observed variables, the maximum likeli­ with polarization (except in the case of Spain); and education level ex­ hood method is used to estimate the model parameters. Given that the erts positive effects in Chile and Portugal, where subjective income is models are over-identified, we can assess how well the model fits the negatively related to our dependent variable. data as a whole. With the goal of testing model fit, different measures are If we use issue consistency instead of issue extremity to measure used: the Root Mean Squared Error of Approximation (RMSEA), the issue-based ideology, the results are the same as those described above Comparative Fix Index (CFI), the Tucker and Lewis Index (TLI), and the (see Model 2 of Tables A13-A17 and Figure A26). If anything, the effects Standardized Root Mean Square Residual (SRMR). of issue consistency on ADBB are slighly weaker than those exerted by Fig. 1. Cross-lagged structural equation model, 3 waves. Notes: Variables are standardized. ADBB = Affective distance between blocs; IdeolIden = Ideological identity. Control variables are: issue extremity/issue consistency, partisan identity, gender, age, education level and subjective income. 6 J.M. Comellas and M. Torcal Electoral Studies 83 (2023) 102615 Fig. 2. Linear regression models. Affective distance between ideological blocs as dependent variable. Argentina, Chile, Italy, Portugal and Spain. Notes: 95% confidence intervals. All variables except gender and education level are standardized. Source: TRI-POL, wave 1 (wave 3 for Argentina). issue extremity. left-right blocs appears to have little or nothing to do with issue-based Are the bipolarizing effects of ideological identity conditioned by ideology. issue-based ideology? To test H3, we have included in the previous Finally, we explore whether bipolar affective polarization in turn models an interaction term between ideological identity and issue ex­ feeds back into ideological identity. Fig. 4 shows the main results of the tremity. Fig. 3 displays, for all five countries, the marginal effects of three-wave cross-lagged structural equation models that test H4 in identity-based ideology on ADBB by levels of issue extremity (see Model Chile, Italy, Portugal and Spain. They include issue extremity and the 3 of Tables A13-A17). The interaction effects are not singificant at any rest of the control variables measured in the first wave. The fit of the conventional level, which means that respondents with high levels of models is good (see Table A18): the RMSEA is below the standard ideological identity are more biased in favor of their ideological bloc’s threshold of 0.10; the requirement that the CFI and the TLI are greater voters than those with low levels of identity even when their average than 0.90 is met in all countries; finally, the SRMR should not be above positions on the selected six policy-issues are moderate or ideologically 0.08 for model acceptability, a requirement that all our models meet. inconsistent. Only for the case of Spain is the interaction term significant For H4 to be supported, we need to observe not only that ideological at a confidence level of 90% and, here, the direction is surprisingly identity feeds subsequent levels of ADBB, but also that bipolar affective negative: that is, ideological identity seems to have a slightly stronger polarization reinforces psychological ties to one’s own ideological label bipolarizing effect when it is inconsisent with issues. over time. First, as is observed in Fig. 4, an increase in ideological If we perform an interaction effect between ideological identity and identity in a given wave leads to an increase in ADBB in the sebsequent issue consistency, the results remain quite the same (see Model 4 of wave. Although these cross-lagged effects seem weak, especially be­ Tables A13-A17 and Figure A27). The main difference is that the tween the second and third waves, they are substantial when taking into negative interaction term found in Spain in the previous model now account the reduced time interval between the waves, which only allows appears significant at a 95% confidence level, although the strength of a modest variation of the considered variable over time. Nevertheless, the moderating effect is quite weak: the marginal effect of identity-based bipolar affective polarization has a much less consistent relationship ideology on bipolar affective polarization is predicted to be 0.51 stan­ with later ideological identity. The lagged effects of polarization on dard deviations when issue consistency is of − 2 (i.e. respondents identity are significant and quite substantive between the first two maintain two more issues that are inconsistent with their ideological waves (although they are weaker than the cross-lagged effects exerted bloc than consistent with it, representing the 10th percentile of the by ideological identity in Chile and Italy). These effects, however, are variable), while the marginal effect is still 0.38 standard deviations only statistically significant at the 95% confidence level between the when issue consistency is of +5 (i.e. respondents hold five more issues second and third waves – when we control for lingering effects – in one that are consistent with their ideological bloc than inconsistent with it, country: Portugal. Specifically, increasing bipolar affective polarization being the 90th percentile of the variable). by one standard deviation in the second wave leads to an increase of All in all, the results are in line with H3: the capacity of ideological 0.09 standard deviations on ideological identity in the third wave. The identity to polarize sentiments towards party voters into two opposing cross-lagged effects of polarization between the last two waves are 7 J.M. Comellas and M. Torcal Electoral Studies 83 (2023) 102615 Fig. 3. Marginal effects of ideological identity on the affective distance between ideological blocs by levels of issue extremity. Argentina, Chile, Italy, Portugal and Spain Notes: 95% confidence intervals. Standardized coefficients. Source: TRI-POL, wave 1 (wave 3 for Argentina). weaker and only significant at a 90% confidence level in Chile and 4.1. Robustness checks and extensions Spain, while in Italy they are very weak and not significant at any conventional level. Some robustness checks and additional analyses have been imple­ If we control the three-wave cross-lagged models on the basis of issue mented. First, there is the possibility that the results obtained are mainly consistency instead of issue extremity, the lagged effects of ADBB on driven by sentiments towards supporters of small extreme parties. To ideological identity between the last two waves seem to be somewhat test this possibility, we recalculated our measure of ADBB by weighting stronger and are statistically significant at the 95% confidence level in sentiments towards voters by the electoral size of their parties.6 The Chile and Spain, while they remain non-significant in Italy (see Figure A28 and Table A19). All in all, the inconsistency of the results (and their dependency on country, waves and model specification) does 6 not support H4. The findings, hence, reinforce the expectation that could We obtained the electoral size of each party, in the cases of Italy, Portugal be derived from social identity theory that identity-based ideology and Spain, by estimating the (weighted) mean voting intention based on all the electoral polls performed 90 days before the first day of the wave’s fieldwork. precedes and stimulates the affective distance between one’s own For more information, see the Data Protocols of the TRI-POL project ideological bloc and the opposite, more than the other way around. (https://www.upf.edu/web/tri-pol/data-protocols-panel-survey). In the cases of Argentina and Chile, we used the results of the most recent national election. 8 J.M. Comellas and M. Torcal Electoral Studies 83 (2023) 102615 Fig. 4. Cross-lagged structural equation models between affective distance between ideological blocs (ADBB) and ideological identity (IdeolIden), 3 waves. Chile, Italy, Portugal and Spain. Notes: **p < 0.01, *p < 0.05, + p < 0.1. Variables are standardized. ADBB = Affective distance between blocs; IdeolIden = Ideological identity. Control variables, means, variances and co-variances are not reported. Control variables are: issue extremity, partisan identity, gender, age, education level and subjective income. Source: TRI-POL, waves 1, 2 and 3. results are quite similar to those of the main models and the fundamental weaker effect on ADBB, similar in strength to that of issues (Tables A39- conclusions hold (Tables A20-A26 and Figures A29-A34).7 A45 and Figures A43-A48).9 Second, we also checked whether ideological identity and issue- Fifth, it is well known that, in most party systems, political conflict is based ideology has an equal impact on both in-bloc liking and out- structured in different dimensions. We conducted a principal component bloc dislike components of ADBB or they are more strongly associated factor analysis with our six policy-issue variables (see the varimax with one of them. The results show that ideological identity fuels both rotated factor loadings in Tables A46 and A47), and the results show the polarization components, although the impact tends to be stronger over presence of three different factors: one dimension is formed by the out-group dislike, especially in Italy and Portugal. Issue-based ideology economic issues (state intervention and public services), other factor tends to be also associated with both components, but it has no signif­ includes the immigration-related ones (immigration level and multi­ icant effects on in-bloc liking in Italy and on out-bloc dislike in Portugal culturalism), and the last the cultural issues (same-sex marriage and (Tables A27-A31 and Figures A35-A36). abortion)10. We then created a composite index for each of these issue Third, we checked the results when those respondents who self- types and computed the respective extremity and consistency variables placed themselves in the centre of the ideological scale (5) – but were (Figures A49-A53 show box plots for each of these variables by country classified in one of the ideological blocs based on their party identifi­ and ideological bloc). Finally, we tested the effect of each issue type on cation or intention to vote – are not considered in the analyses. The new bipolar affective polarization. The results seem to confirm that, in all models exhibit very similar results to the main ones (Tables A32-A38 countries except Argentina, ideological identity exerts a greater effect and Figures A37-A42).8 than each of the different issues.11 In Italy, immigration extremity has a Fourth, we have operationalized ideological identity, in the main strong impact on ADBB, being this effect only weaker than that of analyses, as the extent to which respondents identify solely with their identity-based ideology at the 90% confidence level. This finding may ideological label. We have argued that this operationalization is reflect the significant strength of the radical right and its cultural agenda convenient because, when an individual shows some degree of attach­ in this country (see Table A48-A52 and Figures A54-A55). ment towards the opposite label, her identification with the own bloc is more diffuse and, hence, weaker than that of another individual who 5. Conclusions only shows attachment towards her ideology. If this type of ideology, by contrast, is measured simply as the degree of identification with one’s In multiparty contexts, citizens usually express positive views own label, one might expect it to have weaker bipolarizing effects, precisely because this measure is not capable of identifying those in­ dividuals with an ambiguous ideological identity. We checked that and 9 the results confirm that ideological identity measured in this way has a Only in Spain is the effect of ideological identity stronger than that of issue- based ideology; in Chile, Italy and Portugal, identity and issues have the same impact; and in Argentina, issue-based ideology fuels polarization to a greater extent than ideological identity. 7 10 The main difference is that the lagged effect of ideological identity on The exception is Spain, where there are only two factors: the first mainly (weighted) bipolar affective polarization loses significance between the second contains the immigration-related issues and the second the cultural (same-sex and the third waves in Portugal. However, the rest of cross-lagged effects marriage and abortion) ones; the economic issues present moderate loadings in exerted by ideological identity in all countries remain significant. both factors. 8 11 The main difference is that the lagged effect of ideological identity on ADBB In the case of Argentina, the effect of ideological identity is weaker than loses significance between the second and the third waves in Chile. that of economy extremity/consistency. 9 J.M. Comellas and M. Torcal Electoral Studies 83 (2023) 102615 towards more than one party and its supporters as well as dislike to­ is operationalized using a survey question on the extent to which the wards several parties and their voters. However, many of them tend to respondent identifies with the ‘Left’ and ‘Right’ ideological labels. end up in a two well confronted blocs (e.g. Kekkonen and Ylä-Anttila Although this is a much better measure than the classic ideological self- 2021; Reiljan and Ryan 2021). This paper contributes to the under­ placement scale to capture ideological identity, it could be improved standing of affective polarization in multiparty systems by exploring the through the development of a multi-item scale similar to the one pro­ extent to which ideological identity, a supra-partisan identity that is posed by Huddy et al. (2015) for expressive partisanship (see the prominently relevant in most countries, has the capacity to polarize multi-item measure of identity-based ideology used by Oshri et al., 2021 sentiments towards party voters into two opposing left-right blocs. Until in Israel). Second, we have shown that when ideological identity is now, different studies located in multiparty systems have explored the measured simply as the degree of attachment to one’s own label effect on affective polarization of extreme attitudes towards salient (without considering whether respondents also identify to some degree policy issues and/or extreme positions on the left-right scale (e.g. Gue­ with the opposite label), it has a similar effect on ADBB to that of issues. des-Neto 2022; Harteveld 2021; Torcal and Comellas 2022; Wagner Future research could further explore how respondents with an ambig­ 2021), but without distinguishing between identity-based and uous ideological identity contribute to deflating the polarization be­ issue-based components of ideology or comparing their ability to pro­ tween blocs. Third, we know from previous research that latent rightist mote a bipolar form of polarization. respondents tend to avoid placing themselves on the ideological spec­ We show that, for the multiparty systems of Argentina, Chile, Italy, trum in countries characterized by the legacy of a right-wing dictator­ Portugal and Spain, identity-based ideology is significantly and strongly ship (e.g. Ames and Smith 2010). Since this legacy is present in our associated with the affective distance between ideological blocs, and cases, we cannot completely rule out the possibility that this response that this association is stronger than those of ADBB with issue extremity bias in ideological self-placement could somehow affect our results, or issue consistency. Furthermore, our analysis reveals that ideological especially with regard to the right-wing bloc. Fourth, the paper does not identity increases bipolar affective polarization even when issue-based go into assessing the possible impact of partisan identity on affective ideology is weak or inconsistent with respondents’ ideology. Identi­ polarization compared to ideological identity, especially with regard to fying with an ideological label, therefore, appears to be a crucial factor feelings towards voters of one’s own bloc: while identity-based ideology promoting a bipolar type of affective polarization that has usually been may emphasize similarities between supporters of different parties associated with rather negative social and political consequences (e.g. belonging to the same bloc, partisan identity may fuel differences within McCoy and Somer 2019). All these findings are in line with those found the own bloc (e.g. Huddy et al., 2018). More research in this direction in the US context (Mason 2018b), and seem congruent with the capacity would be appropriate. of ideological identity to promote political engagement in Israel’s Fifth, different studies argue and provide empirical evidence that multiparty system (Oshri et al., 2021). The only partial exception is the some European party systems are increasingly structured around three case of Argentina, where the bipolarizing effects of ideological identity political spaces: the traditional left, the traditional right and a new are weaker and similar to those exerted by issue-based ideology. These populist or anti-establishment pole, which in Europe tends to be repre­ results can be explained by the lower political salience of left-right labels sented by the radical right (e.g. Kriesi et al., 2008). Future research in this Latin American country (e.g. Brussino et al., 2016; Zechmeister could explore the extent to which individuals develop affective attach­ 2006). ments towards this ‘third’ pole and how it, together with left and right Finally, we also explore the possibility that the affective distance identities, may impact the configuration of affective blocs in multiparty between ideological blocs in turn feeds back into ideological identity. systems. And, finally, the Argentinian case reminds us that left and right The results of the implemented cross-lagged models are robust regarding labels do not always constitute the most adequate political identities in the capacity of identity-based ideology to increase subsequent levels of structuring party systems (Torcal and Carty 2023), especially outside the bipolar affective polarization, but are more inconsistent in relation to context of Western industrialized nations. In fact, we do not argue that the lagged effects of polarization on ideological identity. These findings the left-right identity conflict should be the main driver of affective reinforce the argument congruent with social identity theory that the polarization resulting in a two-bloc conflict in all democracies. There direction of the explored relationship mainly goes from group identity to could be other alternative conflicts generating the same results, such as group bias or antipathy. Future research could test whether the reverse the populist/anti-populists one in countries in which one of the main effects of affective polarization on ideological identity are conditioned parties is a promoter of a populist discourse. by some contextual or individual-level factor. For example, a possible Declaration of competing interest research hypothesis is that polarization fuels identity-based ideology None. over time only among those with high levels of education or political sophistication. Funding The main implication of the results presented in the paper is that the tendency of voters in multiparty systems to divide into two affective left- The research was supported by the Spanish Ministry of Science and right blocs is not so much due to policy disagreements as it is a simple Innovation (project reference: PID2019-106867RB-I00/AEI/10.13039/ question of identity. Citizens who are strongly affectively attached to an 501100011033), the BBVA Foundation (Grants for Scientific Research ideological label tend to see themselves as belonging to delimited groups Teams in Economy and Digital Society 2019) and the European Union- opposed to each other and, therefore, are less able to achieve stable NextGenerationEU / Recovery, Transformation and Resilience Plan / compromises around policy agendas even if they do in fact agree on UPF complementary call for requalification - Margarita Salas Grants. particular policy issues (e.g. Mason 2018b). This disturbing dynamic can Mariano Torcal wants to thank the ICREA-ACADEMIA Intense Research represent a major obstacle to the implementation of policies necessary to Award for the additional funding. face the great economic, social and environmental challenges that we face as human beings. We therefore need to continue exploring possible Data availability measures to attenuate this division based on affective ties with ideo­ logical blocs, such as placing greater emphasis on specific policy issues The TRI-POL dataset used in this paper is available at https://osf. in the debates by political parties and the media (e.g. Miller 2020), or io/3t7jz/. For more details on this dataset, see Torcal et al. (2023). promoting consensual institutions that push parties to break political Replication syntax for the analyses reported in the paper is available at blocs and achieve broad agreements that include a greater number of osf.io/uqxtf. Further inquiries can be directed to the authors. actors (e.g. Bernaerts et al., 2022). This study also presents some limitations. First, ideological identity 10 J.M. Comellas and M. Torcal Electoral Studies 83 (2023) 102615 Acknowledgments Huddy, L., Bankert, A., Davies, C., 2018. Expressive versus instrumental partisanship in multiparty European systems. Adv.Polit. Psychol. 39 (S1), 173–199. Huddy, L., Mason, L., Aarøe, L., 2015. Expressive partisanship: campaign involvement, We are very thankful to Laia Balcells for inviting us to present a political emotion, and partisan identity. Am. Polit. Sci. Rev. 109 (1), 1–17. preliminary version of this paper at a panel on polarization at the CES Iyengar, S., Lelkes, Y., Levendusky, M., Malhotra, N., Westwood, S.J., 2019. The origins International Conference of Europeanists in 2022, as well as to Markus and consequences of affective polarization in the United States. Annu. Rev. Polit. Sci. 22, 129–146. Wagner for his excellent comments and suggestion as a discussant on the Iyengar, S., Sood, G., Lelkes, Y., 2012. Affect, not ideology: a social identity perspective paper. We are also especially grateful to Eelco Harteveld and Luana on polarization. Publ. Opin. Q. 76 (3), 405–431. Russo for giving us the opportunity to participate in this special issue on Iyengar, S., Westwood, S.J., 2015. Fear and loathing across party lines: new evidence on group polarization. Am. J. Polit. Sci. 59 (3), 690–707. affective polarization. Finally, we would also like to thank the anony­ Kekkonen, A., Suuronen, A., Kawecki, D., Strandberg, K., 2022. Puzzles in affective mous reviewers for their helpful comments. polarization research: party attitudes, partisan social distance, and multiple party identification. Front. Political Sci. https://doi.org/10.3389/fpos.2022.920567 online. Appendix A. Supplementary data Kekkonen, A., Ylä-Anttila, T., 2021. Affective blocs: understanding affective polarization in multiparty systems. Elect. Stud. 72. Article 102367. Supplementary data to this article can be found online at https://doi. Knudsen, E., 2021. Affective polarization in multiparty systems? Comparing affective polarization towards voters and parties in Norway and the United States. Scand. org/10.1016/j.electstud.2023.102615. Polit. Stud. 44 (1), 34–44. Kriesi, H., Grande, E., Lachat, R., Dolezal, M., Bornschier, S., Frey, T., 2008. West References European Politics in the Age of Globalization. Cambridge University Press, Cambridge. Lelkes, Y., 2021. Policy over party: comparing the effects of candidate ideology and party Alemán, E., Cabezas, J.M., Calvo, E., 2021. Coalition incentives and party bias in Chile. on affective polarization. Polit. Sci. Res. Methods 9 (1), 189–196. Elect. Stud. 72. Article 102362. Lelkes, Y., 2018. Affective polarization and ideological sorting: a reciprocal, albeit weak, Allport, G.W., 1954. The Nature of Prejudice. Addison-Wesley, Cambridge. relationship. Forum 16 (1), 67–79. Ames, B., Smith, A.E., 2010. Knowing left from right: ideological identification in Brazil, Levendusky, M., 2009. The Partisan Sort: How Liberals Became Democrats and 2002-2006. J. Polit. Lat Am. 2 (3), 3–38. Conservative Became Republicans. University of Chicago Press, Chicago. Berlucchi, A.B., 2021. Populism without host ideologies: a new home for voters with Levitin, T.E., Miller, W.E., 1979. Ideological interpretations of presidential elections. Am. exclusionary attitudes in Italy’s Five Star Movement? Party Polit. 28 (5), 811–825. Polit. Sci. Rev. 73 (3), 751–771. Bernaerts, K., Blanckaert, B., Caluwaerts, D., 2022. Institutional design and polarization. Lisi, M., Sanches, E.R., Dos Santos Maia, J., 2020. Party system renewal or business as Do consensus democracies fare better in fighting polarization than majoritarian usual? Continuity and change in post-bailout Portugal. S. Eur. Soc. Polit. 25 (2), democracies? Democratization, online. https://doi.org/10.1080/ 179–203. 13510347.2022.2117300. Lopes, H.F., 2023. An unexpected Socialist majority: the 2022 Portuguese general Bougher, L.D., 2017. The correlates of discord: identity, issue alignment, and political elections. West Eur. Polit. 46 (2), 437–450. hostility in polarized America. Polit. Behav. 39, 731–762. Mainwaring, S., Torcal, M., 2006. Party system institutionalization and party system Boxell, L., Gentzkow, M., Shapiro, J.M., 2022. Cross-country trends in affective theory after the third wave of democratization. In: Katz, R.S., Crotty, W. (Eds.), polarization. Rev. Econ. Stat. https://doi.org/10.1162/rest_a_01160 online. Handbook of Party Politics. SAGE Publications Ltd, California, pp. 204–227. Brussino, S.A., Imhoff, D.S., Paz García, A.P., Dreizik, M.A., Rabbia, H.H., 2016. What are Malka, A., Lelkes, Y., 2010. More than ideology: conservative-liberal identity and left and right in Argentina? Cognitive squemas of citizens from Cordoba. Trends in receptivity to political cues. Soc. Justice Res. 23, 156–188. Psychology 24 (4), 1249–1264. Mason, L., 2016. A cross-cutting calm: how social sorting drives affective polarization. Conover, P.J., Feldman, S., 1981. The origins and meaning of liberal/conservative self- Publ. Opin. Q. 80 (S1), 351–377. identifications. Am. J. Polit. Sci. 25 (4), 617–645. Mason, L., 2018a. Uncivil Agreement: How Politics Became Out Identity. University of Converse, P.E., 2006. The nature of belief systems in mass publics (1964). Crit. Rev. 18, Chicago Press, Chicago. 1–74. Mason, L., 2018b. Ideologues without issues: the polarizing consequences of ideological Cruz Olmeda, J., Soto Licea, J.C., 2021. Las elecciones primarias en Argentina en 2021: identities. Publ. Opin. Q. 82 (S1), 866–887. ¿anticipo de un cambio de rumbo? Elecciones (julio-diciembre) 20 (22), 375–388. McCoy, J., Somer, M., 2019. Towards a theory of pernicious polarization and hot it https://doi.org/10.53557/Elecciones.2021.v20n22.11. harms democracies: comparative evidences and possible remedies. Ann. Am. Acad. Dalton, R.J., 2021. Modeling ideological polarization in democratic party systems. Elect. Polit. Soc. Sci. 681 (1), 234–271. Stud. 72. Article 102346. McGarty, C., Tuner, J.C., 1992. The effects of categorization on social judgement. Br. J. Devine, C.J., 2014. Ideological social identity: psychological attachment to ideological Soc. Psychol. 31, 253–268. in-groups as a political phenomenon and a behavioral influence. Polit. Behav. 37, Miller, L., 2020. Polarización en España: más divididos por ideología e identidad que por 1–27. políticas públicas”, vol. 18. EsadeEcPol Insight. Available online at: https://dobetter. Dias, N., Lelkes, Y., 2022. The nature of affective polarization: disentangling policy esade.edu/es/polarizacion-espana. disagreement from partisan identity. Am. J. Polit. Sci. 66 (3), 775–790. Moraes, J.A., 2015. The Electoral Basis of Ideological Polarization in Latin America”, vol. Druckman, J.N., Levendusky, M.S., 2019. What do we measure when we measure 403. Kellogg Institute for International Studies Working Papers. Available online at: affective polarization? Publ. Opin. Q. 83 (1), 114–122. https://kellogg.nd.edu/documents/1721. Ellis, C., Stimson, J.A., 2009. Symbolic ideology in the American electorate. Elect. Stud. Orriols, L., León, S., 2020. Looking for Affective Polarisation in Spain: PSOE and 28 (3), 388–402. Podemos from Conflict to Coalition. S. Eur. Soc. Polit. 25 (3–4), 351–379. Emanuele, V., Chiaramonte, A., 2020. Going out of the ordinary. The de- Oshri, O., Yair, O., Huddy, L., 2021. The Importance of Attachment to an Ideological institutionalization of the Italian party system in comparative perspective. Group in Multi-Party Systems: Evidence from Israel”. Party Politics. https://doi.org/ Contemporary Italian Polit. 12 (1), 4–22. 10.1177/13540688211044475 online. Finkel, S.E., 2008. Linear panel analysis. In: Menard, S. (Ed.), Handbook of Longitudinal Popp, E., Rudolph, T.J., 2011. The tale of two ideologies: explaining public support for Research. Elsevier Press, New York, pp. 475–504. economic interventions. J. Polit. 73 (3), 808–820. Free, L.A., Cantril, H., 1967. The Political Beliefs of Americans: A Study of Public Rama, J., Cordero, G., Zagórski, P., 2021. Three is a crowd? Podemos, Ciudadanos, and Opinion. Rutgers University Press, New Brunswick, NJ. Vox: the end of bipartisanship in Spain. Front. Political Sci. https://doi.org/ Garry, J., 2007. Making ‘party identification’ more versatile: operationalising the 10.3389/fpos.2021.688130 online. concept for the multiparty setting. Elect. Stud. 26 (2), 346–358. Reiljan, A., 2020. Fear and loathing across party lines’ (also) in Europe: affective Garzia, D., Ferreira da Silva, F., Maye, S., 2023. Affective polarization in comparative polarisation in European party systems. Eur. J. Polit. Res. 59, 376–396. and longitudinal perspective. Publ. Opin. Q. nfad004, 1–13. Reiljan, A., Ryan, A., 2021. Ideological tripolarization, partisan tribalism and Gidron, N., Adams, J., Horne, W., 2020. American Affective Polarization In Comparative institutional trust: the foundations of affective polarization in the Swedish Perspective (Elements in American Politics). Cambridge University Press, Cambridge. multiparty system. Scand. Polit. Stud. 44 (2), 195–219. Goldstone, R.L., Hendrickson, A.T., 2010. Categorical perception. Wiley Interdisciplinary Roberts, K.M., 2002. Social inequalities without class cleavages in Latin America’s Reviews: Cognit. Sci. 1 (1), 69–78. neoliberal era. Stud. Comp. Int. Dev. 36 (4), 3–33. Guedes-Neto, J.V., 2022. The effects of political attitudes on affective polarization: Sendra, M., Ortiz, P., 2022. Can Argentines Be Programmatic Voters? Assessing the survey evidence from 165 elections. Polit. Stud. Rev. https://doi.org/10.1177/ Influence of Policy Preferences in Argentine Coalitions Using the Propensity to 14789299211067376 online. Vote”. Revista de Ciencia Política. https://doi.org/10.4067/s0718- Harteveld, E., 2021. Fragmented foes: affective polarization in the multiparty context of 090x2022005000123 online. The Netherlands. Elect. Stud. 71. Article 102332. Simón, P., 2020. The multiple Spanish elections of April and may 2019: the impact of Heit, E., Nicholson, S.P., 2010. The opposite of republican: polarization and political territorial and left-right polarisation. S. Eur. Soc. Polit. 25 (3–4), 441–474. categorization. Cognit. Sci. 34, 1503–1516. Singer, M., 2016. Elite polarization and the electoral impact of left-right placements: Hogg, M.A., Smith, J.R., 2007. Attitudes in social context: a social identity perspective. evidence from Latin America, 1995-2009. Lat. Am. Res. Rev. 51 (2), 174–194. Eur. Rev. Soc. Psychol. 18 (1), 89–131. Somma, N.M., Donoso, S., 2022. Renewing the political arena: social uprising, Huddy, L., 2001. From social to political identity: a critical examination of social identity constitutional change and Chile’s new government. Revista Mexicana de Política theory. Polit. Psychol. 22 (1), 127–156. Exterior 122, 29–42. 11 J.M. Comellas and M. Torcal Electoral Studies 83 (2023) 102615 Tajfel, H., Billig, M.G., Bundy, R.P., Flament, C., 1971. Social categorization and Van Erkel, P.F.A., Turkenburg, E., 2022. Delving into the divide: how ideological intergroup behaviour. Eur. J. Soc. Psychol. 1 (2), 149–178. differences fuel out-party hostility in a multi-party context. Eur. Polit. Sci. Rev. 14, Tajfel, H., Turner, J., 1979. An integrative theory of intergroup conflict. In: Austin, W.G., 386–402. Worchel, S. (Eds.), The Social Psychology of Intergroup Relations. Brooks/Cole, Vegetti, F., Širinić, D., 2019. Left-right categorization and perceptions of party Monterey, pp. 33–47. ideologies. Polit. Behav. 41, 257–280. Thomassen, J. (Ed.), 2005. The European Voter: a Comparative Study of Modern Wagner, M., 2021. Affective polarization in multiparty systems. Elect. Stud. 69. Article Democracies. Oxford University Press, Oxford. 102199. Torcal, M., Carty, E., 2023. Populismo, ideología y polarización afectiva en Argentina”. Webster, S.W., Abramowitz, A.I., 2017. The ideological foundations of affective Revista Argentina de Ciencia Política, Forthcoming. polarization in the U.S. Electorate. Am. Polit. Res. 45 (4), 621–647. Torcal, M., Carty, E., Comellas, J. M., Bosch, O. J., Thomson, Z. A., Serani, D., 2023. The Wedell, D.H., Hicklin, S.M., Smarandescu, L.O., 2007. Contrasting models of assimilaton Dynamics of Political Polarisation: A Dataset For Spain, Portugal, Italy, Argentina, and contrast. In: Stapel, D.A., Suls, J. (Eds.), Assimilation and Contrast in Social and Chile (2019-2022). Available at SSRN: https://ssrn.com/abstract=4377503 or Psychology. Psychology Press, New York, pp. 45–74. https://doi.org/10.2139/ssrn.4377503. Zechmeister, E., 2006. What’s left and who’s right? A Q-method study of individual and Torcal, M., Comellas, J.M., 2022. Affective polarisation in times of political instability contextual influences on the meaning of ideological labels. Polit. Behav. 28, and conflict. Spain from a comparative perspective. S. Eur. Soc. Polit. 27 (1), 1–26. 151–173. Torcal, M., Mainwaring, S., 2003. The political recrafting of social bases of party Zechmeister, E.J., Corral, M., 2013. Individual and contextual constraints on ideological competition: Chile, 1973-95. Br. J. Polit. Sci. 33 (1), 55–84. labels in Latin America. Comp. Polit. Stud. 46 (6), 675–701. Turner, J.C., Hogg, M.A., Oakes, P.J., Reicher, S.D., Wetherell, M.S., 1987. Rediscovering the Social Group: A Self-Categorization Theory. Blackwell, Oxford. 12