This article investigates how multiple forms of perceived threat map onto horizontal affective polarization (operationalized as social distance toward supporters of rival candidates) in Argentina and Brazil. These nations, mirroring trends observed in other Latin American democracies, have experienced a significant escalation in political polarization in recent decades (Sarsfield, Moncagatta, and Roberts Reference Sarsfield, Moncagatta and Roberts2024). The electoral success of right-wing populist leaders Javier Milei and Jair Bolsonaro, known for their adversarial rhetoric, is often argued to have intensified existing partisan divisions and amplified perceptions of reciprocal threat (Brusco Reference Brusco2023; Sarsfield, Moncagatta, and Roberts Reference Sarsfield, Moncagatta and Roberts2024). This is important because perceptions of threat and, especially, of intergroup threat may represent an important source of affective polarization at the societal level (Stephan, Ybarra, and Rios Reference Stephan, Ybarra, Rios and Nelson2009; Renström, Bäck, and Carroll Reference Renström, Bäck and Carroll2021, Reference Renström, Bäck and Carroll2023; Renström and Bäck Reference Renström, Bäck, Torcal and Harteveld2025).
Affective polarization reflects a partisan bias, wherein individuals hold positive views of their political co-partisans and negative views of opposing partisans (Iyengar and Westwood Reference Iyengar and Westwood2015: 691). Our study focuses on the role that various perceived intergroup threats play in this phenomenon. While political and social identities can foster political antipathy (Tajfel and Turner Reference Tajfel, Turner, Worchel and Austin1986), outgroup derogation typically requires specific aggravating conditions, such as the attribution of threat to other groups (Stephan, Ybarra, and Rios Reference Stephan, Ybarra, Rios and Nelson2009). However, despite recognizing the role that threat perception plays in negative outgroup attitudes, affective polarization research often fails to incorporate direct measures of potential threats. Thus, we still know very little about which specific threats correlate with affective polarization for different political groups and whether this relationship is similar for supporters of the right-wing and left-wing camps.
In this article, we use original survey data from Argentina and Brazil to analyze the link between specific perceived intergroup threats and negative attitudes toward rival presidential candidate supporters. Our findings support the hypothesis that perceptions of threat from various social and political groups are associated with outgroup distance (a measure of negative affect toward supporters of rival candidates). In Brazil, this relationship differs significantly between supporters of the right-wing populist candidate, Bolsonaro, and supporters of his left-wing counterpart, Luiz Inácio ‘Lula’ da Silva. These groups are distinct not only in the magnitude of the relationship between threat perception and outgroup distance (which is significantly stronger among right-wing supporters) but also in the specific sources and domains of the perceived threats that are related to it. This is consistent with research in political psychology developed primarily in the United States, which indicates that conservatives respond more acutely to perceived threats by increasing their animosity toward outgroups. In contrast, while differences in the specific sources and domains of perceived threats are also observed between the left and right camps in Argentina, differences in the magnitude of the relationship do not reach statistical significance, highlighting the importance of accounting for contextual variation in research on this topic.
Our research contributes to the existing literature in two main ways. First and foremost, it shows that directly measuring perceived threats allows for a more comprehensive understanding of affective polarization dynamics than traditional models based solely on political identity. Second, it tests whether threat-based mechanisms identified in Western contexts generalize to democracies characterized by weak party-system institutionalization, fluid electoral alliances, and the presence of successful political outsiders.
This is important because our understanding of affective polarization dynamics remains predominantly informed by research on the United States and European democracies, where politics is more strongly structured by policy issues and traditional partisan identities. Given the prevailing view that affective polarization stems primarily from partisan identities (Iyengar, Sood, and Lelkes Reference Iyengar, Sood and Lelkes2012), this underscores the necessity to test the validity of our hypotheses in contexts beyond advanced post-industrial democracies where positive party identification tends to be weaker. To the best of our knowledge, this represents the first study to directly examine the relationship between outgroup threat and affective polarization within Latin American democracies. While recent studies on Latin America provide valuable insights into this phenomenon (e.g., Areal Reference Areal2022; Brusco Reference Brusco2023; Comellas and Torcal Reference Comellas and Torcal2023), significant gaps remain, particularly regarding the study of interpersonal expressions of affective polarization, rather than attitudes directed toward elites.
Threats and affective polarization: a theoretical framework
Although affective polarization research has increasingly turned to the role of threat perception, important questions remain about how different types of threats are related to outgroup animosity across diverse contexts. In this regard, this article makes two primary contributions to the literature. First, it examines how different types of perceived threat are associated with animosity toward distinct political outgroups and how these associations vary between right- and left-wing supporters. While existing research recognizes the importance of threat perceptions in shaping affective polarization, it rarely distinguishes between their different forms or assesses how they map onto hostility toward specific political outgroups. To cover this gap, we analyze multiple types of perceived threat and their relationship with affective polarization, comparing these patterns across ideological camps.
Second, we test the applicability of the theory in contexts of weak party institutionalization. In doing so, we move beyond the United States and Western Europe, which most of the literature has focused on. By focusing on Brazil and Argentina – two democracies characterized by fluid and weak positive partisanship – we explicitly address the call by Renström, Bäck, and Carroll (Reference Renström, Bäck and Carroll2021) to consider how contextual features shape threat perceptions. Indeed, these cases provide a more rigorous test of the theory, because if threat perceptions are associated with greater outgroup distance even where partisan identity is weak, it suggests that the psychological link between threat and polarization is a fundamental mechanism that transcends institutionalized party systems.
The remainder of this section unfolds in three parts. First, we critically evaluate the dominant approaches to affective polarization in established Western democracies, while also discussing their applicability to Brazil and Argentina. Next, we substantiate the added theoretical value of threat perception when studying affective polarization. Finally, we provide a brief overview of the political contexts of contemporary Argentina and Brazil and present our hypotheses concerning the relationship between threat perception and outgroup distance in these two countries.
Affective polarization research
The literature on affective polarization in the United States and other Western democracies predominantly features two explanatory frameworks. The first framework highlights policy disagreement as the main source of polarization (Rogowski and Sutherland Reference Rogowski and Sutherland2016; Webster and Abramowitz Reference Webster and Abramowitz2017). It argues that negative attitudes toward opposing partisans stem from policy extremity (individuals with extreme policy views exhibit greater hostility toward rival parties and their supporters) and policy alignment (individuals whose policy positions are strongly aligned with those of the party they support display higher levels of polarization than those with inconsistent views). However, empirical research, particularly in the United States, has consistently found that policy disagreement plays only a secondary role in explaining affective polarization compared to social identity dynamics (Mason Reference Mason2018). If this holds true in stable Western democracies, it is even more likely to be the case in countries characterized by weakly institutionalized party systems and prevalent personalistic voting, where programmatic or ideological ties between parties and voters tend to be weak. In fact, cross-national studies reflect this variance. Areal (Reference Areal2022), for instance, finds that policy and ideology play a limited role in driving affective polarization in Brazil, although Comellas and Torcal (Reference Comellas and Torcal2023) report a somewhat stronger association in Argentina.
A second explanatory framework, which is nowadays prevalent in the literature, argues that partisanship is the main driver of affective polarization. This approach views partisanship as a social identity that shapes perceptions, fostering both ingroup favoritism and outgroup hostility (Iyengar, Sood, and Lelkes Reference Iyengar, Sood and Lelkes2012, Reference Iyengar, Lelkes, Levendusky, Malhotra and Westwood2019; Mason, Reference Mason2018). Affective polarization is thus seen as stemming from social stereotyping (Rothschild, Adam, Howat et al. Reference Rothschild, Adam, Howat, Shafranek, Busby and Howat2019) and ultimately from the social identity processes described by Tajfel and Turner (Reference Tajfel, Turner, Austin and Worchel1979) in their social identity theory.
At first sight, this latter perspective may also seem difficult to apply to countries like Argentina and Brazil, where few citizens identify with major parties and partisan loyalties are often fluid (Lupu Reference Lupu, Carlin, Singer and Zechmeister2015; Baker, Ames, Sokhey et al. Reference Baker, Ames, Sokhey and Renno2016). Although expressive party identification, understood as an emotional attachment reinforced by social affiliations, is a well-documented driver of emotional reactions (Huddy, Mason, and Aarøe Reference Huddy, Mason and Aarøe2015), some scholars have claimed that the low levels of partisanship, whether expressive or instrumental (Huddy, Bankert, and Davies Reference Huddy, Bankert and Davies2018), observed in Latin America stem from the methodological limitations of direct party identification questions, which fail to accurately capture true partisan loyalty in contexts where public expressions of partisanship are stigmatized (Meléndez Reference Meléndez2022). In contrast, others, such as Areal (Reference Areal2022), acknowledge the low levels of positive partisanship but put forward alternative explanations. He suggests that, in countries like Brazil, partisan identities are not necessarily defined by a positive attachment to a particular party but tend to emerge from individuals defining themselves by what they oppose rather than what they support. This process is conceptualized in the literature as negational categorization (Zhong, Phillips, Leonardelli et al. Reference Zhong, Phillips, Leonardelli and Galinsky2008) or outgroup categorization (Leonardelli and Toh Reference Leonardelli and Toh2015). Research on negative partisanship shows that individuals can form political identities primarily based on animosity toward an outgroup, specifically rival parties, often resulting in significant political consequences (Samuels and Zucco Reference Samuels and Zucco2018; Druckman, Klar, Krupnikov et al. Reference Druckman, Klar, Krupnikov, Levendusky and Barry Ryan2024; Russo and Areal Reference Russo, Areal, Torcal and Harteveld2024). Thus, while individuals can sometimes define their ingroup as people who identify with a specific party – a notion aligned with expressive party identification, where partisanship serves as a social identity (Huddy, Mason, and Aarøe Reference Huddy, Mason and Aarøe2015) – in other cases, the ingroup is less formally defined and may comprise all those who share an opposition to another party or candidate. Indeed, negative partisanship is now regarded as one of the most influential attitudinal factors shaping contemporary political orientations and behavior across various contexts (Medeiros and Noël Reference Medeiros and Noël2014; Caruana, McGregor, and Stephenson Reference Caruana, McGregor and Stephenson2015; Mayer Reference Mayer2017; Samuels and Zucco Reference Samuels and Zucco2018; Meléndez and Kaltwasser Reference Meléndez and Rovira-Kaltwasser2021; Bankert Reference Bankert2022; Garzia and Ferreira da Silva Reference Garzia and Ferreira da Silva2022).
Beyond party identification, scholars within this strand have more recently explored how other social identities, such as left-right ideological identities, can shape affective polarization between different political blocs (Bantel Reference Bantel2023). Although Latin American political elites make extensive use of left-right labels, often with negative connotations, and the left-right dimension remains central to party competition in the region (Kitschelt, Hawkins, Luna et al. Reference Kitschelt, Hawkins, Luna, Rosas and Zechmeister2010), numerous studies have argued that citizens in many Latin American countries do not consistently apply ideological labels to themselves in ways that align with their policy preferences or voting behavior (Harbers, de Vries, Steenbergen Reference Harbers, de Vries and Steenbergen2012; Zechmeister and Corral Reference Zechmeister and Corral2013; Oliveira and Turgeon Reference Oliveira and Turgeon2015; Saiegh Reference Saiegh2015; Zechmeister Reference Zechmeister, Carlin, Singer and Zechmeister2015). Nevertheless, these results have been recently nuanced, as new evidence highlights the growing importance of the left-right divide as a sociopolitical identity in the continent (see Comellas and Torcal Reference Comellas and Torcal2023; Torcal and Carty Reference Torcal and Carty2023; Sarsfield, Moncagatta, Roberts et al. Reference Sarsfield, Moncagatta and Roberts2024).
Taken together, the key features of Argentina and Brazil highlighted in this section make them theoretically important cases. Because existing research on affective polarization has paid much more attention to contexts with strong and stable partisan identities, examining these dynamics in Argentina and Brazil allows us to assess whether mechanisms identified in those settings also operate under very different political conditions. This article, therefore, treats these contexts not merely as empirical settings but as a critical test of whether existing theories travel beyond stable party systems. In the next section, we turn to threat-based explanations of affective polarization and consider whether these mechanisms, which have been tested almost exclusively in advanced democracies, should also be expected to operate in weakly institutionalized party systems, providing a more comprehensive approach for the study of affective polarization, which can easily travel to a diverse range of democracies.
Threat perception and political outgroup derogation
Existing perspectives on affective polarization often assume that the mere presence of strong group identities (or, alternatively, substantial policy differences) is sufficient to provoke outgroup dislike. This contrasts sharply with established findings in the social psychology literature, which has consistently highlighted the crucial role played by threat perception as a primary driver of outgroup prejudice and animosity.
The proposition that social identities precede affective polarization is firmly grounded in social psychology, particularly self-categorization theory, which posits that social stereotypes and ingroup bias arise from the cognitive process of categorizing oneself and others into ingroups and outgroups (Turner Reference Turner1987). However, the existence of social identities is widely considered to be a necessary but not sufficient condition for outgroup derogation. As Mummendey and Otten (Reference Mummendey and Otten1998) argue, the active imposition of negative consequences upon outgroups requires the presence of specific aggravating conditions. In other words, while individuals may naturally favor their ingroup, this inclination does not generate outgroup animosity without the presence of further instigating factors.
According to intergroup threat theory, negative outcomes in intergroup relations occur when individuals perceive an outgroup as a threat to core dimensions of their social identity. These dimensions include not only resources and physical safety (realistic threat) but also values, norms, and beliefs (symbolic threat) (Stephan, Ybarra, and Rios Reference Stephan, Ybarra, Rios and Nelson2009). In essence, the perception of a threat to one’s identity triggers negative cognitive, emotional, and behavioral responses toward the outgroup. Consequently, animosity toward political outgroups is expected to be more pronounced when individuals perceive supporters of the opposing political camp as a source of threat.
Research from advanced post-industrial democracies offers strong support for the validity of this theoretical framework in accounting for affective polarization. For instance, studies in Italy have shown a strong correlation between perceived threat and emotional bias toward political outgroups (Schepisi, Porciello, Bufalari et al. Reference Schepisi, Porciello, Bufalari, Aglioti and Panasiti2019). Individuals in Sweden and Germany have been found to distance themselves from opposing party supporters and react with anger when perceiving a threat to their ingroup (Renström, Bäck, and Carroll Reference Renström, Bäck and Carroll2023). Similarly, a comparative analysis of Israel and South Korea has linked gaps in the perception of external threats to the country with different degrees of affective polarization (Harel, Nir, Vandermeulen et al. Reference Harel, Nir, Vandermeulen, Maoz and Halperin2025). Research on the United States and other Western democracies suggests that political actors often contribute to heightening intergroup differences by derogating outgroups and portraying them as threats, thereby fostering polarization (Rosenfeld Reference Rosenfeld2018; Harteveld, Mendoza, and Rooduijn Reference Harteveld, Mendoza and Rooduijn2022). As we show in the next section, political actors in both Argentina and Brazil exhibit these dynamics as well.
It is important to note that the applicability of this theoretical model is not confined to democracies with stable positive party identities, which have hitherto been the primary focus of investigation. Nevertheless, as research on the relationship between threat perception and affective polarization is extremely scarce outside of the United States and Europe, the generalizability of its empirical findings remains unclear.
Theoretically, perceived threats can also arise in the context of outgroup-only categorization (i.e., negational categorization or negative identities), where individuals perceive others as an outgroup due to shared differences and the ingroup does not conform to a unified category based on, for example, a common partisan identity (Leonardelli and Toh Reference Leonardelli and Toh2015; for the role of partisan identities see Samuels and Zucco Reference Samuels and Zucco2018). Therefore, it is reasonable to expect that this model’s explanatory power extends to Argentina and Brazil, where threat perceptions are still likely to play a very significant role in driving affective polarization despite the existence of weak political attachments. By applying this framework to these contexts, our research design provides a rigorous test of whether these mechanisms are also present in less institutionalized party systems.
Threat relevance
So far, the literature has paid limited attention to the types of intergroup threat associated with affective polarization in the political domain (but see Kesberg, Feddes, Vogel et al. Reference Kesberg, Feddes, Vogel and Rutjens2025). Most studies drawing from integrated threat theory tend to focus on reactions directed specifically toward the source of the perceived threat. However, research on affective polarization has occasionally revealed a disparity between the perceived source of a threat (e.g., ethnic minorities) and the political outgroup targeted by those who perceive that threat (e.g., supporters of opposing parties). Renström, Bäck, and Carroll (Reference Renström, Bäck and Carroll2021), for instance, found a significant correlation between perceived threats from refugees and animosity toward opposing party supporters in Europe. Similarly, heightened perceptions of immigrant threats have been shown to increase affective polarization toward opposing party supporters among far-right supporters, with anger acting as a mediating factor (Renström, Bäck, and Carroll Reference Renström, Bäck and Carroll2023). Analyses of polarization in the United States have also highlighted the significant role of perceived threats among certain groups, particularly conservative whites, who perceive the empowerment of ethnic minorities and Black people as a threat to traditional values. This perception has not only contributed to the rise of Republican voting in Southern states, but it has also exacerbated partisan hostility, conflict, and division across the country (Valentino and Sears Reference Valentino and Sears2005; Abramowitz and McCoy Reference Abramowitz and McCoy2019).
Although the affective polarization literature has not explicitly theorized this link, the discrepancy between the source of a threat and the target of affective polarization is consistent with the phenomenon of secondary transfer effects. As described in the intergroup contact literature, this phenomenon occurs when prejudice directed toward a specific group generalizes to other groups that are perceived as similar (Pettigrew Reference Pettigrew2009; Meleady and Forder Reference Meleady and Forder2019). The underlying psychological mechanism lies in the metacontrast principle, a core element of classical self-categorization theory (Turner Reference Turner1981; Oakes, Turner, and Haslam Reference Oakes, Turner and Haslam1991). This principle suggests that individuals establish and adjust group boundaries by evaluating perceived differences and similarities among people on certain dimensions of comparison. Individuals perceived as sharing more in common are categorized as part of the same group. Moreover, broader group categories can encompass multiple subgroups considered to share common attributes, depending on the chosen dimension of comparison.
Therefore, when politics serves as the salient dimension of comparison, political divisions will influence which social groups are perceived as a unified political outgroup. For instance, when far-right supporters perceive a threat to their core identity and values attributed to ‘the left’, they may reconstruct or re-signify this category to fit a blame narrative that extends beyond left-wing party supporters to include social groups like immigrants, ethnic minorities, feminists, etc. Consequently, threats emanating from any of these constituent groups can elicit negative reactions not only toward the members of the specific group in question but also toward members of other groups deemed to be affiliated with the overarching category of ‘the left’.
Different political groups are also likely to respond to different types of threats, with the source and content of those perceived threats playing a key role in those responses. In the United States, conservative individuals tend to construe illegal immigration and gun control as primary threats, while liberal individuals are more prone to reporting fears related to police brutality, climate change, and overpopulation (Jost, Stern, Rule et al. Reference Jost, Stern, Rule and Sterling2017). Hibbing (Reference Hibbing2020) argues that, while conservatives and liberals are equally likely to perceive different situations as threatening, conservatives are more likely to feel the need to be vigilant against outsider threats. Although the alignment between different political groups and threats is likely dependent on the political and social context, Brandt, Turner-Zwinkels, Karapirinler et al.’s (Reference Brandt, Turner-Zwinkels, Karapirinler, Van Leeuwen, Bender, van Osch and Adams2021) comparative study shows evidence that economic threats tend to be associated with left-wing beliefs, whereas the perceived threat of social violence is associated with right-wing cultural beliefs in most (although not all) of the countries they study. In a similar vein, Renström, Bäck, and Carroll (Reference Renström, Bäck and Carroll2021) find that perceiving refugees as a threat is only related to affective polarization for right-wing individuals in Sweden, whereas the threat of political extremism is only associated with affective polarization among those on the left. Nevertheless, as the authors also highlight, ‘[c]ontextual features may clearly matter for understanding why certain societal challenges are perceived as threats and ultimately influence affective polarization in a society’ (Renström, Bäck, and Carroll Reference Renström, Bäck and Carroll2021: 569). This call is explicitly addressed in this article by focusing on the understudied contexts of Brazil and Argentina.
While existing research has shown that perceived threats can heighten horizontal affective polarization, it has typically examined a narrow set of threat sources. What remains less understood is how multiple types of threats simultaneously shape hostility toward political outgroups. By examining a broader range of threat domains, this research aims to extend existing work by assessing how diverse threat perceptions map onto horizontal affective polarization in contexts where partisan identities are weak and often negative rather than positive.
Threat and political conservatism
It has been widely assumed that individuals from different political camps respond similarly to threat perceptions, differing only in the type of threat they respond to. However, a number of scholars posit that reactions to threats are not uniform across individuals of different political affiliations. In particular, conservative individuals have been argued to exhibit a greater propensity to respond to negative stimuli and to react more strongly to perceived threats than their left-wing counterparts (Hibbing, Smith, and Alford Reference Hibbing, Smith and Alford2014). The psychological mechanisms explaining this greater propensity to respond to threats are thought to be rooted in the psychological basis of conservatism, which prioritizes order, stability, uncertainty avoidance, and threat minimization (Jost, Glaser, Kruglanski et al. Reference Jost, Glaser, Kruglanski and Sulloway2003, Reference Jost, Stern, Rule and Sterling2017; Hibbing, Smith, and Alford Reference Hibbing, Smith and Alford2014). For Wilson (Reference Wilson1973: 259), conservative attitudes can be understood as a reaction to a ‘generalized susceptibility to experiencing threat or anxiety in the face of uncertainty’. Because of their need for threat control, conservatives tend to display a higher readiness to dislike groups that are seen as threatening to social cohesion, social order, security, and stability (Duckitt Reference Duckitt2001). Thus, they are more likely than other individuals to perceive actions that challenge social conformity (including beliefs, values, or behavior that are inconsistent with social conventions or the government’s ability to enforce social rules) as a threat, with any groups identified as the source of such a threat becoming a target for action (Feldman Reference Feldman2003). These claims are further supported by findings linking childhood anxiety to the development of conservative ideologies (Fraley, Griffin, Belsky et al. Reference Fraley, Griffin, Belsky and Roisman2012). They are also consistent with Altemeyer (Reference Altemeyer1998), who found that individuals scoring high in Right-Wing Authoritarianism (RWA) tend to be more likely to perceive the world as a threatening and dangerous place, thereby justifying the need for society to defend itself from such forces.
A growing body of research has provided support for the hypothesis that individuals with right-wing authoritarian and/or conservative ideologies and party affiliations tend to be more responsive to threats (Hibbing, Smith, and Alford Reference Hibbing, Smith and Alford2014). For instance, Lavine, Burgess, Snyder et al. (Reference Lavine, Burgess, Snyder, Transue, Sullivan, Haney and Wagner1999) found that high authoritarians are more responsive to threat messages than low authoritarians are. Right-wing authoritarians also exhibit more outgroup prejudice when they perceive other groups as threatening to their customs, values, and traditions (Haddock, Zanna, and Esses Reference Haddock, Zanna and Esses1993). Focusing more specifically on political conservatism, studies have shown that conservatives tend to exhibit greater physiological responses (such as skin conductance) than other individuals when exposed to a threatening stimulus (Oxley, Smith, Alford et al. Reference Oxley, Smith, Alford, Hibbing, Miller, Scalora, Hatemi and Hibbing2008). In a similar vein, Dodd, O’Daniel, and Cusato (Reference Dodd, Balzer, Jacobs, Gruszczynski, Smith and Hibbing2012) found that, besides presenting greater physiological responses to threat, individuals on the political right also paid greater attention to the threat than those on the political left. Nash and Leota (Reference Nash and Leota2022), on the other hand, conducted two electroencephalogram (EEG) studies where they found heightened responses under threat among self-reported conservatives. This is consistent with research utilizing magnetic resonance imaging (MRI) scans, such as Pedersen, Muftuler, and Larson (Reference Pedersen, Muftuler and Larson2018), who found self-reported conservatives to show greater reactivity in the neural threat circuitry of the brain while under threat. Although most of these studies were conducted in America, Osmundsen et al. (Reference Osmundsen, Hendry, Laustsen, Smith and Petersen2022) conducted a study in both Denmark and the USA and found that, in both samples, conservatives reported stronger emotional reactions to threatening images.
Despite this evidence, the assertion that conservatives respond more strongly to the perception of threat is not embraced by all scholars. Crawford (Reference Crawford2017), for instance, argues that, while conservatives react more strongly to physical threats, both individuals on the left and those on the right are similarly influenced by meaning threats (challenges to individuals’ perception that the world is ordered and intelligible, disrupting the psychological foundations through which people make sense of their environment). Moreover, some empirical studies that have attempted to replicate the findings of Oxley, Smith, Alford et al. (Reference Oxley, Smith, Alford, Hibbing, Miller, Scalora, Hatemi and Hibbing2008) and Dodd, Balzer, Jacobs et al. (Reference Dodd, Balzer, Jacobs, Gruszczynski, Smith and Hibbing2012) have not found significant differences between conservatives and those on the left in how they react to threat in the USA (e.g., Knoll, O’Daniel, and Cusato Reference Knoll, O’Daniel and Cusato2015; Arceneaux, Dunaway, and Soroka Reference Arceneaux, Dunaway and Soroka2018; Bakker, Schumacher, Gothreau et al. Reference Bakker, Schumacher, Gothreau and Arceneaux2020) and the Netherlands (Bakker et al. Reference Bakker, Schumacher, Gothreau and Arceneaux2020). Thus, the question of whether threat is equally associated with affective polarization for both the right and the left political camps is open to further research. Furthermore, it is possible that these mechanisms are conditioned by specific political contexts, such as the specific elite-driven mobilization strategies deployed by right- and left-wing parties. Thus, the assumption that conservatives are more responsive to threat perceptions may not apply to all democracies.
The politics of division: Argentina and Brazil’s polarizing context
Levels of polarization in both Brazil and Argentina are on par with those seen in other Latin American nations, where recent studies have identified a clear upward trend (Sarsfield, Moncagatta, and Roberts Reference Sarsfield, Moncagatta and Roberts2024; Bergman and Fernández Reference Bergman and Fernández2025). Both countries nonetheless illustrate how populist leaders use derogatory rhetoric targeting various outgroups, emphasizing the symbolic threats they are portrayed as posing, reinforcing intergroup divisions, and thereby intensifying polarization.
In Brazil, Jair Bolsonaro’s 2018 presidential victory was linked to high preexisting polarization, declining trust in traditional parties, and economic recession (Hunter and Power Reference Hunter and Power2019; Samuels, Mello, and Zucco Reference Samuels, Mello and Zucco2024). Once in office, Bolsonaro promoted ‘freeing Brazil from socialism’ and ‘inverted values’, defending ‘the family’ and ‘Judeo-Christian tradition’ and combating ‘gender ideology’ (Phillips Reference Phillips2019). His administration was marked by antagonistic rhetoric, particularly toward ‘the left’, which he dehumanized, claiming they ‘do not deserve to be treated like normal people’ (Correio Braziliense 2020). Bolsonaro’s ultranationalist discourse also targeted groups such as the Lesbian, Gay, Transgender, Bisexual and Queer (LGTBQ) community, refugees, and migrants, calling the latter ‘the world’s filth’ coming to Brazil and adding, ‘like we don’t already have enough problems to fix’ (Quixabeira Reference Quixabeira2018). His government framed migrants as threats to Brazil’s cultural identity and security, adopting restrictive immigration policies and withdrawing from the UN Global Compact for Migration (Fellet Reference Fellet2019; Piauí Reference Piauí2019; Filomeno and Vicino Reference Filomeno and Vicino2021). His efforts to overturn the 2022 election further deepened divisions, and although he was banned from public office until 2030 and condemned to 27 years in prison, he continues to exert political influence (Pina Reference Pina2025).
Argentina has also experienced growing polarization, culminating in the 2023 election of right-wing populist Javier Milei (Brusco Reference Brusco2023). Milei has intensified divisions by insulting opponents as ‘filth’, ‘enemies’, and ‘rats’ (Clarin 2024; La Nacion 2024). An admirer of Bolsonaro’s ‘fight against the international left’ (Moraes Moura Reference Moraes Moura2023), he frequently uses the term zurdos (‘leftists’; literally ‘left-handed’) with derogatory adjectives and profanity, portraying them as seeking to ‘pollute life’, ‘destroy society’, and as morally corrupt (López Marín Reference López Marín2023). Milei has also attacked feminists, the LGTBQ community, immigrants, and other minorities, even equating gay people with ‘paedophiles’ (Tode Jimenez Reference Tode Jimenez2025; World Economic Forum 2025). His government has tightened asylum protocols and proposed new restrictive immigration measures (Centenera Reference Centenera2024a, Reference Centenera2024b). Spokesperson Manuel Adorni stated that ‘miscreants, squatters, and opportunists should remain on the other side of the border and not come to Argentina to harm its citizens’ (Centenera Reference Centenera2024a), while Security Minister Patricia Bullrich has adopted a similarly hardline stance. This rhetoric has been linked to a rise in xenophobic violence (Ferrer Reference Ferrer2024). In a context of increasing migration fluxes, recent research shows a positive correlation between right-wing authoritarianism and prejudice toward other Latin American immigrants in Argentina (Etchezahar, Ungaretti, and Marchiano Reference Etchezahar, Ungaretti and Marchiano2022).
While Milei’s and Bolsonaro’s divisive discourse has arguably fueled polarization, opposition leaders have also contributed to escalating tensions. In Argentina, Cristina Fernández, a leading Peronist and 2022 assassination attempt survivor, referred to Milei and his team as a ‘buffoon’ (mamarracho) and a ‘shameless lot’ (caraduras), claiming, ‘My children and I are being persecuted; they are refusing to investigate those who wanted to assassinate me (…) Am I afraid? I’m afraid that my grandchildren will grow up in such an unfair country, so unequal’ (TN, 2023). More recently, she accused Milei of causing ‘pointless sacrifice’, saying, ‘Your government is a failure and you, as president, are utterly shameful’ (Associated Press 2024). In Brazil, Fernando Haddad labeled Bolsonaro a ‘barbarian’ endorsed by the Ku Klux Klan (Phillips Reference Phillips2018), and during the 2022 debates, Lula da Silva called Bolsonaro a ‘little dictator’ and ‘the king of fake news and stupidity’, and accused him of ‘destroying the country’ (BBC 2022; Vasconcelos Reference Vasconcelos2022).
Hypotheses
Drawing upon the theoretical considerations discussed in the previous sections, we posit that affective polarization is exacerbated when perceived threats target dimensions central to group identity and/or originate from sources identified as political outgroups. Consequently, we hypothesize that affective polarization in Brazil and Argentina will correlate positively with perceived threat levels. Moreover, we anticipate that distinct threat types will differentially polarize supporters of each political camp. Bolsonaro and Milei supporters are expected to be polarized by perceived threats from left-wing supporters, along with threats from other outgroups (e.g., immigrants) and to core values such as national identity. Similarly, Lula and Massa supporters are expected to be polarized by perceived threats from right-wing supporters, along with threats from other outgroups (e.g., exclusionary groups such as racist people) and to core values such as social equality.
These initial hypotheses assume a uniform polarizing effect of perceived threats across all political groups, conditional only on the relevance of the threat. However, as we explained earlier, there is a growing body of work in psychology that argues that right-wing individuals are more responsive to threat perceptions than those on the left. Accordingly, we will test an alternative hypothesis that predicts a significantly stronger association between threat perception and outgroup distance for Bolsonaro and Milei voters compared to those who supported Lula and Massa.
Data and operationalization
We evaluate our hypotheses using original survey data collected in Argentina and Brazil between the 30th of November and the beginning of December 2023.Footnote 1 These nations present a compelling case for comparative analysis due to their shared political context, characterized by a history of relatively recent military dictatorships, similar political systems (presidential federal republics with two-round presidential elections and a bicameral Congress), and the presence of populist leaders (Valenzuela Reference Valenzuela1993). Moreover, both countries exhibit pronounced socioeconomic inequalities, as reflected in their 2022 Gini coefficients (Argentina: 0.41, Brazil: 0.52) and societal poverty rates (Argentina: 22.84%, Brazil: 27.15%) (World Bank 2025). The timing of data collection, following the election of Milei in Argentina (October 22, 2023) and the electoral defeat of Bolsonaro in Brazil (October 2, 2022), enables us to examine our hypotheses in two neighboring countries with presidents from opposing ideological camps.
Although both Argentina and Brazil display sharp political divisions, we did not select both cases under the premise that they both share similarly high levels of affective polarization. Using data from the Comparative Study of Electoral Systems (2023), Bergman and Fernández (Reference Bergman and Fernández2025) show that polarization toward political elites is higher in Argentina than in Brazil; in fact, even after Bolsonaro’s 2018 election, affective polarization in Brazil remained relatively low compared to other Latin American countries. Nevertheless, as we discuss below, we find comparable levels of social distance in both countries.
Our study employed online surveys conducted in Spanish (Argentina) and Portuguese (Brazil) by YouGov, with quota sampling for gender, age (interlocked), education, and region.Footnote 2 Ethical approval for the surveys was granted by the Research Ethics Committee at the Universidad Nacional de Educación a Distancia (UNED) (project code: 2018-T1/SOC-10152).
Initial data collection resulted in 1,506 Argentine and 1,504 Brazilian respondents. As our study centers on affective polarization between supporters of the two candidates that competed in the second round of the 2022 presidential election in Brazil and the 2023 presidential election in Argentina, the effective sample sizes after excluding non-voters (but not accounting for missing data) were 1,050 in Argentina and 1,054 in Brazil.
In Argentina, the presidential candidates were the right-wing populist Javier Milei, from the electoral alliance La Libertad Avanza/Freedom Advances, and the left-wing Peronist Sergio Massa, from the electoral alliance Unión Por la Patria/Union for the Homeland. In Brazil, the candidates were the left-wing Luiz Inácio (‘Lula’) Da Silva, from the electoral alliance Brasil da Esperança/Brazil of Hope, and Jair Bolsonaro, from the electoral alliance Pelo Bem do Brasil/For the Good of Brazil.
To analyze the data, we use linear regression models with post-stratification weights adjusting for level of education, region of residence, and the cross-classification of age and gender. The original questionnaire wordings in Spanish and Portuguese can be found in the online Supplemental Materials.
As this is an observational study, we are unable to make causal claims regarding the relationships examined. While we can assess whether affective polarization is associated with the perception of various threats, whether these associations are consistent across opposing ideological blocs, and whether they persist after controlling for plausible confounding variables, it is important to bear in mind that our data does not allow us to establish a clear causal direction.
Dependent variable
Our dependent variable is an indicator of social distance, which was operationalized by asking respondents how comfortable they would feel with having supporters of the out-candidate as neighbors, with responses ranging from very comfortable (1) to very uncomfortable (5).Footnote 3 Latin Americans are known to engage in political discussions with their neighbors more frequently than individuals in most other countries (Baker, Ames, and Rennó Reference Baker, Ames and Rennó2020: 32). Moreover, evidence from Brazil indicates that these everyday conversations serve as a significant channel for political information sharing (Baker, Ames, and Renno Reference Baker, Ames and Renno2006).
The study focuses on social distance toward the outgroup due to its established use in social psychology as a measure of behavioral intentions to discriminate (e.g., Conner, Abusuampeh, Kosyluk et al. Reference Conner, Abusuampeh, Kosyluk, Tran, Davis-Cotton, Hill and Brown2024). Social distance is argued to provide a more proximal measure of attitudes related to behavioral outcomes than other measures of affective polarization such as feeling thermometers, trust assessments, and outgroup trait ratings, which tend to assess broader, more abstract outgroup evaluations (Druckman and Levendusky Reference Druckman and Levendusky2019). Thus, they are arguably one step closer than other measures to some of the most normatively undesirable behavioral consequences of polarization, such as intolerance, social avoidance, and prejudice. It is important to acknowledge that all measures of affective polarization have faced criticism, and social distance is no exception. For example, some scholars have noted that these measures may be influenced by individual-level traits, such as a general tendency toward conflict avoidance (Wagner Reference Wagner2024). Nonetheless, empirical research has demonstrated substantial intercorrelation between social distance and other affective polarization measures (Druckman and Levendusky Reference Druckman and Levendusky2019), suggesting that it remains a valid indicator of outgroup attitudes.
Descriptive statistics for this variable (as well as the other variables used in the study) can be found in the Supplemental Materials (Tables 1a and 1.1.a). The social distance indicator has a moderate negative skew (Brazil: −0.860, Argentina: −0.286). As we further discuss in the Results section, the majority of respondents in both countries expressed discomfort with having supporters of the rival candidate as neighbors.
Independent variables
Threat
Threat perceptions were measured using six statements that cover various threat types, differing in both the source and content of the perceived threat. Sources included political outgroups (‘I feel threatened by left-wing people’/‘I feel threatened by right-wing people’) and social outgroups (‘I feel threatened by immigrants’/‘I feel threatened by racist people’). Contents encompassed national identity (‘I feel my national identity is under threat’) and social equality (‘I feel threatened by rising inequality’), each assessed on a 5-point agreement scale (1 = Strongly disagree, 5 = Strongly agree). This battery follows Hibbing (Reference Hibbing2020). For similar wordings, see also Schmid and Muldoon (Reference Schmid and Muldoon2015), Schepisi, Porciello, Bufalari et al. (Reference Schepisi, Porciello, Bufalari, Aglioti and Panasiti2019), and Roth et al. (Reference Roth, Steinmann, Loughnane, Devine, Muldoon, Shelly, van Tilburg, Steffens and Campbell2025).
We acknowledge that these measures may capture both general perceptions of threat to the ingroup and threat to the self within the context of the intergroup relations under study. Empirical research has shown that when individuals are exposed to messages framing a threat against their ingroup, they are likely to internalize this as a personal threat – a relationship that is even more pronounced among those with stronger group identities (De Hoog Reference De Hoog2013). Moreover, while these questions capture broad threat perceptions, they are not designed to identify the specific emotions felt by individual respondents, which remains beyond the scope of this study.
These specific threats were selected because of their potential relevance in the political contexts under study. The threats were evenly divided into those that were expected to affect political groups on the left (right-wing people, racist people, and rising inequality) and on the right (left-wing people, immigrants, and national identity).
In the analysis, we first examine the threat items separately. After assessing that the relevance of specific threats differs across political camps, we performed principal component analysis to derive composite indices separating threat perceptions into two distinct components: a left-wing threat index and a right-wing threat index. Further details regarding this procedure can be found in the Results section.
Across the full samples in both Argentina and Brazil, the composite threat indices exhibited very low levels of skewness (coefficient < 0.5), suggesting a near-symmetric distribution (see Supplemental Materials). Many of the individual threat items display moderate levels of skewness, though the absolute value of the Fisher-Pearson coefficient never exceeds one. Specifically, moderate positive skewness (between 0.5 and 1) was found for the threat posed by immigrants (in both countries) and the threats from right-wing people, left-wing people, and national identity (all in Brazil only). These positive coefficients indicate that many respondents declared low levels of perceived threat from these sources. Conversely, moderate negative skewness was found for the threat posed by inequality in both countries. The item concerning racist people remained near-symmetric in both samples. These distributions tend to change in the political subsamples. For example, the skewness for the threat from left-wing people moves toward near-symmetry among right-wing voters (Bolsonaro: −0.09; Milei: 0.11). Similarly, the threat from right-wing people becomes near-symmetric for left-wing voters (Lula: 0.043; Massa: −0.030).
Controls
The following potential confounders are introduced in the models:
Social sorting is defined as the extent to which an individual’s vote choice can be correctly predicted from their sociodemographic characteristics. Following Harteveld (Reference Harteveld2021), we operationalized this by, first, regressing vote choice in the second round of the presidential elections in each country on gender, age, education, religious affiliation, religiosity, household income, region, and urbanization level of the place of residence. We then obtained the residuals and computed a social sorting score for each individual, which is one minus the absolute residual.
Ideological sorting, the alignment between left-right ideology and vote choice, was measured using a similar approach to social sorting. However, instead of sociodemographic characteristics, we used ideological self-placement (an 11-point scale from left to right) as the predictor.
Issue-based sorting, defined as the alignment between policy positions and vote choice (Mason Reference Mason2018), was operationalized using the same inductive approach as social and ideological sorting. In this case, we predicted support for the chosen candidate based on respondents’ positions on the following issues, all measured on a 5-point agree-disagree scale: redistribution (‘the government should take measures to reduce wealth inequality’), social liberalism (‘when jobs are scarce, men should take priority over women’ and ‘homosexual couples should have the same right to adopt children as heterosexual couples’), environment (‘fighting climate change should be a priority for the government’), and immigration (‘my country would be stronger if we stopped immigration’, ‘immigrants steal jobs from real Argentines/Brazilians’, and ‘immigrants take social services away from true Argentines/Brazilians’)
Ideological extremity. Individuals with extreme ideological identities are likely to exhibit more polarization, as well as react more strongly to perceived threats. We measured this variable by calculating the absolute value of the difference between respondents’ self-placement on the eleven-point left-right scale and the scale’s midpoint (5).
Issue extremity. To construct an indicator of issue extremity, we recoded the issue position items to a consistent left/liberal (1) to right/conservative (5) scale, calculated individuals’ mean position across all items, subtracted the midpoint (3), and took the absolute value of that difference.
Negative political identity. To account for the role of negational partisan identities, we included a question asking respondents: ‘Which party or alliance, if any, do you feel most removed from?’. Respondents were then asked about each of the parties and alliances that stood candidates in at least one round of the presidential election. We used their answers to create a dichotomous indicator of whether respondents mentioned the party of the out-candidate (1) or not (0). This formulation is aimed at capturing the focal point of a respondent’s negative partisanship with a single item, but we acknowledge that responses may reflect both a latent negative identity and substantive concerns, such as specific policy disagreements. This is, therefore, a less granular measure than the multi-item batteries typically employed in studies where negative partisanship is the primary object of inquiry (e.g., Bankert Reference Bankert2022).
In-party support. As previously noted, direct measures of positive party identification in Latin America are often criticized for their vulnerability to social desirability bias (Meléndez Reference Meléndez2022). Additionally, Brazil and Argentina present further complexities due to the presence of highly volatile electoral alliances, each involving multiple parties, and two political outsiders. Milei’s Libertarian Party had never exceeded 1.55% of the vote prior to 2023 and was, therefore, virtually unknown to voters before that year’s presidential election campaign. Meanwhile, Bolsonaro left his former party only a year after his election. He then repurposed a small existing party, the then-centrist Liberal Party, as his electoral vehicle a few months before the 2022 presidential election. To circumvent these methodological challenges, we focus on the intensity of respondents’ stated political preferences. This was operationalized by asking respondents: ‘On a scale where 0 means “I am absolutely certain I would never vote for this candidate or alliance” and 10 means “I am absolutely certain I would always vote for this candidate or alliance”, please indicate the likelihood of you voting for one of the following candidates or alliances’. Respondents were asked to rate each of the electoral alliances that participated in the presidential election, including those eliminated in the first round. In-party support was then operationalized as the respondent’s rating of the electoral alliance for which they cast their vote in the election runoff. This measure likely captures various aspects of political support, including positive partisanship, candidate attachment, and other considerations that increase respondents’ proximity to parties (De Angelis and Garzia Reference De Angelis and Garzia2013).
Results
Figure 1 illustrates the reported social distance toward the ingroup, represented by supporters of the in-candidate, and the outgroup, represented by supporters of the out-candidate, in Argentina and Brazil. As can be seen, both samples exhibit remarkably similar levels of negativity directed toward the outgroup, irrespective of the respondents’ political alignment. In Argentina, supporters of Massa appear to express slightly reduced levels of polarization relative to supporters of Milei; however, the observed differences between these two groups do not achieve statistical significance at p < 0.05. This is consistent with Comellas and Torcal (Reference Comellas Bonsfills and Torcal Loriente2026), who show that, in Spain, left-wing supporters exhibit levels of affective polarization toward the radical right that are comparable to the levels of animosity directed by radical-right supporters toward the left. In both Argentina and Brazil, the mean social distance toward the outgroup is approximately 4 on the 1–5 scale, which corresponds to feelings of discomfort with having supporters of the opposing candidate as neighbors. While Brazil appears to show a marginally greater disparity in social distance between the ingroup and the outgroup compared to Argentina, the differences between both countries are really small.
Social distance toward in-candidate and out-candidate supporters in Argentina and Brazil (with 95% confidence intervals).
Threat and social distance
Figures 2 and 3 present coefficient plots from ordinary least squares (OLS) models illustrating the relationships between social distance toward out-candidate supporters and each of our indicators of perceived threat, as well as the coefficients for each of the variables measuring threat perception after adding controls (the complete results are shown in the Supplemental Materials, Tables 3a, 4a, 5a, and 6a). We use these models to assess the extent to which the different types of threat are relevant for different groups of voters.
Social distance toward out-candidate supporters and different types of perceived threat in Argentina.
Social distance toward out-candidate supporters and different types of perceived threats in Brazil.
Figure 3. Long description
The image contains two side-by-side dot plots comparing perceived threats among Lula voters and Bolsonaro voters. Each plot has a vertical axis labeled ‘Type of Threat’ with categories: Left-wing people, Immigrants, National identity, Right-wing people, Racist people, and Rising inequality. The horizontal axis ranges from -0.2 to 0.4. Each plot includes two sets of data points: one without controls (black dots) and one with controls (gray dots). The plots show how different types of threats are perceived by voters of each candidate. For Lula voters, the perceived threats vary across categories, with some threats showing positive values and others negative. For Bolsonaro voters, the perceived threats also vary, with some categories showing higher positive values compared to Lula voters. The plots highlight the differences in perceived threats between the two voter groups.
First, we focus on the coefficients from the bivariate models to ascertain if there is a relationship between different perceived threats and social distance toward out-candidate supporters. As can be seen, feeling threatened by left-wing people and immigrants, and perceiving national identity to be under threat, are associated with increased social distance toward opposing political supporters among right-wing voters (Milei/Bolsonaro voters). The magnitude of the coefficient for immigrant threat is seemingly smaller in Brazil than in Argentina, but it nonetheless retains statistical significance (p < 0.05). In contrast, perceived threats from right-wing people, racist people, and increasing inequality are not significantly associated with greater outgroup distance among right-wing voters. In Brazil, Bolsonaro supporters who feel less threatened by right-wing people display greater social distance toward the outgroup (p < 0.05).
The findings for Lula and Massa voters mirror those for right-wing voters in the bivariate models. Consistent with initial expectations, the perceived threats pertinent to this specific group are identified as right-wing individuals, racist people, and rising inequality. Nevertheless, while there is a positive association between the perception of rising inequality as a threat and social distance in both countries, the coefficient does not achieve statistical significance for Lula supporters in Brazil (at p < 0.05). This might be explained by the substantial political heterogeneity of Lula’s broad anti-Bolsonaro coalition. Polarization among significant segments of this group may have been rooted in perceived threats to other salient values that were not captured in the study. Lastly, none of the perceived threats that exhibit relevance for the right-wing voter group demonstrate a corresponding relevance for left-wing voters. Again, the only partial exception is Brazil, where Lula supporters who perceive a threat from left-wing people display lower social distance toward Bolsonaro supporters (p < 0.05).
These findings offer initial support for two of our proposed hypotheses: firstly, that perceived threats correlate with increased social distance in both national contexts, and secondly, that different political camps are polarized around different threats.
In a second step, control variables were introduced in the models (the coefficients for these are also displayed in Figures 2 and 3). As many of our controls measure potential antecedents of threat perception and social distance, they may account for individual variation in the propensity to perceive threats and therefore absorb the effect of our threat variables. Following the inclusion of controls, however, the association between perceived threats and outgroup distance remains robust, particularly among those who supported the populist right-wing candidates (Milei and Bolsonaro). For Milei voters, all of our perceived threat items remain significantly associated with greater outgroup distance after introducing controls. For Bolsonaro voters, only the perceived threat of immigrants loses significance. In contrast, perceived threats from left-wing people and to national identity are still significantly associated with greater outgroup distance.
Results are somewhat less robust among the left-wing camp. In Argentina, Massa voters who perceive racists as a threat are significantly more likely to display greater social distance toward Milei voters, even after including controls. However, the item referring to inequality as a threat is no longer statistically significant at standard levels (p = 0.051), and the same applies to the threat from right-wing people. In Brazil, the statistical significance of these items also decreases among Lula voters after adding controls, but both the perceived threat from right-wing people and from racists remain statistically associated with greater outgroup distance (p < 0.05).
All in all, these findings suggest that the relationship between threat perception and outgroup distance is not entirely explained by other variables, providing further evidence that different threat types are relevant for different political camps. However, the findings seem to lend support to our hypothesis that threats are more important for those on the right, as demonstrated by the robustness of these associations for those who supported both Milei and Bolsonaro. For the left camp, much of the association between threat perceptions and outgroup distance is explained away by the other variables in the model, such as party support, negative identification, ideological extremity, etc. This is not the case for the right camp, where there is a clearer direct relationship between perceived threats and outgroup distance for most items.
In order to compare if this association between perceived threat and outgroup distance is greater for right-wing supporters than for those supporting the left-wing camp, we created two composite perceived threat indicators using principal component analysis. Eigenvalues and parallel analysis both indicated the presence of a two-component solution, with the first component comprising items that are relevant to right-wing supporters (perceived threats from left-wing people, immigrants, and national identity) and the second component comprising the three items we had identified as being relevant to supporters of the left camp (perceived threats from right-wing people, racist people, and rising inequality).Footnote 4 This approach allows for a more parsimonious assessment of the relevance of group-specific threats.
Figure 4 presents coefficient plots from the models using the perceived threat indices across our four voter groups: Milei and Massa supporters in Argentina and Bolsonaro and Lula supporters in Brazil. The models are detailed in the Supplemental Materials (Table 7a). In order to facilitate the comparison of coefficients, all variables were converted to standardized z-scores.
OLS coefficients with 95% confidence intervals.
The outcomes of these models provide further support for the hypothesized link between threat perception and social distance in both Argentina and Brazil. As expected, perceived threats are significantly associated with greater outgroup social distance across all political groups in both countries. Moreover, this association can only be observed for the perceived threat index that comprises the relevant items for each political camp. Importantly, though, the magnitude of the coefficient for the relevant threat index was significantly larger among Bolsonaro supporters (right-wing index = 0.25) than among Lula supporters (left-wing index = 0.11) (t = 1.92, p < 0.05, one-tailed Wald test). In Argentina, the relevant coefficient for Milei voters (right-wing index = 0.23) was also larger than that for Massa voters (left-wing index = 0.15), with a difference of 0.08. However, this difference was not statistically significant (t = 0.94, p = 0.18, one-tailed Wald test). These results provide partial evidence for our last hypothesis, which predicted a stronger direct association between threat perception and affective polarization among right-wing supporters.
For right-wing voters, the coefficient for the perceived threat (right-wing) index is among the largest in the models, comparable to other relevant variables such as the strength of in-party support, which is also associated with outgroup social distance across all political groups in both countries, and negative political identity. The latter is (ceteris paribus) significantly associated with outgroup social distance for supporters of Massa, Lula, and Bolsonaro, but not for those who supported Milei (the coefficient is positive, but it is not significant at p < 0.05). Areal (Reference Areal2022) also found a strong association between negative partisanship and outgroup social distance in Brazil, although the study did not specify whether this was the case for each of the two political camps. Holding other variables constant, ideological extremity is strongly and significantly associated with increased outgroup social distance among Milei supporters but not among Massa, Lula, and Bolsonaro supporters. Other controls, such as social sorting, ideological sorting, issue-based sorting, and issue extremity, were not significantly related to outgroup social distance in any of the models. This contrasts with research on Western democracies (Mason Reference Mason2016; Harteveld Reference Harteveld2021; Comellas Reference Comellas2022) and highlights the lack of relevance of these variables in Argentina and Brazil, which may also occur in other democracies with weakly institutionalized party systems.
In addition to these models, we also ran models including sociodemographic variables as controls (gender, age, religiosity, religion or denomination, education, household income, and urbanization of the place of residence). Results did not change in any substantive manner (see Supplemental Materials, Table 8a).
Finally, we replicated our models using an alternative specification of the dependent variable, which consisted of subtracting the respondent’s outgroup social distance from their ingroup social distance. This is standard practice for studies relying on like-dislike thermometers, but it is unconventional for social distance scales. Nevertheless, we ran the models as robustness checks (a replication of our final model can be found in the Supplemental Materials, Table 9a). Overall, results remain consistent with our primary analysis. Threats related to right-wing concerns are associated with higher levels of polarization among right-wing supporters in both Argentina and Brazil. Conversely, threats associated with left-wing concerns are only significant for Lula supporters in Brazil, with a smaller coefficient. The main difference with the standard social distance models is that, once controls are added, Lula supporters are less polarized toward Bolsonaro supporters when they perceive a threat from left-wing people, immigrants, or their national identity. The coefficient for right-wing threat among right-wing voters is also larger in magnitude than the corresponding coefficient for left-wing threat among left-wing supporters. However, when using this version of the dependent variable, the difference is only marginally significant in Brazil (p < 0.1, one-tailed Wald test) and continues to be non-significant in Argentina.
Discussion and conclusion
This paper sought to elucidate the dynamics of affective polarization in Argentina and Brazil by specifically examining the relationship between outgroup distance and various dimensions of threat perception. Furthermore, it analyzed the extent to which this relationship remains consistent for supporters of right-wing populist candidates compared to those who support the opposing political bloc.
The findings of this study reveal that, among individuals aligned with the right-wing candidates (Bolsonaro and Milei), animosity toward supporters of the rival candidate is correlated with perceived threats emanating from left-wing groups and other social groups, such as immigrants, as well as abstract threats affecting values such as national identity. Conversely, supporters of Lula and Massa exhibited greater levels of social distance toward supporters of the rival candidate when they perceived threats from right-wing groups, racist people, and (especially for Massa supporters) increasing inequality. However, the association between these threats and outgroup distance was clearly more robust and larger for right-wing supporters in Brazil (also in Argentina, though in this case, the differences did not reach statistical significance).
For the left camp, this association was largely (though not entirely) accounted for by variables such as party support, negative identification, and ideological extremity. In contrast, we found a more robust and direct relationship between perceived threats and outgroup distance for the right camp. This is consistent with prior research, conducted mostly within the United States context, which posited that individuals on the political right exhibit a more pronounced response to perceived threats compared to those on the political left (Hibbing, Smith, and Alford Reference Hibbing, Smith and Alford2014; Jost, Stern, Rule et al. Reference Jost, Stern, Rule and Sterling2017). In this way, our findings help strengthen previous evidence on conservatives’ greater sensitivity to threat (Jost, Napier, Thorisdottir et al. Reference Jost, Napier, Thorisdottir, Gosling, Palfai and Ostafin2007), as well as studies showing that conservatives and progressives tend to respond to different types of threats (Renström, Bäck, and Carroll Reference Renström, Bäck and Carroll2021). Nevertheless, the lack of robust evidence for such differences in Argentina suggests the importance of accounting for contextual variation. The populist and highly polarized nature of the Peronist left may offer a potential explanation for these results, as it may have contributed to slightly narrowing the traditional gap in threat sensitivity typically observed between the left and right camps in other countries.
All in all, this study contributes to the existing body of knowledge on the dynamics of affective polarization in two principal ways. Firstly, it shows that accounting for threat perceptions enables us to have a more comprehensive understanding of the form and nature of political antagonism. By extending this analysis to democracies with weak party institutionalization, we show that these psychological mechanisms transcend the over-studied contexts of the United States and Western Europe. Secondly, while threat perception is a universal psychological phenomenon, the salience and impact of specific threats are subject to contextual variation. Although our analyses only focused on a limited number of threats, the results suggest a relatively consistent association with outgroup distance in both Argentina and Brazil. Nevertheless, as we have mentioned, the lack of robust evidence for the hypothesis that right-wing supporters are more responsive to threat than left-wing supporters in Argentina (in contrast to the findings in Brazil) underscores that threat perception is likely shaped by the unique dynamics of national political competition.
Future studies should therefore build on this research by considering contextual specificity while also investigating a broader range of perceived threats beyond the six examined here. This could include threats posed by different political and social outgroups (e.g., other demographic groups, such as religious and education-based groups, and sociopolitical actors such as trade unions and social movements) and threats to other values and beliefs (e.g., environmental protection, gender equality, or traditional norms).
A further contribution of our study concerns the relevance of different correlates of affective polarization in the cases examined. Perceived threat emerged as one of the variables with the strongest correlation with outgroup distance in both Argentina and Brazil (particularly among right-wing supporters), alongside in-party support and negative political identity. In contrast, variables such as social sorting, ideological sorting, and issue extremity demonstrated limited statistical association with outgroup distance in most models. While social sorting is a well-documented driver of affective polarization in Western democracies (Mason Reference Mason2016; Harteveld Reference Harteveld2021; Comellas Reference Comellas2022), this relationship does not appear to hold for the two Latin American cases under study. This finding aligns with Areal (Reference Areal2022), who demonstrated the limited impact of such variables within the Brazilian context. A likely explanation for this divergence is the presence of weakly institutionalized party systems in both countries, characterized by relatively low levels of positive partisanship and the limited role of ideology in driving vote choice (Moraes and Béjar Reference Moraes, Béjar, Torcal and Harteveld2025). Ultimately, these results underscore the necessity of studying affective polarization beyond the typical European and North American cases that most of our current knowledge of the phenomenon stems from.
At this point, it is also necessary to acknowledge the inherent limitations of this study. Given that our analysis is based on observational data, we could not make any definitive causal claims. It thus remains possible that affective polarization drives threat perception, rather than the other way around, or that both phenomena are mutually reinforcing in a reciprocal process. Recent experimental research conducted in two West European multi-party democracies suggests that threat perceptions do indeed drive negative affect toward out-party supporters (Renström, Bäck, and Carroll Reference Renström, Bäck and Carroll2023). However, even if the causal arrow were reversed, our findings would still be highly significant, as they would indicate that affectively polarized voters (and especially supporters of the populist right) are systematically more likely than other voters to perceive political and social outgroups as a profound threat. This is a key finding for understanding the nature and depth of polarization in contemporary societies.
Future research should apply the theoretical framework using experimental interventions that manipulate threat perception. These experiments can help determine how to mitigate the impact of threat perceptions to prevent detrimental levels of political animosity. While important in all societies, this is particularly critical in democracies with comparatively weak democratic institutions and a lack of party-system institutionalization, which are highly vulnerable to escalating levels of polarization.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1475676526101583.
Data availability statement
For data and replication files for this article, see Supporting Materials.
The replication datasets provided are reduced versions of the original data. The complete, original datasets can be found at the links below:
Olivas Osuna, J. J., Ramiro, L., Vasilopoulou, S., Halikiopoulou, D., Zanotti, L., Santana, A., Rama, J., Clari, E., & Gomez, R. (2026). Socio-political attitudes in Argentina (2023) (Version V1) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.20557661
Brazil Data: Olivas Osuna, J. J., Ramiro, L., Vasilopoulou, S., Halikiopoulou, D., Zanotti, L., Santana, A., Rama, J., Clari, E., & Gomez, R. (2026). Socio-political attitudes in Brazil (2023) (Version V1) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.20559913
Acknowledgements
We are grateful to the four anonymous reviewers for their time and thoughtful feedback. We additionally thank the participants of the Affective Polarization section at the 2024 ECPR General Conference in Dublin and the EPOP 2024 Conference in Manchester for their valuable comments. This article was entirely written by the authors in English, but tools such as Grammarly Go were used to obtain suggestions for ex-post language editing. The authors are entirely responsible for the scientific content of the paper.
Funding statement
The research presented in this study received support from the Madrid Regional Government (project codes: 2018-T1/SOC-10152 and 2022-5A/SOC-24238) and the Spanish Research Agency (project code: PID2020-113182RA-I00).
Competing interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Ethical standards
Ethical approval for the surveys was granted by the Research Ethics Committee at the Universidad Nacional de Educación a Distancia (UNED) on June 7, 2020 (project code: 2018-T1/SOC-10152). This approval was subsequently renewed on October 23, 2023, prior to the interviews conducted in Brazil and Argentina. Participant recruitment, managed by YouGov, was conducted in compliance with the European Union’s Data Protection legislation (GDPR), ensuring participants provided informed consent and had the right to decline or withdraw.
Open access
