Introduction
Affective polarization – e.g., individuals’ proclivity to harbor personal animus and prejudice toward supporters of unfavored parties (Iyengar, Sood, Lelkes Reference Iyengar, Sood and Lelkes2012) – has proliferated in recent years (Janssen and Turkenberg Reference Janssen and Turkenberg2024; Reiljan Reference Reiljan2020). On both sides of the Atlantic (Boxell, Gentzkow, Shapiro et al. Reference Boxell, Gentzkow and Shapiro2024; Garzia, Ferreira da Silva, Maye et al. Reference Garzia, Ferreira da Silva and Maye2023; Gidron, Adams, Horne et al. Reference Gidron, Adams and Horne2023; Wagner Reference Wagner2021), the tendency to see one’s political adversaries not just as opponents but as enemies driven by bad or misguided intentions appears as one of the most pernicious problems facing liberal democracy today, as it may fuel social sorting, political gridlock, support for undemocratic norms, and unwillingness to accept electoral defeat (Berntzen, Kelsall, Harteveld et al. Reference Berntzen, Kelsall and Harteveld2024; Campos and Federico Reference Campos and Federico2025; Harteveld Reference Harteveld2021; Mason Reference Mason2016; McCoy and Somer Reference McCoy and Somer2019).
Our aim in this paper is to study how affective polarization is impacted by national identity, as previous research has suggested this as one of the most powerful candidates (Voelkel, Stagnaro, Chu et al. Reference Voelkel, Stagnaro, Chu, Pink, Mernyk, Redekopp, Ghezae, Cashman, Adjodah, Allen, Allis, Baleria, Ballantyne, Van Bavel, Blunden, Braley, Bryan, Celniker, Cikara and Willer2024) for the kind of superordinate identity that is able to make political adversaries recategorize one another into a common in-group (Gaertner, Dovidio, Anastasio et al. Reference Gaertner, Dovidio, Anastasio, Bachman and Rust1993), even by rather simple cues (Gur, Ayal, Wagner et al. Reference Gur, Ayal, Wagner, Adler and Halperin2024; Siegel and Badaan Reference Siegel and Badaan2020).
Yet, the effects of national identity on affective polarization remain far from settled. On the one hand, Levendusky (Reference Levendusky2018) found that affective polarization was significantly reduced when Americans were exposed to 4th of July celebrations, a vignette about the strength of America and Americans, or a question asking them to list why they were proud to be Americans. Riek, Mania, Lamoreaux et al. (Reference Riek, Mania and Lamoreaux2009) similarly concluded that inter-group hostility between Democrats and Republicans can be reduced by interventions that foster national pride. More broadly, national identity has further been found to enhance unity and cooperation across ethnic and religious divides in the United States (Transue Reference Transue2007), as well as in India (Charnysh, Lucas, Singh et al. Reference Charnysh, Lucas and Singh2015) and Malawi (Robinson Reference Robinson2016).
On the other hand, Brandt, Turner-Zwinkels, Inbar et al. (Reference Brandt, Turner-Zwinkels, Inbar and Quinn2020) were unable to replicate the depolarizing effect of 4th of July celebrations, and Park and Warner (Reference Park and Warner2024) similarly failed to establish a direct effect of national identity primes on affective polarization in the United States and South Korea. Indeed, different understandings of national identity seem to have become one of the primary sources of affective polarization, at least in the United States (Spinner-Halev and Theiss-Morse Reference Spinner-Halev and Theiss-Morse2024), where some groups now explicitly refer to the flag as divisive (Estrada et al. Reference Estrada, Hirsch, Lian, Ruwart and Chau2024). Exposure to the American flag has also been found to polarize democratic-republican ideologies (Chan Reference Chan2017) and appeals to a common in-group to exacerbate affective polarization (Klar Reference Klar2018; Rutchick and Eccleston Reference Rutchick and Eccleston2010), at least for voters who are keen to defend their nation from what they see as the threat of immigration (Wojcieszak and Garrett Reference Wojcieszak and Garrett2018). This has led some to suggest that national identity can ‘no longer serve as a unifying identity that binds the entire country together’ (Dawkins and Hanson Reference Dawkins and Hanson2024).
Moreover, the research front remains divided over whether or not national identity actually dampens or fuels affective polarization. We also note that there has not yet been a single experimental study of this potential relationship in a European context (for a cross-sectional one, however, see García-Sánchez, Turner-Zwinkels, Kesberg et al. Reference García-Sánchez, Turner-Zwinkels, Kesberg, Marot, Rodríguez-Bailón, Willis and Kuppens2024). Nor has previous research investigated the potential effects on polarization of the more subtle aspects of national identity that are known as ‘banal’ (as opposed to ‘hot’) nationalism (Billig Reference Billig1995), ‘everyday nationhood’ (Fox and Miller-Idriss Reference Fox and Miller-Idriss2008), or ‘nationalism in settled times’ (Bonikowski Reference Bonikowski2016). This is important, since previous research tells us that even when national symbols such as flags only hang limply in the background, which in many contexts is more common than for them to be proudly waved and paraded (cf. Hassin et al. Reference Hassin, Ferguson, Kardosh, Porter, Carter and Dudareva2009), such national identity cues can nevertheless exert a powerful influence on people’s political attitudes and behavior (Butz Reference Butz2009). Yet, we currently know virtually nothing about whether this is also true for affective polarization.
We address these research gaps by posing the following research question: What is the effect of exposure to the national flag on affective polarization in Sweden and Denmark?
More specifically, we conducted a survey experiment in both countries in the winter of 2023/2024. Our survey included two treatments: one in the form of a picture of a national flag hanging from a flagpole, and one that featured a birthday cake with a small national flag on it. Our primes capture a more realistic everyday scenario than, for example, asking respondents to list why they are proud of their nationality, as previous research has tended to do (Levendusky Reference Levendusky2018; Park and Warner Reference Park and Warner2024). As opposed to in the traditional 4th of July parade in the United States or the national celebrations in Norway on May 17, the Swedish and Danish flags are rarely explicitly waved. Yet they are ubiquitous, and even in this ‘unwaved’ and more ‘mindless’ form (Skey Reference Skey2017), national identity seems able to inspire unity, as when Swedes rallied around their internationally controversial policy approach to Covid-19 (Gustavsson and Larsson Taghizadeh Reference Gustavsson and Larsson Taghizadeh2023).
The first contribution we make in this paper is empirical and speaks to the literature on affective polarization. To our knowledge, this is the first experimental study to date of the causal effect of national identity primes on affective polarization in Europe. It is also the first ever study, not only in Europe but globally, of how simply being exposed to the national flag in its ordinary, uncommented day-to-day form (as opposed to being primed with national pride specifically) impacts affective polarization. We furthermore take a comprehensive approach to the measurement of affective polarization by estimating the flags’ effects on three outcomes: social distancing, trait stereotyping, and party dislike. Affective polarization is multidimensional, comprising at least three subdimensions – othering, moralization, and aversion (Campos and Federico Reference Campos and Federico2025). Consistent with prior work (e.g., Berntzen, Kelsall, Harteveld et al. Reference Berntzen, Kelsall and Harteveld2024; Boxell, Gentzkow, Shapiro et al. Reference Boxell, Gentzkow and Shapiro2024; Druckman and Levendusky Reference Druckman and Levendusky2019; Janssen and Turkenberg Reference Janssen and Turkenberg2024; Reiljan Reference Reiljan2020; Wagner Reference Wagner2021), our measures primarily capture the aversion subdimension. Aversion has been especially strongly linked to downstream democratic consequences such as reduced support for democratic norms, increased support for authoritarian rule, anti-democratic-candidate voting, and political violence (Campos and Federico Reference Campos and Federico2025). Because aversion is itself multifaceted and common measures have not always been strongly correlated in previous research (Druckman and Levendusky Reference Druckman and Levendusky2019; Kekkonen, Suuronen, Kawecki et al. Reference Kekkonen, Suuronen, Kawecki and Strandberg2022), our study incorporates the three most prominent aversion metrics in previous research (Röllicke Reference Röllicke2023) to provide a comprehensive view of which facets of aversion are responsive to the national identity primes.
Our second contribution is theoretical, as we expand the discussion on how national identity could impact affective polarization to also consider the ‘banal’ (Billig Reference Billig1995) and everyday forms of nationhood (Fox and Miller-Idriss Reference Fox and Miller-Idriss2008) that have hitherto remained overlooked. We also offer a clarification of the common in-group identity model, noting that such more implicit forms of national identity need not work their unifying magic by strengthening a person’s explicit national identity attitudes, for example, in the form of national pride, as previous research has assumed (cf. Levendusky Reference Levendusky2018; Park and Warner Reference Park and Warner2024). In order to test this, we include a more extensive question battery than previous research in our survey, including, for example, both a general kind of national pride (e.g., pride in being Swedish/Danish) and a more issue-specific kind (pride in the country’s institutions and culture, respectively).
Our third contribution is methodological. While previous research has studied how vignettes priming national pride, or national holidays like the 4th of July, impact affective polarization, we offer the first study of whether or not merely seeing the national flag does so. We know of no other studies – either in Europe, North America, or indeed any other contexts – that have examined this specific relationship. This means that our findings speak not only to the literature on affective polarization and the one on banal nationalism and everyday nationhood but also to the one on visual national identity primes and ‘mindless markers of the nation’ (Skey Reference Skey2017), including the national flag, which have hitherto focused on other outcomes, such as political opinions and ideology (Carter, Ferguson, Hassin et al. Reference Carter, Ferguson and Hassin2011b, Reference Carter, Ferguson and Hassin2020; Chan Reference Chan2017), inter-ethnic attitudes (Charnysh, Lucas, Singh et al. Reference Charnysh, Lucas and Singh2015; Hassin et al. Reference Hassin, Ferguson, Kardosh, Porter, Carter and Dudareva2009; Transue Reference Transue2007), and tax compliance (Gangl, Torgler, Kirchler et al. Reference Gangl, Torgler and Kirchler2016).
Our main finding is that, in Sweden but not Denmark, both flag primes reduced affective polarization when measured as social distancing, but not in terms of trait stereotyping or party dislike. The effect on social distancing is in line with our preregistered hypothesis. Moreover, the link between the flag primes and reduced affective polarization does not seem to be mediated by any of our questions capturing explicit national identity attitudes, such as national pride. This, we suggest, further underscores how national flags need not be openly waved in order to impact our political attitudes.
The common in-group identity model: Beyond national pride
Coining the seminal conception of the nation as ‘an imagined community’, Benedict Anderson (Reference Anderson1983) posited that a common national identity enables unacquainted members of the same nation to feel loyalty and fraternity with each other. From a psychological point of view, such an imagined community constitutes a social identity, and this is also how we will approach national identity in what follows.
Social Identity Theory (SIT) posits that there is a universal human need to create and maintain a positive attachment to the groups with which one identifies, and that this in turn leads people to hold more positive attitudes toward those whom they consider members of their in-group (Tajfel and Turner Reference Tajfel, Turner, Austin and Worschel1979). By invoking a shared identity, one can thus raise people’s empathy, loyalty, and solidarity with other in-group members, even at a cost to their own individual well-being (Ellemers, Spears and Doosje, Reference Ellemers, Spears and Doosje2002; Ellemers, Kortekaas and Ouwerkerk Reference Ellemers, Kortekaas and Ouwerkerk1999; Hornstein Reference Hornstein and Berkowitz1978; Mullen, Brown, Smith et al. Reference Mullen, Brown and Smith1992; Ouwerkerk, Ellemers, de Gilder et al. Reference Ouwerkerk, Ellemers, de Gilder, Ellemers, Spears and Doosje1999). Experiments within the ‘minimal group’ paradigm tell us that this can happen even when the shared identity is highly superficial and contrived. Moreover, while individuals often feel belonging to multiple social groups, which of them is most salient and thus likely to affect their perception of others as in- vs. out-group members largely depends on contextual cues (Gaertner, Dovidio, Anastasio et al. Reference Gaertner, Dovidio, Anastasio, Bachman and Rust1993).
In the following, we zoom in on the national flag – one of the perhaps most ubiquitous and powerful triggers of such a social identity, even when it remains ‘unwaved’. Previous studies on ‘banal nationalism’ (Billig Reference Billig1995), the unnoticed and mundane nationalism in ‘settled times’ (Bonikowski Reference Bonikowski2016), and the ‘everyday nationhood’ that goes ‘beyond pomp and circumstance’ (Fox Reference Fox2017) all insist that national flags do not ‘have to be saluted or waved to work their national magic’ (Fox and Miller-Idriss Reference Fox and Miller-Idriss2008, p. 549). We believe this makes a lot of sense. Yet, whether or not this ‘magic’ also extends to dampening affective polarization remains an open question, since previous experimental studies using subliminal or explicit flag primes, as well as exposure to other national symbols, have focused on other outcomes, such as political ideology (Carter, Ferguson, Hassin et al. Reference Carter, Ferguson and Hassin2011b) or attitudes to minorities (Becker, Enders-Comberg, Wagner et al. Reference Becker, Enders-Comberg, Wagner, Christ and Butz2012).
Some of these studies have found that exposure to the American flag increases political aggression (Ferguson and Hassin Reference Ferguson and Hassin2007) and ideological polarization (Chan Reference Chan2017). Yet, others have argued that the ideological effects of exposure to the American flag have in fact decreased over time (Carter, Ferguson, Hassin et al. Reference Carter, Ferguson and Hassin2020). The bulk of these studies, however, have found that exposure to national flags has a unifying rather than divisive effect, bringing people together across political, racial, and religious divides. This seems to hold true in contexts as varied as the United States (Butz, Plant, Doerr et al. Reference Butz, Plant and Doerr2007; Transue Reference Transue2007; Wright and Citrin Reference Wright and Citrin2011), India (Charnysh, Lucas, Singh et al. Reference Charnysh, Lucas and Singh2015), Pakistan (Kalin and Siddiqui Reference Kalin and Siddiqui2020), Israel (Gur, Ayal, Wagner et al. Reference Gur, Ayal, Wagner, Adler and Halperin2024; Hassin et al. Reference Hassin, Ferguson, Kardosh, Porter, Carter and Dudareva2009), Italy and Russia (Hassin et al. Reference Hassin, Ferguson, Kardosh, Porter, Carter and Dudareva2009), Austria (Gangl, Torgler, Kirchler et al., Reference Gangl, Torgler and Kirchler2016), Australia, New Zealand, and Great Britain (Chan Reference Chan2019; Sibley, Hoverd, Duckitt et al. Reference Sibley, Hoverd and Duckitt2011), and Malawi (Robinson Reference Robinson2016).
Why, one might ask, would exposure to nationhood reminders such as the national flag reduce affective polarization specifically? We follow Levendusky’s (Reference Levendusky2018) suggestion that the answer lies in the very cause of this polarization, at least according to Social Identity Theory: the human tendency to categorize people as either in- or out-group members. For SIT, it bears reminding that it does not only stress that such divisions lead to in-group favoritism. It also emphasizes that who is considered part of the in-group versus the out-group varies greatly, depending on which group identity happens to be the most salient in any particular situation. This means that group identities may not only fuel polarization; if they are inclusive enough, they can also dampen it.
Levendusky thus theorizes that when a Democrat thinks of a Republican, it is likely that her own partisan identity – dividing her from the Republican – will be most salient, potentially fueling affective polarization. When she is reminded of her national identity as American, however – an identity she shares with the Republican – she will recategorize him into her in-group. This means that she suddenly sees him as more similar to herself than she would otherwise, and thus in turn also becomes less hostile and feels more warmth toward him.
More precisely, the assumed mechanism here stems from the Common Ingroup Identity Model (CIIM) and consists of re-categorization (Gaertner, Dovidio, Anastasio et al. Reference Gaertner, Dovidio, Anastasio, Bachman and Rust1993). The CIIM holds that by cueing a ‘superordinate group, the favoritism reserved for in-group members is extended to former out-group members and mitigates conflict between the subgroups’ (Kalin and Suddiqui Reference Kalin and Siddiqui2020, p. 62; also see Charnysh, Lucas, Singh Reference Charnysh, Lucas and Singh2015). In other words, if I think of myself and others mainly in partisan terms, I will categorize my opponents as members of an out-group; but if an overarching identity that we instead share – such as the national one –becomes salient, I will think of both myself and my opponents in terms of that shared identity, thus re-categorizing them from members of my out-group to members of my in-group, which in turn is likely to reduce my animosity towards them.
Extending this reasoning to national flags, which Levendusky did not explicitly study, we formulated the following pre-registered hypothesis to be tested empirically:
Exposure to the national flag will lead to attenuated levels of affective polarization compared to the control group.
Note that while we agree with previous research in expecting reminders of the national identity to reduce affective polarization via the re-categorization mechanism described above, we do not assume that national pride holds a key role in this process. We thus differ from Park and Warner (Reference Park and Warner2024, p. 3), who state that ‘priming national identity should heighten the sense of national pride, and the intensified national pride should reduce political intergroup tensions between partisan groups’. We also believe it is a mistake to simply assume, as does Levendusky (Reference Levendusky2018), that national identity and national pride are so ‘deeply entwined’ empirically that priming one of them (in his case pride) is likely to also prime the other (in his case national identity).
Our objection here is based on the observation that, strictly speaking, the CIIM is rooted in Social Categorization Theory, an offshoot of Social Identity Theory that stresses the cognitive rather than affective aspects of group membership. This means that the CIIM has to do with perceptions of group boundaries, not the feelings, such as pride, that the groups themselves might invoke. The re-categorization it assumes in a person who is cued with a common in-group identity is thus that of a different cognitive representation of the former out-group member, and perhaps also of herself (Gaertner, Dovidio, Anastasio et al. Reference Gaertner, Dovidio, Anastasio, Bachman and Rust1993), but not necessarily a change in her feelings towards this group as such.
Of course, it is empirically possible that flag exposure reduces affective polarization not via cognitive re-categorization only, but also via heightening in-group pride – but if we only prime national pride, as previous research has done, we will never know if that is the case. If we only ask explicit questions about national pride, moreover, we also miss the other sub-dimensions of national identity that exist. Already, the classic definition of a social identity as ‘that part of a person’s self-concept that derives from the knowledge of his membership in a social group (or groups) together with the value and emotional significance attached to the membership’ (Tajfel Reference Tajfel1981, p. 255) suggests that pride is but one of several sub-dimensions of such an identity.
Contrary to what Levendusky (Reference Levendusky2018) seems to assume, several studies of the dimensionality of national identity have indeed shown that national pride forms an empirically distinct dimension from, for example, feeling close to one’s nation or co-nationals (cf. Gustavsson Reference Gustavsson, Gustavsson and Miller2020; Huddy and del Ponte Reference Huddy, del Ponte, Gustavsson and Miller2020), the latter of which can, by contrast, be combined with both national shame and constructive critique (Schatz, Staub, Lavine Reference Schatz, Staub and Lavine1999). The effects of national pride on outcomes like trust also diverge and sometimes even run in the opposite direction from those of national attachment (Gustavsson and Stendahl Reference Gustavsson and Stendahl2020). This, then, might also be the case for an outcome like affective polarization.
In the following, we have thus included a question battery with items gauging explicit national pride, national identification, and national attachment. Since openly expressing general national pride remains controversial for some, perhaps especially in a European context, we also make sure to measure issue-specific national pride, which suffers less from social desirability bias (Meitinger Reference Meitinger2018).
We refrain from formulating specific hypotheses about whether or not and, if so, which of the items in our explicit national identity battery mediate the expected effect of our flag primes on affective polarization. Instead, we see this as an open question that we here offer a first attempt at probing empirically. It is quite possible, indeed, that none of the explicit national identity attitudes are involved in this process, and that the only thing that flags ‘do’ is to lead to a cognitive re-categorization of ourselves and our political adversaries, without affecting our attitudes towards that shared (national) identity itself. After all, several studies have found that implicit reminders of nationhood, such as flag primes, can have significant effects on other political attitudes and behavior than affective polarization, while leaving the subjects’ explicit national identity attitudes, including pride, virtually unchanged (cf. Becker, Enders-Comberg Wagner et al. Reference Becker, Enders-Comberg, Wagner, Christ and Butz2012; Carter, Ferguson, Hassin Reference Carter, Ferguson and Hassin2011a; Olsson Reference Olsson2024).
Contextualizing the cases: Sweden and Denmark
To our knowledge, this study is the first to examine implicit nationalism via flag primes in Scandinavia – specifically Sweden and Denmark. National identity in these countries is at least as strongly felt as in the more studied United States case, based on aggregate levels of national attachment (Gustavsson and Miller Reference Gustavsson, Miller, Gustavsson and Miller2020), which, like general national pride, national chauvinism, and uncritical patriotism, aligns with European averages. However, Scandinavian national identity often manifests more subtly. For example, the ‘rally around the unwaved flag’ helped explain the Swedish public’s tenacious support for national authorities’ controversial Covid-19 policies (Gustavsson and Larsson Taghizadeh Reference Gustavsson and Larsson Taghizadeh2023).
To outside observers, Sweden and Denmark appear strikingly similar: both feature banal nationalism via unwaved flags on public buses, birthday cakes, and community flagpoles (Löfgren Reference Löfgren, Eriksen and Jenkins2007), and a Nordic welfare model characterized by universal benefits (Breidahl Reference Breidahl2017; Esping-Andersen Reference Esping-Andersen1990), high social trust (Rothstein and Uslaner Reference Rothstein and Uslaner2005), and a strong emphasis on political consensus (Bergman and Strøm Reference Bergman and Strøm2011). But Sweden and Denmark are not carbon copies of each other. Although they share features often viewed as making them ‘suitable for ‘most similar systems’ comparisons’ (Bergman and Strøm Reference Bergman and Strøm2011, p. 35), they nonetheless differ in ways that may shape the relationship between national identity and affective polarization.
First, the Danish flag (nicknamed Dannebrogen), often regarded as the world’s oldest national flag still in use, is a deeply entrenched source of pride (Nielsen Reference Nielsen2018). It is omnipresent in both public and private settings and frequently appears in nonpolitical or commercial contexts. The Dannebrog plays an integral role on social occasions, from birthday celebrations to football matches, and serves as a deep-rooted symbol for Danes (Warburg Reference Warburg2008). Although the use of the Swedish flag bears notable similarities to the Danish flag, it tends to be less omnipresent in private spaces, which may also be the reason why some observers have noted that, to a Swede, the more widespread use of flags in Denmark and Norway can even come across as ‘vulgar’ (Eriksen and Jenkins Reference Eriksen and Jenkins2007). Thus, while both flags tend to be omnipresent and rather ‘unwaved’, flag ubiquity is particularly pronounced in Denmark.
Second, since the far-right Sweden Democrats entered parliament in 2010, affective polarization has risen more sharply in Sweden than in Denmark. The Sweden Democrats are notably disliked, even compared to similar parties in neighboring Nordic countries (Ryan Reference Ryan2023). Indeed, one study found that Sweden experienced the largest increase in party antipathy among European countries after 2008 (Bettarelli, Reiljan, Van Haute Reference Bettarelli, Reiljan and Van Haute2022).
Our exploratory approach compares Sweden and Denmark to assess how national flag exposure affects affective polarization. Higher baseline polarization in Sweden, along with the relatively widespread dislike of the Sweden Democrats among a majority of voters, may limit the flag’s unifying potential. In contrast, the greater ubiquity of the flag in Denmark might reduce its political salience and, consequently, its effectiveness as a recategorization mechanism. This comparative design allows us to examine whether and how the flag–affective polarization relationship differs across these contexts.
Data and methods
The results in this paper derive from an original survey fielded by YouGov between 30 November 2023 and 1 January 2024, using population-representative samples of Danish and Swedish citizens aged 18 and older. Samples were drawn randomly from YouGov’s internet panels in each country. Summary statistics appear in Table 1 (online Appendix C). Our primary analytical approach is OLS regression. To explore potential mediation pathways from the flag priming experiments to affective polarization via explicit national-identity attitudes, we also employ structural equation modeling (SEM) with maximum likelihood estimation as a robustness check (see online Appendix G).
Experimental treatment
To test our hypotheses, we implemented a visual national identity priming experiment designed to contrast the effects of two subtle national flag primes: one in which the national flag was the central focus of the image (cf. Chan Reference Chan2017), and another in which the flag appeared embedded in a social context and was not prominently featured. We preregistered our hypotheses via AsPredicted before fielding the survey and experiment.Footnote 1
After completing a battery of background questions, the respondents for this study were randomly assigned to one of four conditions consisting of two treatment groups and two matching controls. In the first treatment group (T1), participants received brief information about the remainder of the survey accompanied by an image of the national flag. The first control group (C1) received the exact same text but no image. In the second treatment group (T2), participants were shown an identical introductory text as in T1 and C1 and were then asked when they last attended a birthday party, whereby a picture of a birthday cake with a small national flag appeared next to that question. The second control group (C2) was identical to T2 except that the cake image contained no flag. English translations of all stimuli appear in online Appendix A, and balance checks based on age, gender, ideology, and education appear in Appendix B. None of the differences between treatment and control conditions reached statistical significance at p < 0.05, confirming experimental balance.
After excluding respondents who rushed through the survey or straightlined across grid items,Footnote 2 we retained 632 Swedish participants in the pure flag sample and 640 in the cake flag sample. In Denmark, the corresponding counts were n = 622 (pure flag) and n = 612 (cake flag). An a priori power analysis (reported in Appendix F), based on effect sizes from prior subtle flag-priming studies (Butz, Plant, Doerr Reference Butz, Plant and Doerr2007; Gangl, Torgler, Kirchler Reference Gangl, Torgler and Kirchler2016), indicated that this sample size would detect a ≥3-4 percentage-point difference between treatment and control with 80 percent power.
Our treatment was deliberately subtle in two respects. First, respondents were not prompted to consciously reflect on the flag image. Second, we did not ask participants to explicitly write about what they like or dislike about their country (cf. Gangl, Torgler, Kirchler Reference Gangl, Torgler and Kirchler2016; Levendusky Reference Levendusky2018). We opted for this weaker intervention to enhance external validity: real-world exposure to national symbols is typically wordless and uncommented. Finding any effect under these conditions would suggest that repeated and more overt flag exposure could indeed influence intergroup attitudes.
Because the treatment was subtle, all respondents did not actively notice the images. About 30 percent of retained participants in the pure flag treatment (T1) reported seeing no picture at all; in the cake experiment (T2 and C2), this figure rose to almost 48 percent. Prior research on subliminal priming has shown that national symbols can affect attitudes even when processed only subconsciously (Butz, Plant, Doerr Reference Butz, Plant and Doerr2007; Hassin et al. Reference Hassin, Ferguson, Kardosh, Porter, Carter and Dudareva2009; Sibley, Hoverd, Duckitt Reference Sibley, Hoverd and Duckitt2011). Hassin et al. (Reference Hassin, Ferguson, Kardosh, Porter, Carter and Dudareva2009) found that subliminal symbol primes produced effects similar in magnitude to those where the treatment was ‘far from subtle’ (Hassin et al. Reference Hassin, Ferguson, Kardosh, Porter, Carter and Dudareva2009, 138). To retain statistical power to detect small effects, our primary analyses include all respondents regardless of whether they passed our attention check. Importantly, to validate robustness, we conduct sensitivity analyses on the attentive and inattentive subsamples separately.
Measures
Dependent variable: Affective polarization
Because affective polarization is multidimensional, we employ a comprehensive approach that distinguishes three subtypes within the aversion dimension (Campos and Federico Reference Campos and Federico2025): social distancing, trait stereotyping, and party dislike (Druckman and Levendusky Reference Druckman and Levendusky2019).
The social distancing measure assesses willingness to form close social ties across partisan divides. Respondents answered, ‘How would you feel if a close family member married a supporter of…’ followed by a randomized list of all national parliamentary parties (12 in Denmark, 8 in Sweden). Responses ranged from 1 = very unhappy to 6 = very happy.
The trait stereotyping measure assesses negative prejudice toward opposing partisans. Respondents indicated their agreement with statements of the form, ‘To what extent do you agree or disagree that each of the following traits describes supporters of [party]?’ The traits were intelligent, reliable, selfish, and mean. Responses range from 1 = completely agree to 7 = completely disagree, repeated for each party.
The party dislike measure is based on a feeling thermometer that asked, ‘How do you feel about the parties in the Swedish/Danish parliament?’ followed by the same list of parties. Responses range from 1 = strongly dislike to 7 = strongly like.
For all three measures, the party order was randomized between respondents but held constant within each respondent to avoid order effects and reduce cognitive load.
Following other recent studies (e.g., Bettarelli, Reiljan, Van Haute Reference Bettarelli, Reiljan and Van Haute2022; Janssen and Turkenberg Reference Janssen and Turkenberg2024; Nordbrandt Reference Nordbrandt2021), we adopt Wagner’s (Reference Wagner2021) Affective Polarization Index (API) approach adapted for multiparty systems. Wagner proposes two overarching measurement methods: spread of like–dislike scores (spread) and Mean Distance from the Most Liked Party (MDLP). Each of these can be implemented using either an unweighted or a vote share-weighted method. The first approach calculates the spread of rating scores across all parties without weighting for vote share. It captures each respondent’s average deviation in liking from their own individual mean. Low spread indicates similar feelings for all parties, while high spread indicates large differences in evaluations, i.e., higher levels of individual-level affective polarization. The alternative version of this measure weights party evaluations according to each party’s vote share. The second approach uses the respondent’s most liked party as a reference point and calculates the mean distance between this party and all others. A weighted version accounts for party size when computing these distances. Accordingly, we derive four indices of affective polarization: spread unweighted, spread weighted, MDLP unweighted, and MDLP weighted. For full methodological details, see Wagner (Reference Wagner2021).
All four measures were normalized such that they took on the value of 0 if there was no difference in an individual’s feelings for different voter groups and 1 if there was a maximum deviation from the mean (spread) or the most favored part (MDLP). The scores generated by each trait rating were averaged into a final additive index, ranging between 0 = no difference in the trait ratings for different supporter groups and 1 = maximum deviation.
Mediating variables: Explicit national identity attitudes
To test whether exposure to the national flag strengthened explicit national ingroup identification, attachment, and pride, which in turn might serve as mechanisms for reduced affective polarization, we included an extensive question battery tapping explicit national identity attitudes (cf. Levendusky Reference Levendusky2018; Park and Warner Reference Park and Warner2024).
We measured national identification by asking respondents, ‘How strongly do you identify as Swedish/Danish?’ on a 1–5 scale (1 = very strongly; 5 = not strongly at all). National attachment was assessed by asking, ‘How close do you feel to Sweden/Denmark?’ (1 = very close; 5 = not close at all). For national pride, we divided the concept into general pride and two issue-specific subdimensions. General pride was captured through the question, ‘How proud are you of being Swedish/Danish?’ (1 = very proud; 5 = not proud at all). To address potential social-desirability bias and cross-country variation (Meitinger Reference Meitinger2018), we also collected issue-specific pride items and used factor analysis to verify two distinct dimensions. The first dimension, institutional pride (α = 0.84), combines items measuring pride in how democracy works; the country’s political influence in the world; economic development; the social welfare system; and fair and equitable treatment of all groups. The second dimension, cultural pride (α = 0.70), combines items measuring pride in scientific and technological advancements, sporting successes, art and literature, and the national history. All pride items were asked on a 1–5 scale (1 = very proud; 5 = not proud at all).
After reverse-coding so that higher values consistently indicate stronger identification, attachment, or pride, each variable was normalized to range from 0 (lowest possible level) to 1 (highest possible level).
Results
Figure 1 displays the results of OLS regressions testing the treatment effects of our two flag primes on the three measures of affective polarization in Sweden and Denmark, highlighting effects statistically significant at the 95 percent confidence level (p < 0.05.)
Figure 1. Treatment effects of exposure to the national flag on social distancing, trait stereotyping, and party dislike using Wagner’s (Reference Wagner2021) four measurement approaches.
Notes: Orange color = significant at the 95% confidence level; grey color = not significant at the 95% confidence level. n flag = 585, n cake = 593 in Sweden. n flag = 510, n cake = 497, in Denmark.
The key finding is that both subtle flag primes – the standalone flag image and the flag on a birthday cake – significantly reduced social distancing scores in Sweden but not in Denmark. Specifically, exposure to the Swedish flag lowered social distancing by about 4–5 percentage points, compared to the no-flag control. Seven of eight coefficients for social distancing, using both unweighted and weighted spread and MDLP measures, are significant at the 95% level. Given the subtlety of our primes, this is not a negligible effect, and it is comparable to effect sizes in prior subtle flag priming studies (e.g., Butz, Plant, Doerr Reference Butz, Plant and Doerr2007).
Meanwhile, none of the treatments had a statistically significant impact on the other aversion measures, i.e., trait stereotyping and party dislike, in Sweden. Although the trait stereotyping coefficients are consistently negative (indicating lower prejudice), they do not reach statistical significance. In Denmark, effects were generally weak, and none of the coefficients exceeded the 95% significance threshold. Notably, however, the direction of the Danish effects, particularly for the cake prime, tends to be positive. In other words, socially embedded flag exposure produced opposite substantive effects in Sweden and Denmark, even though statistical confidence in this cross-country contrast is low.
Robustness checks
Appendices D and E report the results of a host of sensitivity analyses. First, the Swedish social distancing effects remain robust when including controls to account for potential compositional differences between treatment and control groups (see Appendix C for control specifications and Appendix D for results). Figure D1 in the Appendix examines whether the flag-treatment effects on social distancing are consistent among attentive Swedish respondents – i.e., those who correctly identified the picture they were shown – and inattentive respondents – those who did not. As expected, confidence intervals widened when the sample was split, reflecting what is to be expected given the reduced power in these analyses3. Nevertheless, the negative pattern persisted in both subgroups as the cake flag lowered social distancing in both the attentive and inattentive samples, consistent with Figure 1 (cf. Butz et al. Reference Butz, Plant and Doerr2007; Hassin et al. Reference Hassin, Ferguson, Kardosh, Porter, Carter and Dudareva2009; Sibley et al. Reference Sibley, Hoverd and Duckitt2011). No corresponding estimates could be produced for the inattentive pure-flag subsample because all participants in the pure-flag control group passed the attention check.
Appendix E disaggregates the average flag effects by the partisan groups that were rated by respondents, largely reflecting the patterns shown in Figure 1. In Sweden, respondents in the treatment groups exhibited substantially lower social distancing scores toward seven out of eight partisan groups, indicating evaluations that were closer to the overall average compared to those in the control groups. The only exceptions were evaluations of Social Democrat (S) voters in the pure-flag condition – the only partisan group toward which the treatment amplified social distancing slightly – and Christian Democrat (KD) voters in the cake condition, where treatment and control group responses did not differ meaningfully. This pattern suggests that the overall reduction in social distancing was not driven by perceptions of any single partisan group, although the effect was most pronounced in ratings of the Sweden Democrats. For the other two outcome measures in Sweden, most treatment effects were either negative or null, with the only positive effect observed in ratings of Left Party (V) voters in the cake-flag condition. In Denmark, evaluations in the pure-flag condition ranged from negative to positive, while cake-flag effects were predominantly positive, suggesting increased affective polarization in the treatment groups.
Additional heterogeneity analyses where the effects are disaggregated based on explicit national identity attitudes, political ideology, gender, household income, and education (Appendix E) reveal few pronounced subgroup differences. With the caveat of limited statistical power, centrist respondents in Sweden showed a larger reduction in trait stereotyping following the pure flag prime than left-leaning ones. Likewise, Danes with above-average pride in national institutions exhibited higher affective polarization than those with below-average institutional pride in the pure flag treatment. Across both primes, we observed no significant subgroup differences for social distancing.
In summary, exposure to the national flag reduced social distancing by about 4–5 percentage points in Sweden, but not in Denmark. This result was replicated in two distinct samples that received similar but slightly different treatments, strengthening our conclusion. The observed decrease in social distancing was further not driven by reduced antipathy toward any single partisan group but occurred more or less across the board (Appendix E). These results partly support our preregistered hypothesis that flag exposure would attenuate affective polarization, but this effect was limited to social distancing and did not extend to trait stereotyping or party dislike.
Mediation analysis
In Appendix G, we present a SEM-path model examining whether explicit national identity attitudes mediate the combined flag treatments’ effect on the unweighted social-distancing measure in Sweden. Figure G.2 shows a direct effect of the flag treatments on reduced social distancing of −0.05 (p < 0.01), consistent with the OLS results in Figure 1. However, none of the explicit national-identity attitudes significantly mediates this effect, as the indirect effect of the national identity primes was effectively zero (b ≈ 0.001). These findings indicate that the flag effect in our experiment does not operate by strengthening explicit national pride, identification, or attachment, contrary to expectations in prior studies (Levendusky Reference Levendusky2018; Park and Warner Reference Park and Warner2024). We discuss alternative mechanisms in the concluding section.
Concluding discussion
Ever since J.S. Mill claimed that in order for representative democracy to function, it requires a shared national identity. The unifying versus divisive effects of such an identity have been debated both empirically and philosophically, not least in the literature on liberal nationalism (c.f. Charnysh, Lucas, Singh Reference Charnysh, Lucas and Singh2015; Gustavsson and Miller Reference Gustavsson, Miller, Gustavsson and Miller2020; Gustavsson and Stendahl Reference Gustavsson and Stendahl2020; Huddy and del Ponte Reference Huddy, del Ponte, Gustavsson and Miller2020). Our results speak to this wider debate, as well as to the growing literature on whether affective polarization in particular is reduced (cf. Levendusky Reference Levendusky2018; Park and Warner Reference Park and Warner2024; Riek, Mania, Lamoreaux Reference Riek, Mania and Lamoreaux2009) or, on the contrary, heightened (cf. Dawkins and Hanson Reference Dawkins and Hanson2024; Wojcieszak and Garrett Reference Wojcieszak and Garrett2018) by national identity reminders – a debate that has hitherto failed to consider Europe or the potential influence of ‘banal’ (Billig Reference Billig1995) and ‘everyday’ nationhood (Fox and Miller-Idriss Reference Fox and Miller-Idriss2008) symbols, such as the ubiquitous yet uncommented presence of the ‘unwaved’ flag (Skey Reference Skey2017).
We have sought to address these research gaps by conducting a national flag priming experiment in Sweden and Denmark – two countries where the national flag appears to be rarely explicitly waved or revered but is still often present as a background feature in all kinds of day-to-day settings.
The central finding of our study is that in Sweden, but not in Denmark, exposure to the national flag did reduce affective polarization, but only when measured as social distancing and not when measured as trait stereotyping or party dislike. Thus, the results partially support our preregistered hypothesis that exposure to the national flag would reduce affective polarization. Our findings – showing that our primes exerted different effects depending on the outcome measure – underscore the importance of a multidimensional approach to measuring affective polarization so as to avoid oversimplified conclusions. Previous research indicates that individuals express greater aversion toward parties and their representatives than toward party supporters (Kekkonen, Suuronen, Kawecki et al. Reference Kekkonen, Suuronen, Kawecki and Strandberg2022). It is possible, then, that national-identity primes influence perceptions of co-nationals only at a horizontal, citizen-to-citizen level. At the same time, the null findings for trait stereotyping suggest that not all horizontal measures of affective polarization tap into the same underlying processes. While identity primes may increase willingness to form personal bonds with out-group partisans, reshaping more deep-seated stereotypes may require other types of interventions. This highlights the need for future research to investigate more carefully the scope conditions under which different dimensions of affective polarization are responsive to interventions of various kinds, including identity primes.
Although we did not find an effect across all three measures of affective polarization – thus we cannot claim that national flags attenuate every facet of it – we did find it for one important dimension: social distancing. This form of polarization is itself a driver of social sorting, which in turn increases the homogeneity of social ties and thus tends to further heighten other aspects related to affective polarization, such as intolerance and anger (c.f. Mason Reference Mason2015, Reference Mason2016). Our results suggest that subtle reminders of a common national identity may help disrupt this cycle, reducing the social distancing that underpins social sorting.
Our flag prime was intentionally subtle, meaning that respondents were exposed to the flag only briefly and without additional prompts or cues to ensure they actively noted or reflected on its meaning. This means that our experiment has high external validity, as it mimics real-world expressions of ‘banal nationalism’ much more closely than the intervention experiments by previous research, which were designed to foster an inclusive national identity through vignettes or reflection tasks (e.g., Levendusky Reference Levendusky2018; Park and Warner Reference Park and Warner2024; Voelkel, Stagnaro, Chu et al. Reference Voelkel, Stagnaro, Chu, Pink, Mernyk, Redekopp, Ghezae, Cashman, Adjodah, Allen, Allis, Baleria, Ballantyne, Van Bavel, Blunden, Braley, Bryan, Celniker, Cikara and Willer2024). Everyday national identity cues rarely come with such reflective prompts. Our findings are thus a relatively strong testimony that the presence of national flags, even when they are subtly woven into the fabric of everyday life, may shape political attitudes and intergroup relations.
In contrast to previous studies, which have tended to assume that the depolarizing effect of national identity should be mediated by a strengthened sense of national pride (e.g., Levendusky Reference Levendusky2018; Park and Warner Reference Park and Warner2024), we theorized that the recategorization process of the Common Ingroup Identity Model (Gaertner, Dovidio, Anastasio et al. Reference Gaertner, Dovidio, Anastasio, Bachman and Rust1993) need not involve any heightened explicit national identity attitudes, such as national pride, or even national identification and attachment. Indeed, our mediation analysis lends support to our assumption, which is in line with previous findings on implicit nationalism (cf. Olsson Reference Olsson2024). Admittedly, however, our study was limited in that it did not include a measure allowing us to conclude that re-categorization actually took place. We encourage future studies to fill this gap, for example, by asking to what extent respondents see their political opponents as co-nationals or more implicitly testing whether they prefer terms like ‘we’ rather than ‘they’ in the treatment condition (cf. Gaertner, Dovidio, Anastasio et al. Reference Gaertner, Dovidio, Anastasio, Bachman and Rust1993, pp. 16–18). We would also like to raise the possibility of an alternative mechanism, based on the observation that national symbols may prompt not only re-categorization but also remind us of the norms and stereotypes stipulating how a ‘typical’ national should behave (Breton Reference Breton2015; Dawkins and Hanson Reference Dawkins and Hanson2024; Siegel and Badaan Reference Siegel and Badaan2020). This suggests that, in some contexts, national symbols could remind people of a national norm that emphasizes getting along and overcoming division – whereas in other contexts, where the norm is to be more uncompromising, national symbols may not have the same depolarizing effect. Perhaps such a difference in national norms could have something to do with the difference we find between Sweden, where the flag does dampen social distancing significantly, and Denmark, where the effects were less systematic? While some suggest it is in decline, Sweden, in particular, has been widely known as the model for consensus-oriented, open, rational, and deliberative policymaking for decades (Bergman and Strøm Reference Bergman and Strøm2011; Pierre Reference Pierre and Pierre2016). Although systematic comparative studies of national norms are scarce, previous research suggests that Denmark’s political culture is less consensual than Sweden’s, as evidenced by the Danish parliamentary model being the one among the five Nordic countries that most closely resembles the Westminster ‘debating parliament’ model (Bergman and Strøm Reference Bergman and Strøm2011). This is arguably most clearly reflected in political and public debates about non-Western migrants – particularly Muslims – where Danish rhetoric is notably direct and harsh in comparative perspective, including compared to Sweden (Simonsen Reference Simonsen2021).
It could also be that this difference is due to the Danish flag being even more ubiquitous than the Swedish one, as it appears in non-political and commercial contexts in ways unmatched by its Swedish counterpart (Eriksen and Jenkins Reference Eriksen and Jenkins2007). This omnipresence may dilute its political salience and symbolic potency as a common in-group priming tool. Perhaps nationalism cannot be too banal in order for it to retain its unifying potential?
In any case, further comparative research is essential to better understand the contextual scope conditions that shape when, how, and under what circumstances national flags influence inter-partisan relations, and to clarify whether the recategorization mechanism they may activate is generalizable across diverse settings. Such research should also investigate alternative mechanisms, such as how dominant cultural values shape flag associations, how historical events influence perceptions of the flag, and how current events affect its meaning (cf. Becker, Butz, Sibley et al. Reference Becker, Butz, Sibley, Barlow, Bitacola, Christ, Khan, Leong, Pehrson, Srinivasan, Sulz, Tausch, Urbanska and Wright2017). Nonetheless, at this point, there is no strong reason to believe that the results of this study only have relevance within a Nordic context. Multiparty systems are prevalent all over Northern Europe, and flags are omnipresent to various degrees. We nonetheless encourage future research to study this further in other settings.
Our efforts to examine heterogeneous effects based on the strength of explicit national-identity attitudes and other demographics revealed few pronounced subgroup differences. However, these analyses were constrained by sample size and statistical power. Further unpacking whether responses to flag exposure vary by attitudes and demographics remains an important avenue for future studies to clarify whether reduced social distancing manifests equally across different segments of society.
All in all, our study highlights the potential of national identity symbols such as the national flag to help bridge the partisan divide that increasingly haunts social relations today. We would nevertheless like to end on a note of caution. Excessive use of national identity symbols in public spaces could inadvertently contribute to sharpening divisions between national in-group members and groups whose inclusion in the national in-group is more contested (c.f. Park and Warner Reference Park and Warner2024). We cannot be certain that the strengthening of national in-group cohesion does not come at the cost of exclusion or alienation of minority groups such as immigrants and other ethnic minorities (c.f. Breton Reference Breton2015). Therefore, it is important that future research investigate whether frequent use of national flags in public and private spaces can contribute to fostering more inclusive national identities – or whether their unifying magic is reserved only for those co-nationals whose status in the common in-group is uncontested.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1475676525100649.
Data availability statement
Replication data is available at Harvard Dataverse with the title: Replication Data for: The unifying magic of the unwaved flag: Do national identity primes reduce affective polarization?.
Acknowledgements
We would like to thank our two anonymous reviewers, as well as the participants at the following seminars, for their valuable feedback: the Centre for Comparative Welfare Studies research seminar at the Department of Politics and Society, Aalborg University; the Political Sociology and Economy seminar at the Department of Government, Uppsala University; and the Comparative Politics seminar at the Department of Political Science, Lund University.
Funding statement
This research was funded by the Swedish Research Council for Health, Working Life, and Welfare (Forte).
Competing interests
The authors declare no competing interests.
Open access
