Affective partisan polarization is on the rise. Scholars first documented affective partisan polarization in the US (Iyengar et al., Reference Iyengar, Sood and Lelkes2012; Doherty et al., Reference Doherty, Kiley and Asheer2019), but similar trends exist in countries across the world (Westwood et al., Reference Westwood, Iyengar, Walgrave, Leonisio, Miller and Strijbis2018; Boxell et al., Reference Boxell, Gentzkow and Shapiro2020; Helbling and Jungkunz, Reference Helbling and Jungkunz2020; Reiljan, Reference Reiljan2020; Wagner, Reference Wagner2021). These trends are worrisome in light of recent scholarship indicating that affective partisanship has divisive implications for many political and nonpolitical situations. The core idea is that partisan group identities trigger in-group and out-group thinking (Tajfel et al., Reference Tajfel, Turner, Austin and Worchel1979) that can extend beyond politics. For example, in-group/out-group divides are influential in dating behavior (Iyengar et al., Reference Iyengar, Sood and Lelkes2012; Huber and Malhotra, Reference Huber and Malhotra2017), economic activities (Gift and Gift, Reference Gift and Gift2015; Michelitch, Reference Michelitch2015; McConnell et al., Reference McConnell, Margalit, Malhotra and Levendusky2018), as well as personal and professional health decisions (Hersh and Goldenberg, Reference Hersh and Goldenberg2016; Lerman et al., Reference Lerman, Sadin and Trachtman2017).
The broad ramifications of affective partisanship raise questions about how far its effects may extend. Taken to the extreme, affective partisanship degrades and denies fundamental rights to the out-party, which can potentially have lethal consequences (Kalmoe and Mason, Reference Kalmoe and Mason2019). However, measuring such extreme dynamics is difficult, because there are not many situations in which people's chance of surviving is directly and publicly debated.
In this research note, we measure affective partisanship in an extreme scenario, by studying a fundamental rights dilemma that gained notoriety during the COVID-19 pandemic: the triage for access to intensive medical care. We conducted conjoint experiments in five countries (Brazil, Germany, Italy, Poland, and the United States) in August 2020. We presented respondents with the attributes of two hypothetical patients and asked which patient they would prioritize for intensive care.Footnote 1
We find that the partisanship of the patient plays a measurable role in respondents’ triage decisions. While respondents mainly rely on characteristics that are in the guidelines of medical associations, out-party animus also influences their priorities. Partisans on both the left and right are likely to withhold medical support from patients on the other political side. Moreover, right-party patients are more heavily penalized. We also find some evidence of in-group favoritism. Right respondents prioritize right partisan over independent patients.
This research note expands our understanding of the extreme ramifications of partisan polarization. Medical associations have clear guidelines for triage decisions and most agree on “the role of justice and benefit maximization as core principles” (Jöbges et al., Reference Jöbges, Vinay, Luyckx and Biller-Andorno2020, p. 948).Footnote 2 Indeed, recent studies show that public opinion seems to side with the utilitarian core of the guidelines (Kneer and Hannikainen, Reference Kneer and Hannikainen2020). Given the prominence of these utilitarian concerns, medical triage is a least likely case for discrimination along partisan lines. Our findings of partisan discrimination are therefore all the more striking. Previous research finds that partisans are willing to dehumanize members of the out-party (Martherus et al., Reference Martherus, Martinez, Piff and Theodoridis2021) and to rationalize physical harm against opponents (Kalmoe and Mason, Reference Kalmoe and Mason2019), all of which has worrisome implications. Our study adds to this line of research by measuring out-group animus in an extreme non-political setting, and expanding our understanding of partisan affective polarization. Moreover, this is one of the first empirical studies to analyze non-political consequences of affective partisanship in a comparative perspective, suggesting these are broad dynamics that exist across the globe.
1. Affective partisan polarization during the COVID-19 pandemic
The COVID-19 pandemic provides a unique opportunity to study cross-national differences in consequences of affective partisanship. The pandemic is a global problem that appeared on the social and political agenda around the world. In addition, several studies have found clear partisan divides in beliefs about the risks and development of the pandemic (Allcott et al., Reference Allcott, Boxell, Conway, Gentzkow, Thaler and Yang2020; Bhanot and Hopkins, Reference Bhanot and Hopkins2020), which suggests partisan divides might play a role in how people evaluate pandemic-related issues. Yet, partisan responses to the pandemic are not uniform across countries. In the US, the Republican party has largely opposed strict lockdown measures, and studies suggest Republicans have been less likely to comply with social distancing measures (Clinton et al., Reference Clinton, Cohen, Lapinski and Trussler2021; Gadarian et al., Reference Gadarian, Goodman and Pepinsky2021). In contrast, Poland's right-wing PiS has supported strict lockdown measures (see supplementary material (SM) 2.3.2). The COVID-19 pandemic is an opportunity to examine how affective partisanship may have different implications in different national contexts.
The COVID-19 pandemic also raised acute moral dilemmas about who deserves access to life-saving medical care. Providing life-saving support to all patients is a moral imperative, but it can be unattainable because of limited medical resources. Throughout the COVID-19 pandemic, all countries debated how to handle potential medical resource shortages, which made the issue of how to triage both prominent and salient. This allows us to study extreme partisan polarization in a realistic real-world scenario.
Clearly, most people will never need to make medical triage decisions because they are not qualified medical professionals. However, we pose our survey questions as measures of moral judgments. In that sense, the real-world scenario of medical care during COVID-19 is a more realistic and accessible scenario than, for example, the question of whether to divert a trolley from killing people, which is extensively used by psychologists to study moral judgments (Greene et al., Reference Greene, Sommerville, Nystrom, Darley and Cohen2001).
2. A conjoint experiment on triage decisions
Case selection and samples
We conducted a conjoint experiment in five countries (Brazil, Germany, Italy, Poland, and the US) in August 2020. The country selection is motivated by four interrelated factors: infection trajectory, party system, incumbent party, and policy responses toward the pandemic. At the time of our data collection, the five countries had different trajectories and experiences with COVID-19. The numbers of infected and deaths were relatively high in Italy, the US, and Brazil, while Germany and Poland were much less affected. Second, we compare the two-party system of the US to the multi-party systems in the other countries. Third, during our study period the governing party in the US, Brazil, and Poland had a right-wing leadership, whereas Germany and Italy were governed by centrist coalitions. Fourth, we observe the policy stance toward COVID-19, where our sample displays variation with respect to position of right parties (see SM 2.3.2). The survey sample includes 6415 individuals from the five countries, recruited from a commercial access panel using quotas to reflect the age, gender, and education distribution of the adult population.Footnote 3
Conjoint experiment and attribute dimensions
The task is framed by the notion that an increasing number of COVID-19 cases could overwhelm the healthcare system and medical staff would have to decide which patients should receive intensive medical care. It then presents respondents with two patient profiles and asks them which of the two they would prioritize.Footnote 4 The patients’ profiles include age (at four different levels: 21, 42, 61, and 76), gender (female and male), whether the patient has children or not, occupation (unemployed, cook, professor, physician, and nurse), arrival at the hospital (first, second, at the same time), the survival chance of the patient (20, 50, and 80percent), and the partisanship of the patient (left partisan, right partisan, none). Respondents see four patient pairs. The attribute levels are randomized, except for the arrival at the hospital.Footnote 5 The order of the attributes is randomized over respondents but kept constant for the four decisions.
The survival chance of the patient most directly aligns with utilitarian ethical guidelines, according to which patients with higher survival chances should be prioritized.Footnote 6 Other aspects are in some of the guidelines: the age of the respondent (younger patients should be prioritized), the job of the patient (patients with “system-relevant” jobs are prioritized), children, and the arrival at the hospital (“first-come-first-serve” principal). Based on these, we derive hypotheses of how the patient profiles influence respondents’ priorities. The full list of hypotheses is described in the pre-analysis plan (see SM 6).
The attribute that we use to test the implication of the affective partisan polarization theory is the partisanship of the patient. We choose different partisan attributes in the five countries. In the bi-partisan system of the US, we use Republican, Democrat, and none. In other contexts, in addition to a non-partisan level we selected parties that stand on the left and the right of the political spectrum. For Germany, we choose the Alternative for Germany (AfD) and the Alliance90/The Greens. For Poland, the Law and Justice Party (PiS) and the Civic Platform (PO). For Brazil, the Social Liberal Party (PSL, Bolsonaro) and the Workers’ Party (PT, Lula), referring to the presidential candidates. For Italy, the Lega and Italia Viva. The supplementary material provides more details on the selection and information about the ideological and lock-down positions of the parties (SM 2.3).
The effects of the partisan attributes of the patients should be pronounced among respondents who are affectively polarized. In line with previous research (see e.g. Druckman et al., Reference Druckman, Klar, Krupnikov, Levendusky and Ryan2020; Wagner, Reference Wagner2021; Harteveld, Reference Harteveld2021), we measure affective polarization using rating-scores for the different parties. The survey contains a battery of 11-pt. rating scores for the different parties. The difference between the rating scores for the left party and the right party serves as our measure.Footnote 7 We also describe the results when using a partisan identification question instead of the rating score-difference in SM 5.4.
To evaluate the effect of patient attributes on respondents’ priorities, we present the marginal means (MM) and the average marginal component (interaction) effects (AMCIE) of the prioritization for intensive care. We present results country-wise and pooled, combining all respondents (weighting for sample size). We estimate the MM and AMCIE using linear regression models and clustered standard errors.
Figure 1 shows the marginal means for the different attributes in the five countries, and the sample size-weighted pooled effects. The results reveal that respondents’ decisions broadly align with the medical guidelines. First, the core criteria of the guidelines—the chance of survival—plays a central role in the decisions. With 54% selection probability a patient with an 80percent survival chance has a 10pp (percentge points) higher chance of being prioritized compared to the 44% selection probability of a patient with 20percent chance of survival. This difference is particularly pronounced in Germany (23pp), Italy (14pp), and Poland (10pp). Second, the age of the patient is crucial. Respondents generally prioritize young patients: they prioritize patients with an age of 76 only 41% of the times, but young patients with an age of 26 56% of times. In Italy, this difference is particularly large with 27pp. Third, respondents prioritize patients with system-relevant jobs but the effects are relatively small.Footnote 8 Nurses and physicians roughly have a 6pp higher chance compared to cooks. Fourth, respondents prioritize patients with children (45% versus 54%) and slightly favor female patients (51%). Finally, arriving second at the hospital decreases the selection chances from 55% to 45%, with particularly strong differences in the US (17pp).
The results further show that when averaging across all five countries, the partisanship of the patient has a small influence on the priorities of the respondents. Across all countries, respondents prioritize a patient with no partisanship with 53%, left partisans with 50%, and right partisans with 46%. The difference varies between the country samples. While there is no difference in the US sample, this pattern is more pronounced in Germany, Italy, and Poland. In Germany, a AfD patient is prioritized only 43% of the times. It has to be pointed out that given the varying context, size, tradition, and supporter group of the right and left parties in our sample, these average differences are difficult to compare across context. However, the results show that compared to the other differences described above, patients’ partisanship is less relevant on average. When asking respondents in an open question format what attributes should matter, only 1–2.5% name partisanship. A majority names survival chances as a important feature (see SM 5.6). Furthermore, an analysis of attribute importance using random forests in the supplementary materials reveals that age, children, and the survival chance seem to matter most (see SM 5.7).
To evaluate the partisan affective polarization in triage decisions, we present the average marginal component interaction effects conditional on the difference in rating scores. Figure 2 shows the effects among respondents with low, middle, and high rating score difference between the respective right and left party.Footnote 9 The panel in the top left reveals out-party animus. Affectively polarized left respondents, who give a low evaluation of the right parties and a high evaluation of the left, downgrade patients with a right partisanship by 16pp compared to patients with no partisanship. The same left respondents do not prioritize patients with a left partisanship. For affectively polarized right respondents we find they are 5pp less likely to prioritize left partisan patients but 5pp more likely to prioritize patients with right partisanship compared to independent patients. Respondents in the middle slightly prioritize patients with no partisan attachments (52%) to both left (49%) and right partisan (48%) patients, resulting in ACMIE of −3pp for the left and −5pp for the right. The pattern could be due to the fact that independents dislike partisans and therefore downgrade both groups (Klar et al., Reference Klar, Krupnikov and Barry Ryan2018).
The country results mimic these general patterns. In the US, respondents with high ratings of the Democratic party and low ratings of the Republican party downgrade Republican patients with 10pp. We find in-group favoritism among Republicans, with polarized Republican supporters prioritizing Republican patients (7pp) compared to non-patients. There is no clear difference between polarized Republican supporters when comparing Democratic patients to patients with no party affiliation. In other country contexts, out-party animus toward the right partisan patients is even more pronounced. In Poland, Brazil, and Germany, left-polarized respondents demote patients with right partisanship with 15, 17, and 20pp. Right-polarized respondents also downgrade left patients in these countries (with 5pp in Brazil, 9pp in Poland, and 10pp in Germany), with the extend of the out-party animus effect being slightly smaller. In Germany and Poland, right-polarized respondents show signs of in-group favoritism, prioritizing right-partisan patients with 6 and 8pp compared to independent patients. In Italy, polarized respondents do not differentiate as strongly based on Italia Viva or Lega partisanship. In some context, it shows that respondents in the middle downgrade right and left-party patients (Brazil, Germany, Poland).
These patterns reveal evidence of out-party animus among affectively polarized left and right respondents and in-group favoritism among affectively polarized right respondents. Out-party animus may be particularly pronounced among left respondents because they believe right-wing partisans are less compliant with COVID regulations. This would influence the perceived deservingness for prioritized treatment, a potential driver of the decisions in moral dilemmas (van Oorschot, Reference van Oorschot2000; van Oorschot, Reference van Oorschot2006). Left-right differences in out-party animus, hence, might be a compound effect of beliefs about the compliance behavior of the patient and the partisanship itself. We cannot tease apart these two possibilities with our research design. Yet, suggestive evidence from the lock-down positions of the parties suggests it is not merely about right-wing approaches to COVID. The AfD in Germany has actively opposed stronger lock-down measures, but the PiS in Poland supported lock-down measures and called for compliance (see SM 2.3.2). Both right partisan patients’ cues trigger equally strong reactions among left respondents, suggesting that the effect is not only due to the position advocated by the patients’ party, but a general reaction toward the right party partisanship.Footnote 10 Another possibility is that partisan cheerleading effects induce differential levels of partisan responses in surveys (Bullock and Lenz, Reference Bullock and Lenz2019), but we leave this for future research to examine.
How far-reaching are the consequences of affective partisan polarization in times of the global COVID-19 pandemic? Our results suggest affective partisan polarization has nonpolitical consequences in countries across the globe. When asking which patient should receive potentially life-saving treatment for COVID-19, respondents penalize out-party patients. This reveals that out-party animus matters for difficult moral decisions when healthcare resources are distributed. The pattern is particularly strong in Poland and Germany, showing the behavioral nonpolitical relevance of affective polarization beyond the US context.
The findings address pressing issues during the global pandemic. Can the COVID-19 pandemic provide citizens with a sense of shared fate that bridges political divides in favor of a common identity? Or does the zero-sum nature of scarce health resources further increase political conflicts and inflame tribal partisan behavior? The answer to these questions is a vital ingredient for successfully combating the current global health crisis (Van Bavel et al., Reference Van Bavel, Baicker, Boggio, Capraro, Cichocka, Cikara, Crockett, Crum, Douglas, Druckman, Drury, Dube, Ellemers, Finkel, Fowler, Gelfand, Han, Alexander Haslam, Jetten, Kitayama, Mobbs, Napper, Packer, Pennycook, Peters, Petty, Rand, Reicher, Schnall, Shariff, Skitka, Susan Smith, Sunstein, Tabri, Tucker, van der Linden, van Lange, Weeden, Wohl, Zaki, Zion and Willer2020).
There is no reason to expect that partisanship matters in real-world hospital care. Medical staff typically do not know the partisan affiliation of their patients, and medical association guidelines strongly prohibit such discrimination. We are confident that these morally and emotionally difficult decisions are in the hands of trained medical personnel. Nonetheless, the hypothetical triage decision helps us understand how ordinary people think about people with different party affiliations. Our triage experiment provides a more realistic insight into the effect of affective polarization on moral judgments than other more abstract moral dilemmas like the trolley problem. The triage of patients with COVID-19 symptoms is an existing problem that is widely discussed in public and we show that affective partisan polarization can impair moral judgment in these extreme situations.
This research was generously funded by a grant from the Hertie School Faculty Activity Fund and supported by the Cluster of Excellence “Contestations of the Liberal Script” (EXC 2055, Project-ID: 390715649), funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany's Excellence Strategy. We thank participants of the ECPR Joint Sessions 2021 and the WZB colloquium “Sociological Perspectives on the Corona Crisis” for valuable feedback on the paper. We thank Johannes Latmann for excellent research assistance. Finally, we are grateful to Respondi for supporting the data collection.