To punish or to assist? Divergent reactions to ingroup and outgroup members disobeying social distancing.

In response to the COVID-19 pandemic, societies face the formidable challenge of developing sustainable forms of sociability-cumsocial-distancing - enduring social life while containing the virus and preventing new outbreaks. Accordant public policies often balance between retributive (punishment-based) and assistance (solidarity-based) measures to foster responsible behaviour. Yet, the uncontrolled spreading of the disease has divided public opinion about which measures are best suited, and it has made salient group disparities in behaviour, potentially straining intergroup relations, elevating heated emotions, and undercutting coordinated international responses. In a 2 × 2 between-subjects experiment, British citizens (N = 377) read about national in-group or outgroup members (categorical differentiation), who were either conforming to or deviating from the corona regulations (normative differentiation). Participants then reported moral emotions towards the target national group and indicated support for public policies. In general, support for assistance policies outweighed support for retributive measures. Second, however, norm deviation was associated with less positive and more negative moral emotions, the latter category further relating to more punitiveness and less assistance support. Finally, respondents who read about norm-violating outgroup members especially reported support for retributive measures, indicating that people might use norm deviation to justify outgroup derogation. We discuss implications for policymakers and formulate future research avenues.

RCT Entities:   Population Interventions Outcomes

Background

The rapid outbreak of the COVID‐19 disease has made salient disparities in responding to the pandemic. Country governments are adopting a wide array of potentially effective containment measures, often walking a tight rope between assistance (e.g., information and sensitization campaigns) and retributive measures (e.g., penalty fees for citizens disobeying the regulations). Retributive measures, in particular, risk an unprecedented devolution of democratic rights and individual freedom to the government. To reconcile the temporary suspension of civil liberties with democratic accountability, it is crucial that the public deems retributive measures legitimate (Staerklé, Falomir‐Pistachor, Pereira, Berent, & Butera, 2015; Tyler & van der Toorn, 2013). The first aim of the current study was therefore to compare public support for retributive and assistance measures to slow down the spreading of the coronavirus. Given the negative implications for wanted rights, we expected less support for retributive than for assistance measures.

The second aim was to examine the mechanisms underlying support for retributive and assistance measures. Opinions regarding containment measures are typically informed by attributions of responsibility, blame, and deservingness (Joffe, 2011; Mondragon, Gil De Montes, & Valencia, 2017; Van Vugt & Park, 2009). Accordingly, support for containment measures is often exemplified by the lay logic: ‘Good people deserve good treatment, bad people deserve bad treatment’ (Crandall & Beasley, 2001; Lerner, 1977; Staerklé & Clémence, 2004). When people are depicted as ignoring regulations, they can be seen as a threat to the group’s safety (Brambilla & Leach, 2014), which might elicit more negative (and less positive) moral emotions. Moral emotions are defined as emotions involved in forming moral judgements and motivating behavioural responses to one’s own and others’ moral behaviour (Tangney, Stuewig, & Mashek, 2007). Positive (e.g., empathy, gratitude, and pride) and negative (e.g., anger, disgust, and contempt) moral emotions might mediate support for containment measures because they elicit moral judgements and transform moral thoughts into action. Put differently, people use a moral frame to appraise norm deviation (Brambilla, Sacchi, Pagliaro, & Ellemers, 2013), with moral emotions explaining conative responses directed towards norm‐deviating people. This emotional reappraisal could therefore relate to increased support for retributive (and decreased support for assistance) measures.

Finally, a third aim was to examine the boundary conditions of support for retributive and assistance measures. Based on recent theorizing (Staerklé, 2009, 2013), we expected two socio‐cognitive processes of differentiation to guide participants’ moral judgement: On the basis of normative differentiation, the social environment is organized as a function of norm‐conforming (good) and norm‐deviating (bad) people. On the basis of categorical differentiation, the social environment is organized as a function of in–groups (us) and outgroups (them).

Interestingly, competing hypotheses can be formulated about the interaction between categorical and normative differentiation in support for containment measures. On the one hand, the so‐called ‘black sheep effect’ (Marques, Abrams, & Seôdio, 2001; Pinto, Marques, Levine, & Abrams, 2010) states that ingroup members ignoring regulations are particularly threatening for ingroup safety and identity. From this perspective, people are motivated to punish norm‐deviating ingroup members more harshly than norm‐deviating outgroup members (Branscombe, Wann, Noel, & Coleman, 1993; Jetten & Hornsey, 2014). On the other hand, studies examining responses to infectious diseases have shown an ‘othering effect,’ whereby blame and responsibility for spreading the virus are sooner attributed to outgroups than to the ingroup (Eicher & Bangerter, 2015; Joffe, 1999; Ungar, 1998). Because people search for positive distinctiveness in relation to outgroups in any given comparative context, norm deviation by outgroup members further justifies outgroup derogation (Green et al., 2010; Joffe & Staerklé, 2007). As a result, people might be motivated to more harshly punish norm‐deviating outgroup than norm‐deviating ingroup members.

The present study

Using an embedded priming design (see Makhanova, Miller, & Maner, 2015), we investigated public support for retributive and assistance measures in relation to normative and categorical differentiation. British respondents first read about members of their national ingroup (Britons) or members of a national outgroup (Italians; categorical differentiation) conforming to or deviating from their respective government’s coronavirus guidelines (normative differentiation). Participants then wrote down examples of such behaviours and their thoughts about this group (i.e., ‘good’ or ‘bad’ Britons or Italians) before reporting (1) moral emotions and (2) support for assistance measures aimed at informing and educating the target group (e.g., sensitization and awareness programs) and for retributive measures aimed at punishing the target group (e.g., army patrols and imprisonment; for a similar differentiation, see Pinto et al., 2010).

By crossing categorical and normative differentiation in a 2 × 2 between‐subjects design, we were able to compare emotions and political attitudes exemplifying ingroup ostracism (i.e., punitiveness vs. solidarity towards ‘bad’ Britons) and outgroup derogation (i.e., punitiveness vs. solidarity towards ‘bad’ Italians) in reaction to the COVID‐19 pandemic. We focused on the United Kingdom (U.K.) and Italy for a number of reasons. First, both countries were severely affected by the disease and both governments have implemented similar containment policies (e.g., general lockdown, increased hygiene standards, and imposed social distancing). Hence, the meaning of compliance and deviance in these two national contexts is similar, justifying the normative differentiation manipulation. Second, the current U.K.‐EU tensions (e.g., concerning Brexit) have pressurised intergroup harmony between Britons and other national groups, making intergroup boundaries particularly salient. Moreover, Italy was the first European country where the COVID‐19 disease hit hard, and was readily accused of spreading the virus around Europe. The sudden spark in contagions in Italy might have made salient this national outgroup even more, justifying the categorical differentiation manipulation.

In accordance with the three study aims, we specified different hypotheses. First, as consent rather than coercion is the default standard underlying British democratic rights and liberties, we predicted less support for retributive than assistance measures overall (Hypothesis 1). Second, because pandemic threat may make some Britons (under some conditions) more punitive, we predicted increased support for retributive measures and decreased support for assistance measures in the context of norm deviation (Hypothesis 2a), with this effect mediated by an increase in negative and a decrease in positive moral emotions (Hypothesis 2b). Third, we expected interaction effects of normative and categorical differentiation for support for retributive and assistance measures (and emotional responses), but we did not have clear‐cut predictions about the direction, given the two different rationales explained above.

Method

Participants and design

British adults were recruited for this online study via Prolific Academic (https://prolific.ac). After providing informed consent, participants completed questionnaires assessing demographics, worldviews, and corona‐related behaviours (for another, unrelated study) and then completed the current study. This research was conducted according to the General Ethical Protocol of Leuven University. Information about respondent characteristics, assessment, experimental design, and supplementary analyses is available in the Appendix S1 and in the Open Science Framework (https://osf.io/ja7s4/).

The total sample comprised 399 individuals with a mean age of 32 years (SD age = 11.32), consisting of 27% men and 72% women (1 respondent indicated ‘X’ as their gender category). Education and income levels showed a fairly normal distribution. In terms of ethnic background, 85% identified as White British, and 93% were born in the U.K. (see Appendix S1 for more information). All participants had British citizenship and were currently living in the U.K., except for six participants who were omitted from further analyses.

The experiment involved a 2 (normative differentiation: scenario about norm‐violating or norm‐conforming group members) × 2 (categorical differentiation: scenario about in‐ or outgroup members) between‐subjects design. Participants were randomly assigned to one of the four conditions.

Instruments and procedure

Manipulation of normative and categorical differentiation

Participants read a scenario about a target group’s behaviour. Particularly, the vignettes read: ‘Sadly, [target country] is one of the countries in Europe the most affected by the coronavirus, both in terms of number of cases and casualties. Still, reports stated that a number of [target country] citizens are [not] doing a good job following the regulations introduced by the [target country] government to reduce the spreading of the coronavirus. Try to think about which behaviours these individuals are doing and which behaviours these individuals are not doing. Please provide some examples, and briefly describe what you think about this group of people.’

Manipulation check

After completing the open‐format question, respondents completed two manipulation checks: (1) ‘Please indicate to which national group the instructions above were referring’ (open‐ended item to check the categorical manipulation) and (2) ‘The national group referred to in the instructions above was complying with the regulations introduced by its government.’ As all other items in the survey, the latter item was measured on 7‐point Likert scales anchored by 1 (‘Strongly disagree’) and 7 (‘Strongly agree’). At the very end of the survey, a final check item tapped into the credibility of our manipulation. Respondents failing two out of three check questions (N = 16: see Appendix S1 for details) were excluded from subsequent analyses. All remaining respondents (N = 377) completed the full questionnaire, yielding no missing data.

In the survey preceding the current study, respondents indicated to what extent they themselves complied with the governmental containment policies. Responses indicated general compliance (M = 6.56, SD = 0.79) by participants in all conditions, ensuring that they considered non‐compliance with the COVID‐19 containment measures as deviant, norm‐violating behaviour.

Mediators

Three items tapped into each mediator (Hutcherson & Gross, 2011; see Table 1 for exact item wordings). Internal consistency of the positive and negative emotion measures was high (Cronbach’s αs of .76 and .90, respectively). To check the distinctiveness of the two measures, we conducted an exploratory factor analysis (EFA) including all six items. The best‐fitting model solution showed two factors, with each item loading primarily on its corresponding valence factor (see Tables S1‐S4 in Appendix S1).

Table 1

Exact item wordings of mediators and outcomes

Positive other‐based moral emotions	Negative other‐based moral emotions
1. I feel empathic towards [target country] citizens	1. I feel angry when I think about [target country] citizens
2. I am grateful to [target country] citizens	2. I condemn [target country] citizens
3. I feel proud of [target country] citizens	3. I experience disgust when I think about [target country] citizens
Support for (extreme) assistance measures	Support for (extreme) retributive measures
1. The [target country] government should deploy professional support teams during the quarantine for people who are living alone	1. The [target country] government should send the army on the streets, to check if people obey the rules
2. The [target country] government should broadcast targeted information campaigns to explain the content and the duration of the regulations put in place	2. The [target country] government should send to prison people who commit severe violations of the governmental regulations
3. The [target country] government should organize consultations with representatives of minority groups to better address the specific challenges they are facing	3. The [target country] government should ask security services to monitor civilians to detect violations of the governmental regulations

Outcomes

Three items assessed both outcomes (see Table 1). The reliabilities were acceptable (retribution: α = .77; assistance: α = .66). EFA revealed two distinct factors, indicating that support for retributive and assistance measures forms separate constructs.

Results

Descriptive analyses

Supporting Hypothesis 1, overall support for assistance measures (M = 5.73, SD = 0.85) was larger than overall support for retributive measures (M = 3.74, SD = 1.44), t(376) = 23.07, p < .001. Next, we conducted 2 (normative differentiation) × 2 (categorical differentiation) analyses of variance on support ratings for both containment measures.

Contrary to Hypothesis 2a, we found no main effect of normative differentiation on support for retributive measures (F(1, 373) = 0.32, p = .57) or assistance measures (F(1, 373) = 2.21, p = .14). Interestingly, we did find a main effect of categorical differentiation on support for retributive (F(1, 373) = 7.95, p = .005) and assistance measures (F(1, 373) = 8.81, p = .003). Participants who read, thought, and wrote about (conforming and deviating) outgroup members scored significantly higher on both outcomes than participants in both ingroup conditions (see Figure 1).

Figure 1

Mean plots (with 95% confidence interval error bars) of the effects of normative and categorical differentiation on support for retributive (a) and assistance measures (b) to contain the virus. [Colour figure can be viewed at wileyonlinelibrary.com]

Importantly, we also found an interaction effect on support for retributive measures (F(1, 373) = 5.40, p = .02), but not on support for assistance measures (F(1, 373) = 0.15, p = .70). Specifically, participants in the deviant outgroup condition endorsed retributive measures to a greater extent than those in the other conditions (Figure 1a). It is noteworthy, though, that punitiveness for such norm‐deviating outgroup members was not significantly above the scale midpoint (t(94) = 1.08, p = .28).

Model tests

To test the underlying emotional mechanisms (Hypothesis 2b), we ran a conditional process analysis using Hayes’ Process macro (2013, Model 59; 5,000 bootstrap samples). At both levels of categorical differentiation, we calculated the indirect effects of normative differentiation on policy attitudes through positive and negative moral emotions, controlling for self‐reported personal compliance. The results of these analyses are portrayed in Figure 2 (and Table S5 in Appendix S1).

Figure 2

Unstandardized results (standard errors between brackets) of the models testing the effect of normative and categorical differentiation on support for retributive (a) and assistance measures (b) to contain the virus via moral emotions. Note. *p < .05; **p < .01; ***p < .001.

Compared to participants in the norm‐conforming condition, participants in the norm‐deviating condition reported less positive and more negative emotions, for in‐ and outgroups alike (i.e., these paths were not moderated by categorical differentiation). Among both groups, increased negative emotions further related to greater support for retributive measures. Interestingly, among outgroup targets only, these negative emotions were also further associated with less support for assistance measures (i.e., categorical differentiation significantly moderated this path). In sum, we found partial support for Hypothesis 2b, as negative, but not positive, moral emotions significantly mediated the effect of normative differentiation on support for retributive measures (for both in‐ and outgroups) and on support for assistance measures (for outgroups only).

Discussion

The present research examined public support for containment measures implemented by governments to slow down the spreading of the coronavirus. In line with the idea that retributive measures are supported less because they go against core democratic values (e.g., personal freedom; Pereira, Falomir‐Pichastor, Berent, Staerklé, & Butera, 2015), results showed stronger support for assistance measures. Support for punitive policies did not exceed the scale mean in any condition, indicating that such containment measures are not well supported in any scenario. A second aim was to investigate whether support of retributive measures increases when people learn about others disobeying the government guidelines and, hence, violating norms. In line with the ‘othering’ idea (e.g., Joffe, 1999) and going against a notion of ingroup ostracism (Marques et al., 2001), retributive measures were more strongly supported for deviant outgroup members than for ingroup deviants. Allegedly, outgroup norm violators are seen as prototypical outgroup members, this ‘bad’ outgroup behaviour is more ‘internally’ attributed to group character, and negative moral emotions are generalized to the outgroup as a whole (Meeussen et al., 2012). The third aim was to examine this underlying emotional process. Results indicated that increased support for punitiveness is mediated by negative moral emotions directed towards the target group (Brambilla et al., 2013).

Intergroup relations in the face of a pandemic

Prior work on categorical and normative differentiation proved the two dimensions to be closely related to policy support (Staerklé, 2009). Yet, they have rarely been crossed orthogonally (see Politi, Gale, & Staerklé, 2017, for an exception) and never related to retributive measures directed towards ingroup and outgroup members. Extending previous findings, we showed that categorical differentiation between ingroup and outgroup members qualified public support for retributive measures during the corona crisis. Reading about norm‐deviating ingroup members elicited less support for retributive measures than reading about norm‐deviating outgroup members. This resonates with the idea that norm deviation serves more as justification for outgroup derogation than for ingroup ostracism (Abrams, Marques, Bown, & Henson, 2000; Pinto et al., 2010).

Notably, support for assistance measures showed a different pattern. Respondents consistently supported assistance measures more for outgroup than for ingroup members, regardless of whether these members were depicted as norm‐conforming or norm‐violating. Possibly, the increased support for measures aimed at informing and educating outgroup members signals a generalized perception of outgroup members as lacking competence to deal with the COVID‐19 outbreak (Fiske, Cuddy, Glick, & Xu, 2002; Yzerbyt, 2016).

Future research might address this perceived competence and effectiveness of national groups and authorities in dealing with crises to better understand public support for confinement measures and could also include a ‘pure’ control condition to examine whether assistance measures are favoured over retributive measures without priming respondents beforehand. Likely, the mere outline of conforming and deviant lines of conduct by group members might have inflated moral judgements and shaped the support for containment measures.

Another limitation of the current study is that we did not assess respondents’ national and supranational identification. Reinforcing the notion that ingroup ostracism and outgroup derogation serve an identity‐enhancing function (Packer, 2008; Riek, Mania, and Gaertner, 2006), high and low national identifiers might react differently to deviance within and across group boundaries. Politi et al. (2017), for instance, found that norm‐conforming outgroup members were particularly appreciated by high ingroup identifiers. Concurrently, Hutchison, Abrams, Guitierrez, and Viki (2008) found that norm‐deviating ingroup members were particularly depreciated by high ingroup identifiers. Different levels of self‐categorization should also be considered (Turner, Oakes, Haslam, & McGarty, 1994). Whereas feelings of nationalism should lead participants to derogate national outgroups, high Europeanism should lead to outgroup solidarity (Visintin, Green, & Sarrasin, 2018; Wagner, Becker, Christ, Pettigrew, & Schmidt, 2012).

Conclusion

To conclude, the current pandemic poses many challenges for the affected societies. Our results indicated that assistance measures to deal with these challenges are favoured over retributive measures. However, perhaps the most notable finding of this study is that support differed in reference to in‐ and outgroups, in combination with the framing of events. News about fellow nationals and other national groups breaking the ‘corona laws’ has the potential to elicit very strong negative emotions towards these norm‐violating (sub)groups. When it comes to other national groups, such emotions can further translate into a stronger endorsement of punishment‐based governmental decisions to contain the virus.

The consequences of such shifted attitudes towards outgroups should not be underestimated. As the Head of the United Nations, António Guterres, stated, the COVID‐19 outbreak is unleashing ‘a tsunami of hate and xenophobia, scapegoating and scaremongering’ (Hudson, 2020). This is evinced in the rise in anti‐Asian sentiment (Wen, Aston, Liu, & Ying, 2020), as it was 700 years ago in the increased anti‐Jew hatred during the Black Death (Cohn, 2007). It also applies to other contexts, such as media portrayals of murders and assaults by outgroup members (Meeussen et al., 2012). Moreover, when governments blame outgroups (e.g., refugees), they might opt for harsher measures, and citizens might perceive such decisions as more legitimate and potentially become more hostile towards the outgroups.

The increased nationalism fuelled by populist rhetoric paired with such outgroup blaming might readily exacerbate international tensions. It is therefore of utmost importance to avoid framing normative conduct along group boundaries, as this increases the ‘us‐versus‐them’ divide. Echoing recent suggestions addressed to political leaders to cope with the COVID‐19 pandemic (Jetten, Reicher, Haslam, & Cruwys, 2020), we advocate social identities to be framed inclusively, in order to create a sense of togetherness, and promote collective resilience and international solidarity. As an optimistic endnote, it seems that people endorse such solidarity‐based measures to a greater extent than retributive measures, even for deviant outgroups. The future will tell whether or not these stronger methods, on both sides of the continuum, will be approved, applied, and shown effective.

Conflicts of interest

All authors declare no conflict of interest.

Author contributions

Jasper Van Assche (Conceptualization; Data curation; Formal analysis; Funding acquisition; Investigation; Methodology; Project administration; Resources; Software; Visualization; Writing – original draft; Writing – review & editing) Emanuele Politi (Conceptualization; Funding acquisition; Methodology; Project administration; Validation; Writing – original draft) Pieter Van Dessel (Data curation; Investigation; Methodology; Writing – review & editing) Karen Phalet (Conceptualization; Funding acquisition; Supervision; Writing – original draft).

Appendix S1. Supplemental Online Materials.

Table S1. Fit indices of EFA on all eight emotion items (left panel) and on the six moral emotion items (right panel).

Table S2. GEOMIN rotated factor loadings for the models with best fit on EFA on all eight emotion items (left panel) and on the six moral emotion items (right panel).

Table S3. Fit indices of EFA on all six containment items.

Table S4. GEOMIN rotated factor loadings for the model with best fit on EFA on all six containment items.

Table S5. Unstandardized estimates (standard errors and 95% confidence intervals between brackets) of the total, direct, and indirect effects of priming norm‐violating (vs. norm‐conforming) behaviour of ingroups and outgroups on support for retributive and assistance measures to contain the corona‐virus, via group‐based moral emotions.

Click here for additional data file.