Parochialism, social norms, and discrimination against immigrants

We created a highly realistic and carefully controlled “microenvironment” (44) of intergroup social interactions designed to observe the degree of assistance offered to strangers, some of whom were German and others immigrants. The experiment, depicted in Fig. 1, involved the staged violation of the norm against littering in public spaces* followed by an interaction that was designed to elicit behavioral responses by unknowing bystanders (see SI Appendix, Figs. S1 and S2 for a schematic representation of the experiment). The intervention proceeded as follows: A male German confederate (violator) violates a widely held norm against littering by dropping an empty coffee cup on a train platform. A female confederate (punisher) immediately sanctions the violator by asking him to pick up his trash. The violator unwillingly but promptly complies with the request and leaves the scene. As the violator is walking away, the punisher receives a call in view of bystanders who have witnessed this previous interaction. While engaged in the call, she accidentally drops a bag of her possessions, the contents (oranges) of which disperse on the platform. The confederate appears to be in need of help retrieving her possessions, and we measure whether bystanders provide that assistance.

Two key dimensions of the intervention were manipulated experimentally. First, to vary the female confederate’s perceived membership in the ingroup (natives) or outgroup (immigrants), we randomly assigned the ethnoreligious attributes of the confederate to one of four conditions: an immigrant wearing a hijab, the same immigrant in plain attire without a hijab, the same immigrant wearing a clearly visible Christian cross, or a German female, who would serve as our control condition (skin tone and phenotype were the variables that would help the bystanders identify confederates as immigrants in the control condition). Second, we also manipulated the level of cultural integration by randomly varying whether the female confederate enforced the antilittering norm. This action signaled to bystanders that the confederate shared their norms and was a civic-minded person. In roughly half of our sample, the female confederate sanctioned the norm violator before requiring assistance. In the remaining half, a different confederate enforced the norm instead (see Fig. 2 for the matrix of treatment conditions). Apart from these two dimensions, we controlled for social class by having confederates wear similar attire across the different teams and iterations. We minimized the potential for differences in attractiveness to affect assistance rates by having the same minority confederate (immigrant female) play all three roles. Since it is not possible for the same actor to portray immigrant and native, we rotated the person playing the role of the German female confederate in each team. In SI Appendix, we also show that our results hold using team fixed effects, which analyze within-team variation in assistance rates across iterations.

One design feature of the intervention warrants note: The decision to manipulate the religious characteristics of the confederate in addition to her ethnic attributes stemmed from the central role that religious difference plays in political debates on immigration in Germany and Europe (Fig. 3). Given media coverage of cultural conflicts between European Christian and immigrant Muslim populations, we had reason to believe that religious identity would be central for the delineation of ingroups and outgroups in Germany.

The interventions were conducted in 29 train stations across three German states (North Rhine-Westphalia, Saxony, and Brandenburg). We implemented a total of 1,614 iterations of the intervention, involving 7,142 bystanders, in major train stations of these cities over a 3-wk period between July and August 2018. During each iteration, confederates were tasked with recording the behavior of bystanders who observed the intervention (coders were not blinded; see SI Appendix for more discussion). The main outcome of interest, which was coded at the iteration level, was whether any bystanders offered assistance to the female confederate in retrieving her possessions. Confederates also noted the total number and gender of bystanders within a prespecified radius, as well as other characteristics of each iteration.

The research protocol was reviewed and approved by University of Pennsylvania’s Institutional Review Board (IRB Protocol 829824). A waiver of the consent process was obtained (see SI Appendix for additional information on ethical and safety considerations).

SI Appendix provides further details regarding the design of the intervention, timelines for the study, data collection locations, and protocols for data collection, as well as auxiliary analyses. Auxiliary analyses in SI Appendix include regression-based analysis of treatment effects including state fixed effects, bystander fixed effects, and various controls; comparison of effects in East vs. West Germany; effects of language assimilation on bias toward immigrants; and an online survey to explore Germans’ attitudes toward littering and provide manipulation checks for some of our treatments. All analyses presented in the main text, unless otherwise noted, were prespecified in an analysis plan registered with the Evidence in Governance and Politics (EGAP) network (20180725AB) before commencement of data collection.

We expected that immigrants would receive less assistance than natives and that identity markers (hijabs) that increase the difference (social distance) between natives and immigrants would decrease assistance. Furthermore, since enforcement of local social norms regarding appropriate behavior signals cultural integration, we hypothesized that enforcing the antilittering norm would offset the negative bias toward immigrants. We chose this particular norm because it constitutes a noncontroversial civic norm that is particularly widespread and deeply internalized in Germany (see SI Appendix, section 7, for suggestive evidence consistent with this assumption).

First, our analyses provide strong evidence in support of our hypothesis of bias against immigrants and suggest that religious difference is what defines immigrants as an outgroup. Female German confederates, who serve as our controls, were assisted in retrieving their possessions in 78.3% of all iterations. By contrast, confederates of an immigrant background (immigrant with hijab and immigrants in plain attire) were assisted less, at 71.3% of iterations (see SI Appendix, Table S11 for further discussion of these analyses using regression with state-fixed effects and bystander fixed effects). The difference between the level of assistance offered to immigrants vs. that to natives is therefore around 7 percentage points and is statistically distinguishable at conventional levels (t = −2.11, P $<$ 0.05, two-tailed). The results shown in Fig. 4 pool across other conditions (i.e., among norm enforcers and nonenforcers within the native and immigrant groups).

The negative bias against immigrants, however, is only due to intergroup differences in religious identity. Phenotypical (ethnic) differences on their own are not sufficient to cause bias. As shown in column 3 of Fig. 4, immigrant confederates wearing a hijab, which clearly signals that they are of Muslim faith, were assisted only 66.3% of the time (these results pool across other treatment conditions, including norm enforcement). This is 12 percentage points less than assistance offered to German confederates (t = −3.22, P < 0.001, two-tailed). In the subsample of interventions from the former East German state of Saxony, which was the site of violent far-right anti-immigrant protests in August 2018, this differential increases to almost 22 percentage points (SI Appendix, Table S9). Differences in discrimination in East vs. West Germany are explored in SI Appendix, Table S7, where we show that bias is larger in the East, although we cannot establish the cause of these differences. We also find that the rise in assistance levels due to good citizenship (i.e., prosocial behavior toward immigrants due to norm enforcement) is significantly larger in the East (SI Appendix, Table S15). Economic differences between East and West, the legacy of communism, as well as differences in religiosity and in the degree of contact with immigrants could explain those results. The magnitude of this negative bias is especially noteworthy given the nature of the items dropped in the intervention; the oranges dispersed in a manner that made it seem challenging for our confederates to retrieve them expediently by themselves, which should have created strong pressures for bystanders to offer assistance regardless of the identity of the confederate.

Our analysis provides corroborating evidence for the centrality of religious differences as the basis of parochialism in our context. Specifically, the average rate of assistance to immigrant confederates in the control condition (column 3 in Fig. 4) is statistically indistinguishable from the assistance given to native Germans. Thus, we do not find any evidence of ethnically driven racism or discrimination in the context of our experiment. This is despite the fact that bystanders recognize our immigrant confederates as non-German (see SI Appendix, section 6, for evidence from a follow-up survey). Racial and phenotypical differences alone are not sufficient to induce discrimination in the context of minor everyday interactions in our sample. That might be due to the fact that immigrants in the control condition are dressed similarly to the native confederates and therefore signal some degree of cultural integration. Immigrants who wear a Christian cross (column 2 in Fig. 4) also send a signal of cultural integration. The difference between the level of assistance offered to a German confederate and that to the cross-wearing immigrant confederate is a mere 1.96 percentage points and is statistically indistinguishable from zero (t = 0.583, P = 0.56, two-tailed). However, there is a decrease of around 10 percentage points in the assistance offered to the immigrant wearing a hijab relative to the immigrant control group or the immigrant group wearing the cross (t = 3.29, P < 0.01, two-tailed). Given that the level of assistance offered to immigrants in the control and in the cross condition is roughly the same as the assistance offered to natives, we can conclude that negative attitudes toward Muslim immigrants are what drive the results in Fig. 4 and we find no evidence that racial or ethnic differences alone generate discrimination. In SI Appendix, Table S5, we conduct analyses using alternate outcome measures and show consistent results with respect to the share of bystanders who provide assistance.

Can a shared understanding of good citizenship—demonstrated through an immigrant’s enforcement of local social norms—help partially counteract the bias against Muslim immigrants? Analyses presented in Fig. 5 show that it can. In the first step, we compare between-treatment conditions wherein an immigrant confederate wearing a hijab enforces the antilittering norm (column 3) and conditions in which she does not (column 4); the mean level of assistance provided to immigrant enforcers is more than 12 percentage points higher than that to immigrant nonenforcers (we compare only native Germans to immigrants with a hijab in Fig. 5) and is statistically significant at conventional levels (t = 2.772, P < 0.01, two-tailed).

The magnitude of the offsetting effects of norms is clearly demonstrated when we examine the difference in assistance rates for immigrant enforcers and native nonenforcers. As columns 2 and 3 in Fig. 5 demonstrate, norm enforcement brings the assistance provided to an immigrant confederate within 0.4 percentage points of that to a native who did not enforce the norm; the difference is statistically indistinguishable from zero (t = 0.093, P = 0.926).

While the evidence presented above highlights the capacity of cultural integration to counteract bias against immigrants generated by ascriptive (religious) differences, it nonetheless suggests that norm enforcement is unable to eliminate the bias in its entirety; the t test between assistance offered to native (column 1 in Fig. 5) and immigrant (column 3 in Fig. 5) enforcers reveals a substantial difference of 11 percentage points, which is statistically significant at the P $<$ 0.05 level (t = 2.211, P $<$ 0.05, two-tailed).

A common social identity can serve as the foundation of democratic citizenship. Nationalism often creates that common identity, but in an increasingly multicultural world, adherence to national identities is supplanted by parochial attachments. Our study, set against the backdrop of increasing intergroup conflict between native and immigrant populations in Germany, was designed to measure the effect of shared norms on discrimination against immigrants and minorities of immigrant background. Theories of social psychology and political science suggest that a common set of norms can unify individuals who would otherwise be divided by their ethnic or religious identities. Importantly, we explore voluntary adherence to norms rather than forced assimilation by immigrants into the native society. Norms need not reflect the one-sided imposition of cultural practices that immigrant communities find foreign or threatening to their own identities; rather, they could emerge as the result of a dialogue and mutual adaptation by host and immigrant communities. Our experiment is focused on measuring responsiveness to the norm against littering—a norm that is both broadly shared among Germans and not inherently antithetical to cultural values or religious practices of immigrant groups. Adherence to this norm by immigrants signals that they care about their local environment and that they consider themselves part of their German communities.

Our experimental evidence suggests that eliminating or suppressing ascriptive differences is not a necessary path to conflict reduction in multicultural societies. We find no evidence of ethnic discrimination per se. The majority of the subjects in our experiment do behave in a cooperative manner toward both Germans and immigrants and shared norms of civic behavior reduce the differences in the level of assistance offered to the two groups. Our findings suggest that norms can form the basis for the reduction in discrimination and improved cooperation. We present evidence that good citizenship among socially distant immigrant minorities is not discounted by the majority population. Yet we also find that the effectiveness of norms in forging integration is constrained by the salience of intergroup differences. In the German context, religious difference increases social distance between native and immigrant populations, but that distance is not insurmountable. Public policies and political rhetoric have heightened the salience of religious markers so that cultural practices that are perceived foreign increase conflict. It is an open question whether societies will evolve to embrace the pluralism that is the inevitable consequence of ongoing macrolevel processes of globalization. Leaders and opinion makers could help by redirecting the public’s attention from the ascriptive differences that divide groups to the importance of cultural integration as the way to promote cooperation.

We thank Loukas Balafoutas, Carsten De Dreu, Eugen Dimant, Thad Dunning, Don Green, Guy Grossman, Dan Hopkins, Brendan O’Leary, Eunji Kim, Dorothy Kronick, Marc Meredith, Nikos Nikiforakis, Anne Norton, Paul Sniderman, and Nan Zhang; participants at the Penn Identity and Conflict Laboratory Working Group; Workshop for Norms and Behavioral Change (NoBec) 2018 at the University of Pennsylvania; and the Cultural Transmission and Social Norms (CTSN 3) workshop at Massachusetts Institute of Technology for valuable comments and suggestions. We are also grateful to our excellent team of 34 confederates and enumerators for their assistance in the implementation of the experiment.