Do principals discriminate against school parents less when having close minority colleagues? A field experiment

ABSTRACT

Interethnic contacts are generally assumed to reduce discrimination and prejudice. According to this optimistic view – and in conflict with Allport’s well-known theory – contacts have positive effects even when they are rather distant. However, findings of recent experimental field research indirectly cast doubt on this belief. To our knowledge, this is the first field experiment thoroughly investigating close as well as more distant contacts simultaneously. In a correspondence study (n = 3015), Swedish elementary school principals were randomly contacted by fictional parents with Arabic- or Swedish-sounding names asking school-related questions. The unique design also used registry data (e.g., on ethnicity). The results support Allport’s more pessimistic view: working closely with minority members of school management was associated with lower levels of ethnic discrimination regarding important qualitative aspects of the principals’ communication with the school parents, but no such pattern was observed for more distant workplace contacts with minority teachers.

KEYWORDS:

• Contact theory
• field experiment
• close contacts
• discrimination
• interethnic contacts
• school parents

Can intergroup contacts reduce discrimination? Allport’s (1954) seminal contact theory leads us to believe so. For some time, a general optimistic belief has been that the theory holds even if the ideal – according to Allport – contact situation is not completely present (Pettigrew et al. 2011). For example, positive effects such as reduced prejudices have been believed to occur even when interethnic contact is not particularly close (cf. “casual contacts” and “acquaintances;” Allport 1954, 264–265).

Interestingly, recent findings of field experiments relevant to political science cast doubt, in passing, on this optimistic view. Studies acknowledging close social interaction do report decreasing prejudicial behaviors and attitudes (Finseraas and Kotsadam 2017; Mousa 2020), but the opposite goes for distant contacts (Condra and Linardi 2019; Hangartner et al. 2019). However, these field experiments concern vastly different social situations: Norwegian military conscripts (Finseraas and Kotsadam 2017), workers in Afghanistan (Condra and Linardi 2019), refugee-exposed inhabitants of a Greek island (Hangartner et al. 2019), and football teams consisting of Iraqi Christian refugees and Muslims (Mousa 2020). In addition, either close or distant contacts are investigated in each study, not both simultaneously. Therefore, something other than contact proximity may explain the varying findings: for example, the effects may be context dependent (see, e.g., Enos and Gidron 2016), so close and distant contacts should preferably be compared directly in similar social situations. Interestingly, a recent registry-based study found close social interaction to decrease support for a right-wing populist party more than did distant interaction (Andersson and Dehdari 2021). However, to attain evidence that is as applicable as possible, field research that more closely observes social situations would be superior.

Our study, therefore, adds important knowledge. As far as we know, it is the first field experiment simultaneously examining both close and more distant contacts. If the pattern discussed above is repeated, further evidence is accumulated that contacts should be close, as argued by Allport; if not, the prevailing and more optimistic view of contacts is supported.

Under investigation here are Swedish elementary school principals’ email replies to minority school parents. Principals were randomly contacted by fictional aliases with either Swedish- or Arabic-sounding names asking for information about choosing schools for their children. Our purpose is to analyze whether the presence of minorities among the school management and teachers, respectively, is associated with principals treating minority school parents in less biased ways. In Sweden, head principals work closely with school management members (e.g., deputy principals). Together, they handle everything from long-term planning to day-to-day problems, so close long-term cooperation is common. Contacts with teachers, on the other hand, tend to be more occasional. Although working at the same school, teachers do not participate in principals’ everyday work tasks nearly as much as do members of school management.

Empirically, it has repeatedly been confirmed that public officials do discriminate against clients (Costa 2017; Hemker and Rink 2017). Audit studies have, in particular, found individuals with Arabic-sounding names to be exposed to discrimination, in many different societal contexts, for example, when interacting with officials working in/with bureaucracy, housing, education, and political institutions (e.g., Lajevardi 2018; Larsson and Adman 2022). However, our field experiment is the first to analyze whether minority work-related contacts may prevent public officials from acting in such biased ways.

The field experiment data were uniquely combined with high-quality Swedish registry data. The latter were used, for example, to capture whether teachers and school management members had been born in the Middle East or Africa, i.e., the same general part of the world as “Arabic background” refers to in the experiment (for details, see the following section and Appendix C).

Case, design, and method

From a migration perspective, Sweden is often described as a European outlier, for example, with public opinion being particularly positive towards ethnic minorities versus in other countries. Furthermore, as a Scandinavian welfare state, this country is characterized by a fairly universal and generous welfare system.

In several ways, however, Sweden is less of an exception than it used to be, with economic equality and earnings/educational mobility having decreased rapidly in recent decades. Moreover, several waves of large-scale immigration have occurred. The population is now significantly ethnically, linguistically, and culturally diverse, with about one fifth of the population being foreign born. Moreover, the right-wing populist party, the Sweden Democrats, has become one of Sweden’s biggest parties, and the general discourse concerning immigrants – particularly those from the Middle East and Africa – has hardened and become politicized. As in European countries such as the UK, France, and the Netherlands, Muslims are found at the bottom of ethnic hierarchies, and a substantial share of the native-born population views Muslims in a negative light and considers refugees a threat to security and social welfare (e.g., Abdelkader 2017; Myrberg 2010). Field experiments with immigrants belonging to these groups report discrimination in several societal sectors, including during interactions with public officials and to some extent also politicians (e.g., Eriksson and Vernby 2020; Larsson and Adman 2022; Larsson, Åström, and Adman 2022).

Regarding the school system, today parents can freely choose schools for their children. Municipalities (i.e., local governments) have administrative responsibility for organizing and financing the elementary school system. Still, publicly funded private schools constitute a substantial share (for details on the Swedish school system, see Appendix A).

In sum, when comparing Sweden with other large immigrant-receiving Western European democracies, we see clear similarities, particularly with respect to the increasing politicization of immigrants from the Middle East as well as increasing inequalities, including in the school system. Therefore, our findings may be generalizable to such countries and to other Scandinavian countries in particular.

Moving on with the concept of ethnic discrimination, theoretically, we define it as unequal treatment based on physical characteristics (e.g., skin/hair) or cultural factors (e.g., name or language). In the experiment, ethnic discrimination was captured by randomly contacting principals via emails from fictional parents with Swedish- or Arabic-sounding names (regarding research ethics, see Appendix B). The parents claimed to be interested in placing their children at the school and asked three relatively simple but important questions regarding the school and its admission procedure (for details and complete emails, see Appendix E). The emails were written as if sent by someone considering moving to the municipality, to avoid potential suspicion in small municipalities.

A factorial design was employed and randomization was used as regards ethnicity, socioeconomic status (SES), and gender. Schools receiving the Arab treatment were not statistically different from other schools in terms of the ethnic composition of their deputy principals and teachers (for more details about the randomization checks, see Appendix F). Gender was randomized to ensure that the results were generalizable to both women and men. SES was randomized to improve generalizability and to control for the SES associations of the names in the experiment. This was done by signaling a highly skilled profession (i.e., dentist) in half of the emails and signaling a low-skilled profession in the other half (i.e., care assistant), two professions common among both immigrants and native-born individuals. In addition, we tried to ensure that all names chosen for the field experiment were associated with similar SES levels. The choice of names was based on a smaller pre-study of the SES associations of different names, using upper-secondary-school principals (for details, see Appendix C).

The email addresses from which the messages were sent were randomly divided into eight groups corresponding to the aliases and SES signals used in the emails (for treatment details, see Appendices C and E).

The coding scheme consisted of three variables. Two of these measured formal aspects, i.e., whether any reply was received within 2 weeks and how many of the three questions were answered. The third variable was an index measuring the friendliness of the principals’ replies. It captured whether the principal invited future contact (1 point), welcomed the sender to the municipality/school (1 point), invited the sender to visit the school (1 point), provided additional information not directly related to the three questions (1 point), and used the name of the sender (1 point). A dimensional analysis supported treating the friendliness items as one dimension.¹ Non-responses were coded as zero for all variables to avoid conditioning on a post-treatment variable and inducing selection bias (cf. Coppock 2019). Emails were independently coded by two research assistants following a coding form formulated by the authors of this article. The names of the fictitious email senders were removed before coding started (for more details and for the results for the individual friendliness variables, see Appendix E).

Registry data provided information on the proportion of teachers – as well as colleagues in school management – born in the Middle East or Africa (for details, see Appendix F). The registry data also allowed access to the principals’ email addresses. Our study targeted all elementary Swedish school principals with unique email addresses working at schools for which registry data were available. Only one unique email address per principal was allowed, and if there were vice principals only, the email address of the main principal was used. A few principals were working at several schools but were contacted only once. We had access to personal email addresses for 97% of the principals (results are highly similar when excluding the remaining 3%, see Table D.1). The analyses below included 3015 principals, which is close to a population study of all Swedish school principals (for more details, see Appendix D).

Results

First, we tested whether discrimination in fact occurs in this social situation, as claimed by prior studies. Models 1, 4, and 7 in Table 1 help us in this regard. Here the variable “Arab” shows the treatment effect of signing emails with Arabic-sounding names (1) instead of Swedish-sounding names (0). In line with prior European studies (Hemker and Rink 2017; Adman and Jansson 2016), for formal dimensions – i.e., response rate and number of questions answered – we find substantially small and non-significant negative effects (Models 1 and 4). Importantly, however, and in line with the abovementioned research, we observe qualitative discrimination, as the coefficient for the friendliness index is negative: responses to Arabic-sounding names are rated 0.26 points less friendly than responses to Swedish-sounding names on a scale of 0–5.

Table 1. Direct effects of the Arab treatment, and interaction effects of the Arab treatment and the proportions of teachers and of school management members born in the Middle East or Africa, on reply, questions answered, and friendliness.

	Reply			Questions answered			Friendliness index
	(1) Discrim. effects	(2) Interact. school managers	(3) Interact. teachers	(4) Discrim. effects	(5) Interact. school managers	(6) Interact. teachers	(7) Discrim. effects	(8) Interact. school managers	(9) Interact. teachers
Prop. MENA × Arab		0.265 (0.217)	0.308 (0.277)		1.017^+ (0.526)	–0.205 (0.491)		1.171** (0.408)	0.836 (0.662)
Prop. MENA		–0.263^+ (0.135)	–0.344^+ (0.176)		–0.681* (0.315)	–0.769^+ (0.441)		–0.820** (0.252)	–0.715 (0.489)
Arab	–0.011 (0.018)	–0.016 (0.018)	–0.023 (0.017)	–0.019 (0.046)	–0.036 (0.046)	–0.014 (0.050)	–0.260*** (0.047)	–0.280*** (0.047)	–0.292*** (0.053)
Constant	0.776*** (0.017)	0.780*** (0.016)	0.790*** (0.016)	1.716*** (0.046)	1.726*** (0.045)	1.745*** (0.045)	1.492*** (0.064)	1.504*** (0.063)	1.520*** (0.065)
N	3015	3015	3015	3015	3015	3015	3015	3015	3015
Adj. R^2	0.000	0.001	0.001	0.000	0.001	0.001	0.016	0.018	0.016

Notes: Ordinary least squares coefficients from regressions. Standard errors in parentheses are clustered at the municipal level. The models also include controls for the gender and SES treatments. Prop. MENA = proportion born in the Middle East or Africa (not including head principal[s]).

⁺ p < .10.

*p < .05.

**p < .01.

***p < .001.

We now turn to our main question, regarding the importance of the principals’ close (i.e., school management members) and more distant (i.e., teachers) minority contacts. Here, Models 8 and 9 in Table 1 are of particular interest. As for the former model, we find a substantially large and statistically significant coefficient for the friendliness index. When all school management colleagues were born in the Middle East or Africa, head principals responded in a more friendly way to the Arabic aliases than to the Swedish aliases (difference of 1.2 friendliness points) than did those without such colleagues. We do not find a significant coefficient concerning contacts with teachers (Model 9). Hence, close, but not distant, social interactions are associated with less discrimination against Arabic parents.

As for questions answered, the results are found in Models 5 and 6. Principals whose close colleagues were all born in the Middle East or Africa answered about one more of the posed questions, while there was no significant coefficient in the case of contacts with teachers. However, the former coefficient is significant only at the 0.10 level. As for replying at all to the emails, no significant interaction effects were found (Models 2 and 3). Hence, the importance of close contacts is most evident in the qualitative aspects of principals’ replies (for a graphical presentation of the school management interaction effects, see Figure H.1–H.3).

Worth mentioning is that self-selection could be present, for example, if ethnically tolerant principals were drawn to more diverse schools or recruited minority staff members to a higher degree than did other principals. Therefore, we have undertaken several robustness checks, enabled by our assistants’ coding of which continent each principal’s name most likely originated from (e.g., whether or not the names sounded Middle Eastern). Table 2 (Model 3) shows only a slight overrepresentation of principals with Middle Eastern- or African-sounding names at schools with more diverse managements.

Table 2. Effect of the head principal’s name on the proportion of colleagues in the school management born in the Middle East or Africa.

	Prop. MENA school management
	(1)	(2)	(3)
Woman’s name	0.003 (0.006)
Foreign name		0.042^+ (0.023)
MENA name			0.220* (0.102)
Constant	0.016*** (0.003)	0.015*** (0.003)	0.015*** (0.003)
N	3015	3015	3015
Adj. R^2	0.000	0.005	0.024

Notes: Ordinary least squares coefficients from regressions. Standard errors in parentheses are clustered at the municipal level.

⁺ p < .10.

*p < .05.

**p < .01.

***p < .001.

This overrepresentation is in itself not correlated with discrimination (Table 3, models 1, 3, and 5). Moreover, Models 2, 4, and 6 in Table 3 show that discrimination levels were unaffected by recruitment taking place after the head principals started working at the schools. The interaction effects are substantially weak – for questions answered, they are even negative – and not statistically significant (for additional robustness tests, see Appendix F).² In sum, these robustness tests do not support self-selection/recruitment having a substantial impact. If the findings were strongly driven by such factors, we would most likely see other results. However, the presence of these methodological problems cannot be completely ruled out based on these additional analyses; hence, some caution should still be exercised regarding the interaction analyses in Table 1.

Table 3. Interaction effects between the head principal’s names and the Arab treatment (models 1, 3, and 5) – and interaction effects between the difference in the proportion of the school management born in the Middle East or Africa (after the principals started working at the schools) and the Arab treatment (models 2, 4, and 6) – on reply, questions answered, and the friendliness index.

	Reply		Questions answered		Friendliness index
	(1)	(2)	(3)	(4)	(5)	(6)
MENA name × Arab	0.019 (0.018)		–0.364 (0.417)		0.062 (0.621)
Diff prop. MENA Managers × Arab		–0.109 (0.208)		0.138 (0.764)		0.521 (0.548)
MENA name	0.236 (0.016)		0.028 (0.277)		0.323 (0.363)
Diff prop. MENA Managers		0.051 (0.178)		0.064 (0.476)		–0.311 (0.411)
Arab	–0.010 (0.018)	–0.016 (0.018)	–0.018 (0.046)	–0.037 (0.046)	–0.260*** (0.047)	–0.277*** (0.047)
Constant	0.774*** (0.017)	0.774*** (0.017)	1.715*** (0.046)	1.736*** (0.045)	1.489*** (0.064)	1.489*** (0.065)
N	3015	2944	3015	2944	3015	2944
Adj. R^2	0.016	0.000	0.000	0.000	0.016	0.016

Notes: Ordinary least squares coefficients from regressions. Standard errors in parentheses are clustered at the municipal level. The models also include controls for the gender and SES treatments.

⁺ p < .10.

*p < .05.

**p < .01.

***p < .001.

Conclusion

To our knowledge, this is the first field experiment investigating close as well as more distant interethnic group contacts simultaneously. A unique design was employed, combining field experiment and registry data. As discussed above, our findings may not be completely accurate due to possible self-selection processes; however, we believe this problem to be limited, as indicated by the robustness tests.

Keeping this method problem in mind, our results are consistent with what, in passing, has been found in recent field experimental research: close contacts matter (Finseraas and Kotsadam 2017; Mousa 2020) while more distant ones do not (Condra and Linardi 2019; Hangartner et al. 2019). To be precise, this is the case as regards important qualitative aspects of the principals’ communication with school parents (as for more formal communicational aspects, some similar but not as clear tendencies were found). To conclude, our study adds real-life empirical evidence in support of Allport’s theory and contradicts the more optimistic view that most forms of contacts are beneficial (Pettigrew et al. 2011).

Interestingly, our study also touches on the longstanding debate about Allport’s (1954) equality condition (e.g., Pettigrew et al. 2011). More specifically, even though the professional relationship between Swedish head principals and other school management members is of a cooperative character, ultimately the head principal has the formal power and responsibility over decisions. Despite this inequality, we still find that management members’ ethnicity matters. Admittedly, however, we are hesitant in generalizing this finding to contexts characterized by large inequalities.

Moreover, the situation we studied bears clear similarities to several other important social situations. Our findings might be generalizable to, for example, the way staff members at unemployment and housing agencies treat clients seeking work and housing, respectively. From a policy perspective, increasing minority representation in such contexts – particularly among staff working closely with those who frequently interact with and make decisions regarding welfare clients – may lead to less biased client treatment and better conditions for societal integration in the longer term. However, before recommending such measures more generally, field experiments in these other contexts too would be preferable.

In addition, why close contacts seem to be superior to distant ones deserves more scholarly attention. It has been suggested that closer contacts to a larger extent foster feelings of guilt about one’s own privilege as well as learning about out-group members’ perspectives and experiences (e.g., Jordan, Lajevardi, and Waller 2022). Experimental designs able to capture such mechanisms would be desirable.

Finally, Allport’s theory concerns a rather extensive range of discriminatory-related behaviors and attitudes. Results may of course differ depending on the exact measure in focus (Pettigrew et al. 2011). Future studies – similar to ours – are warranted investigating whether findings are similar for other types of discriminatory and prejudicial attitudes and behaviors.

Acknowledgements

I want to thank Jonas Larssson Taghizadeh for helping me design the field experiment.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Forskningsrådet om Hälsa, Arbetsliv och Välfärd [grant number 2019-00504]; Vetenskapsrådet [grant number 2016-03070]; Vetenskapsrådet [grant number 2019-03305].

Notes

1 Using a principal component analysis with varimax rotation, the eigenvalue of the only dimension that survived the Kaiser criterion was 1.8. This dimension explained 36% of the total variance. The factor loadings varied between 0.38 and 0.52.

2 In Sweden, recruitment is often carried out by local school agencies, with school managements having limited or no influence.