Reconciling the contact and threat hypotheses: does ethnic diversity strengthen or weaken community inter-ethnic relations?

Abstract

The literature on whether community diversity has a positive effect on individuals' inter-ethnic attitudes (contact hypothesis) or a negative effect (threat hypothesis) remains inconclusive. Most studies infer mechanisms of contact or threat based on the relationship between diversity and mean levels of prejudice in a community. We suggest that both processes of threat and contact may be occurring with increasing diversity. By applying a measure of individual-level contact, this paper demonstrates that increasing community diversity does have a negative effect on inter-ethnic attitudes but only among individuals without inter-ethnic ties. Among those who do form ties, increasing diversity has no effect – that is, contact moderates the negative effect of community diversity. However, this relationship is further moderated by levels of disadvantage in the community. This paper has important implications for the use of the contact/threat hypotheses in studies of contextual diversity and the wider debate on rising diversity in the UK.

Keywords:

• ethnic diversity
• prejudice
• contact hypothesis
• threat hypothesis
• contextual effects

Notes

1. Differentiated into ‘symbolic’ (Stephan et al. 1998) and ‘realistic/perceived’ threat (Esses, Jackson, and Armstrong 1998).

2. Crucially, Laurence (2011) did not disaggregate by ethnicity. This is vital considering that the threat and contact hypotheses are posited to be stronger among superordinate groups

3. A fifth option, ‘don't know’, was recoded as ‘neither agree nor disagree’. However, coding these individuals as missing did not change our substantive findings.

4. This question was only asked in the 2008–09 Citizenship Survey (not our current data set) and correlates at r = 0.23 with the measure of ‘respect ethnic differences within the community’.

5. The CS did not ask these questions of 2,632 white British individuals who reported that ‘their local area was all the same ethnic group’ as themselves. Therefore, an individual's response to these questions will be based on actual experience of living in an (at least marginally) diverse community. This acts as a more robust test of whether actual exposure to greater proportions of out-group within a community affects inter-ethnic relations between residents

6. Our final analytic sample of White British individuals (N = 4,504) is thus the full sample (N = 8,030) minus a combination of those in ‘self–rated homogeneous areas’ (n = 2,632), those with missing values for ‘inter-ethnic contact’ (n = 662), and those with missing values on the remaining dependent and independent variables

7. As our measure of inter-ethnic ties is not community specific, an alternative interpretation is that individuals experience more contact in diverse communities because such communities tend to be found in more diverse local authorities (LA) and counties. However, if we enter percentage of non-White British at the LA and county level into Model 2 to control for this, we find that the coefficient for community percentage of non-White is reduced (coef of the% non-White British in the community (MSOA) = 0.229 p-value <.001) but remains strong and highly significant – that is, community diversity still matters

8. We substituted our MSOA diversity measure for the ‘self-rated measure of diversity’ and find the interaction term is again strong, positive and significant (0.195 p-value <.001).

9. Neither did we find a significant interaction between percentage of non-white British and ‘working class’.

10. Again, we did not find significant interactions with ‘working class’

11. A likelihood ratio test between Model 3 and Model 2 (Prob > χ² = 0.000) suggests that adding the three-way interaction term into the model provides a significantly better model fit than simply controlling for the contact * diversity interaction.

12. The predicted scores for above 60 per cent non-white British at low levels of ‘socio-economic disadvantage’ should be treated with some caution as in reality we have very few communities in such areas in our sample.

13. We also observe that percentage of non-white British does not have a significant negative effect (regardless of contact) in ‘medium’ ‘socio-economically disadvantaged’ areas (i.e. quartile ‘2 of 3’).

Additional information

Notes on contributors

James Laurence

JAMES LAURENCE is a Post-Doctoral Research Associate at the Institute for Social Change at The University of Manchester.