Direct and indirect predictors of opposition to immigration in Europe: individual values, cultural values, and symbolic threat

ABSTRACT

The current study examines the following questions: (1) the extent to which individual basic human values are linked with attitudes towards immigration; (2) whether symbolic threat by immigration mediates this relation; and (3) whether cultural values moderate the relations between individual values, threat, and attitudes towards immigration. The empirical analysis relies on the 2014/2015 data from the immigration module of the European Social Survey (ESS) for West and East European countries. We find that universalistic individuals expressed lower threat due to immigration and higher support of immigration while conservative individuals displayed the opposite pattern. Symbolic threat mediated the association between values and immigration attitudes, but in most countries the mediation was partial. The associations between values, symbolic threat, and attitudes towards immigration were stronger in countries characterised by higher levels of intellectual and affective autonomy and weaker in countries characterised by higher levels of cultural embeddedness. The findings provide support for the centrality of human values in the formation of threat and attitudes towards immigration.

KEYWORDS:

• Individual basic human values
• cultural values
• universalism and conservation
• cultural embeddedness
• affective and intellectual autonomy
• symbolic threat
• attitudes towards immigration
• mediation
• European Social Survey (ESS)

Acknowledgements

The first author would like to thank the University of Zurich Research Priority Programme ‘Social Networks’ for their support. The authors would like to thank Lisa Trierweiler for the English proof of the manuscript. The work of Peter Schmidt was supported by the Alexander von Humboldt polish honorary fellowship granted by the foundation for Polish Science for the international cooperation of Jan Cieciuch with Peter Schmidt.

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Eldad Davidov http://orcid.org/0000-0002-3396-969X

Daniel Seddig http://orcid.org/0000-0003-1595-6948

Anastasia Gorodzeisky http://orcid.org/0000-0002-9244-8501

Notes

1 The relations between values and opposition to immigration can be seen in Figure 1 in the introduction to this special issue (Heath et al. 2020).

2 Although it may make sense theoretically, the 2014/2015 ESS data do not allow differentiating clearly between economic and symbolic threat because the measures correlate too highly (Davidov, Cieciuch, and Schmidt 2018). Cognitive interviews may shed more light about how strongly respondents differentiate between different types of threat. Running the model with economic threat items in addition to symbolic threat items produced similar results.

3 As indicated earlier, the 2014/2015 ESS data do not allow differentiating clearly between economic and symbolic threat because the measures correlate too highly (Davidov, Cieciuch, and Schmidt 2018). Running the model with economic threat items in addition to symbolic threat items produced similar results.

4 We preferred using these items to measure Allow because they also appeared in other ESS rounds, they correlated highly, and they have been shown to load strongly on a single factor, thus, tapping into a single dimension. Furthermore, previous studies demonstrated that they are measurement invariant across ESS countries (Davidov et al. 2015). Running the model with the item ‘people from poorer countries in Europe’ instead of the item ‘people of the same race or ethnic group’ produced similar results. However, no data was collected for ‘people from poorer countries in Europe’ in CZ. Therefore, this country was excluded from the robustness analysis.

5 East European countries in our sample scored significantly higher on cultural embeddedness (r = 0.86) and significantly lower on affective autonomy (r = −0.74) and intellectual autonomy (r = −0.55).

6 For this reason we decided to run multigroup analyses rather than a multilevel analysis which would not allow us to examine the associations country by country.

7 We controlled in Models 1 and 2 for age, education, gender, perceived income, political orientation, and religiosity. The control variables displayed the expected relations with younger, more educated individuals with higher (perceived) income and a left political orientation scoring higher on universalism values and support of immigration and lower on conformity/tradition values and threat. We also acknowledge here the possibility that threat and opposition to immigration may affect the values. However, we do not have panel data to examine this possibility, and therefore we try to minimise the causal language in this manuscript. A detailed output on the effects of age, education, gender, perceived income, religiosity, and political orientation may be obtained from the second author upon request.

8 In the four East European countries we observed high correlations between conformity/tradition and universalism values (CZ: r = .84; EE: r = .71; HU: r = .83; PL: r = .77). We interpreted this correlation as a multicollinearity problem (Grewal, Cote, and Baumgartner 2004). The consequence of allowing the two values to correlate in the multigroup SEM model was that the effects of the two values and their respective standard errors were severely inflated in these countries. Therefore, to circumvent this problem, we decided to omit the correlation. The effects of the values on Allow were then much more in line with our theoretical expectations and, in addition, very similar to the coefficients obtained when only conformity/tradition or universalism were used as single predictors of Allow. The difference in model fit between the two models (with and without a correlation) was minimal.

9 See Appendix B for the indirect and total effects in Model 2 for each country.

10 We examined the cross-country variation of the unstandardised effects, because metric invariance allows comparing unstandardised but not standardised coefficients (Steenkamp and Baumgartner 1998).