Perceived discrimination and support for democracy among immigrants

ABSTRACT

Does perceived discrimination and exclusion promote or hinder support for democracy among immigrants? While many studies investigate the drivers of prejudice and discrimination toward immigrants, relatively less is known about the impact of discrimination on immigrants’ political attitudes. In this paper, we assess whether perceived discrimination is associated with higher levels of support for democracy among Muslim immigrants using the EURISLAM survey dataset, which includes data from immigrants from Muslim-majority countries residing in four European countries. We find that in particular, perceived discrimination toward the ethnic or religious in-group is associated with increased support for democracy. These results are robust to alternative control variables, model specification, matching procedures and coefficient stability analysis. Our findings make an important contribution to understanding the implications of discriminatory experiences for immigrants.

KEYWORDS:

• Discrimination
• democratic attitudes
• immigrants
• Muslims
• Western Europe
• political psychology

Acknowledgements

We thank Andrej Cvetic, Eser Sekercioglu, Miceal Canavan and the participants of Second Scientific Meeting of the German Political Psychology Network, Columbia University Comparative Politics seminars and Humboldt University Berlin Institute for Migration and Integration Research Colloquium for their helpful comments on the earlier versions of this paper. All errors are our own.

Disclosure statement

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

Notes

1 The dataset and the codebook are publicly available at https://doi.org/10.17026/dans-xx7-5×27.

2 A surname-based sampling method that made use of digital phone book records (including both land lines and cellular phones) was used to construct the sampling frame because statistical categories and possibilities to rely on official registries differ greatly across the countries in question (for more details, see Hoksbergen and Tillie 2016, 9–11)). Recent research underlines possible problems related to inferring gender and race identity through names (Lockhart, King, and Munsch 2023). While we acknowledge concerns, we believe the EURISLAM dataset is not likely to suffer from this problem. The countries of origin covered in the sample are Bosnia, Morrocco, Pakistan and Turkey. These countries have distinct cultures and even use different alphabets. Therefore, inferring the country of origin through names within this sample is not likely to cause threats to our inferences.

3 All items were originally measured on a 4-point Likert scale.

4 As a robustness check, we also run the analysis with an index created via factor analysis (Appendix Table A2) and separate analyses for each item (appendix Tables A3–A5). The results are still supportive of our argument.

5 As may be expected, non-natives are significantly more likely to indicate having experienced personal discrimination than natives (Natives: M = 0.14, s.d. = 0.36; non-natives: M = 0.34, s.d. = 0.47; t = 14.64, p < .001).

6 Interestingly, natives are significantly more likely to think that Muslims experience unfair treatment in the host countries compared to non-natives (Natives: M = 0.65, s.d. = 0.26; non-natives = 0.50, s.d. = .34; t = 15.18, p < .001). The ethnic discrimination question was not asked to natives in the survey, so we are not able to compare the means for non-natives and natives for this variable.

7 There are four host countries: Germany, Switzerland, United Kingdom, and Belgium as the base category.

8 In the main analysis, we used an additive index of three items. As a robustness check, we also run the analysis with a dependent variable created via factor analysis (appendix Table A2). We also run a separate model for each item (Appendix Tables A3–A5). All of these analyses point to a positive relationship between perceived discrimination and support for democracy.

9 We replicated this analysis by including the squared term of the age variable to inspect if the relation-ship between age and support for democracy is non-linear. However, neither age nor the squared term is statistically significant.

10 First-generation Muslim immigrants often have ”low expectations” of the political institutions in their destination country, possibly due to coming from countries with weaker institutional performance (Maxwell 2010; Röder and Mühlau 2012). They may thus use their country of origin’s institutions as a reference point, resulting in more favorable evaluations of their host country’s institutions, including the democratic system.

11 The results of random slopes model is reported in the appendix Table A7. Furthermore, when we run the analysis in Table 1 country by country, the results are very similar to Figure 1.

12 Personal discrimination is binary, and respondents who experienced discrimination are matched to those who did not experience it. Religious and ethnic discrimination variables are measured in four categories: never, rarely, occasionally, and frequently. The last three categories are separately matched to the first (“never”) category. First, respondents who perceive discrimination rarely are matched to those who never perceive discrimination, while those who perceive occasional or frequent discrimination are excluded. Second, respondents who perceive occasional discrimination are matched to those who never perceive discrimination, while those who perceive rare or frequent discrimination are excluded. Finally, respondents who perceive frequent discrimination are matched to those who never perceive discrimination, while those who perceive rare or occasional discrimination are excluded.

13 In other words, scholars need to assume the ratio of variation explained by observable confounders to unobservable confounders. Page (2019) suggests using proportional selections (i.e. unobservable confounders explain as much as observable confounders do). Below, we ran additional analysis with respect to this assumption.

14 In this study, short regression is defined as Support for Democracy_i= β₁Discrimination_i + ϵ_i.

15 The difference in support for freedom of speech between the natives and the non-natives is statistically different than zero (t = 2.31, p < .05)

16 In general, a more stringent test of whether people fully embrace freedom of speech involve asking respondents whether they endorse the freedom of speech of groups whose ideas they disagree with (or groups that they dislike) (Gibson 2013). However, the question is asked in very broad terms in EURISLAM survey.

17 Only about 23% of the respondents in our dataset was native born.