A Note on the Measurement of Generalized Trust of Immigrants and Natives

I realize that the term ‘native’ may be taken to refer to the indigenous population in some contexts. In this paper, however, the term is meant to refer to the majority population residing in a given country of residence of immigrants. In Denmark, the country studied in this paper, this effectively means ‘native Danes’.

To take into account the fact that the children are clustered within schools (see the section on data and measures) and hence that within cluster correlation of errors may occur, I clustered the analysis at the school-level. Obviously the children’s parents are not in school, but they may share other characteristics correlated with having children in the same school and so I also cluster the analysis at the school-level for this group.

As pointed out by Rensvold and Cheung (1998) the choice of referent indicator in multiple group measurement models builds on the a priori assumption that the factor loadings for this specific item is equal across groups. However, this assumption cannot be tested and if it proves wrong, the test of metric invariance may be misleading. While it is generally difficult to accommodate this potential problem, results from a recent simulation study (Johnson et al. 2009) found that it has little effect of the conclusions drawn in overall scale-level tests of metric invariance (as opposed to test of factor loadings of specific items). Given that I only apply scale-level tests of metric invariance in this analysis, the results of the tests should be insensitive to the potential problem of the chosen referent indicator being invariant across groups.

It should be noted that the definition of first and second generation immigrants employed by Statistics Denmark also includes refugees and asylum seekers.

The sample was stratified by the share of non-Western immigrants in the schools (in the last three grades) in order to secure representation of different segments of immigrants in Danish society. Pupils in the 8th and 9th grade were slightly oversampled. The response rate of the survey was 59.2 per cent for children and 53.0 per cent for parents. The number of respondents having given valid responses to all three trust questions were 3,224 for the children (956 Danes, 553 Turks, 528 Iraqis, 670 Ex-Yugoslavs, and 517 Lebanese/Palestinians) and 2,811 for the parents (958 Danes, 439 Turks, 461 Iraqis, 521 Ex-Yugoslavs, and 432 Lebanese/Palestinians).

In the case that the selected parent did not wish to or was unable to participate in the survey, it was accepted that the other parent participated. Only immigrant parents, who indicated that they were immigrants, refugees or marriage migrants, were included in the analysis. While both child and parent responded in the majority of the families, only one of the two responded in some families.

All trust differences between immigrants and Danes are significant at the 0.01-level or better for both parents and children.

This is witnessed when the primary language of the telephone interview is compared with how the interviewer assessed the respondents’ language skills.