Decomposing educational disparities between immigrants and natives in Oslo: how gender, parents, and place matter

ABSTRACT

Identifying the determinants of social mobility for children of non-Western immigrants is of vital concern as immigration pressures grow across the West. Prior studies have minimized the role of place as one such determinant but suffers from several unsupportable assumptions. We interrogate this issue by quantifying the relative contributions made by family and neighborhood characteristics in explaining the educational attainments of young adults, paying special attention to how effects differ by gender and parental immigrant status. Our statistical models provide plausibly causal estimates by controlling for geographic selection bias. We estimate model parameters using several cohorts of children raised in Oslo, Norway, for whom government registers provide exceptionally detailed information about cumulative exposures of children to family and neighborhood contexts. Our decomposition demonstrates that differences in childhood neighborhood have a substantial effect on male (but not female) immigrant-native gaps in the probabilities of completing secondary school and of enrolling in university, though these effects are smaller than those reflecting differences in family background. We discuss implications of our findings for people- and place-based social policies aimed at enhancing immigrant opportunities.

KEYWORDS:

• Immigrants
• educational attainment
• neighborhood effects
• decomposition
• gender

Disclosure statement

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

Notes

1 This term is standard in Norwegian research; see Hermansen, Hundebo, and Birkelund (2022).

2 Here we focus only on studies pertaining to immigrant descendants’ educational attainments in Northern Europe since findings are likely contingent on the macro context of a comprehensive social welfare state. A few studies examine the impact of immigrant concentration on other educational outcomes in a Northern European context. Brännström (2008) finds that the percentage of neighbors receiving social assistance was negatively associated with the secondary school grades of Swedish immigrants, especially boys. Åslund et al. (2011) find that higher shares of highly educated co-ethnic neighbors spurred immigrant children’s academic achievement. Nieuwenhuis et al. (2017) find stronger internalization of educational commitment norms by non-Western immigrant (but not native) youth in Utrecht if they lived in neighborhoods with at least 10 percent immigrant families, but with diminishing returns if the neighborhood exceeded 20 percent.

3 Fleischmann et al. (2013) find that the numbers of co-ethnics, university-educated co-ethnics, and home-owning co-ethnics in the neighborhood are positively associated with immigrant descendant educational attainment. Hällsten and Szulkin (2009) and Hermansen (2016) employ neighborhood fixed effects, so the ‘active ingredients’ of place cannot be discerned.

4 Hermansen (2016) distinguishes exogenous and endogenous family characteristics that are unaffected and potentially affected by neighborhood, respectively. He argues that excluding the latter set provides an upper-bound estimate of the explanatory power of neighborhood fixed effects.

5 Hermansen (2016) stratifies by gender but not immigration status in his models.

6 Hermansen (2016) makes this assumption explicitly.

7 A single parent is defined as a family if one or more children are co-resident.

8 Oslo is a region of approximately one million inhabitants comprising about a fifth of the nation’s population. Specifically, we include in the Oslo region: Oslo and 10 surrounding municipalities where at least 39% (or more than 9,000) of the working-age (20-66 years) population commute to Oslo.

9 Our non-Western categorization is conventional; see: Hedman (2013); Boschman and van Ham (2015). Children of mixed native-immigrant parentage are excluded since their numbers are insufficient for analysis. Details about the national origins of our non-Western sample are presented in Appendix Table 1.

10 Most youth complete secondary school by age 19 at the earliest.

11 See, e.g., Toft and Ljunggren (2016); Hermansen (2016); Galster (2019). For details about neighborhoods used in our research, see the Online Supplement.

12 All neighborhood attributes are measured as of January 1.

13 Typically, this is the move before children enter elementary school or secondary school.

14 We do not report on the second component of the Blinder-Oaxaca decomposition, which measures the contribution of the same characteristics having different effects (i.e., different coefficients) between immigrants and natives due to its ambiguity of interpretation in this application; for the full decomposition, see Online Appendix Table 14. The second component of the decomposition is typically used to explore inter-group wage differentials, wherein it is plausible to assume that external discrimination was the only reason one might observe a smaller coefficient for, say, years of education for the lower-income group, all else equal. Such lack of ambiguity does not apply here inasmuch as there likely are immigrant-native differences in how children respond to familial and contextual circumstances, independent of external constraints related to discrimination in educational contexts.

15 By contrast, Fekjær and Birkelund (2007) find that a larger share of immigrant students in secondary school (and, by implication, neighborhood) was associated with Oslo natives’ higher grades and going on to university.

16 On the contrary, Bygren and Szulkin (2010) find that the years of education attained by immigrant descendants who finished compulsory school in Sweden was negatively correlated with the average number of young co-ethnics residing in their neighborhoods during late adolescence, unless those neighbors themselves were successful in school. Sykes and Kuyper (2009) find Dutch immigrant children’s achievement was not affected by immigrants in the neighborhood.

17 See Appendix Table 7 for a correlation matrix of observed parental and neighborhood characteristics.

18 These models cannot control for selection. There are 12,164 children in our cohorts who did not move in the Oslo region between the ages of 1-18: 92.6% natives and 7.4% immigrant descendants. Of the descendants, 733 have a non-Western background, evenly distributed between men and women.

19 This interpretation is also consistent with our decomposition revealing that giving immigrant descendants natives’ coefficients would widen the attainment gaps, especially for university enrollment; see Online Appendix Table 14.

20 See the review in Heath, Rothon, and Kilpi (2008).

21 For a recent international review of social mix policies, see Galster, Turner, and Santiago (2021) and the special issue it introduces.

Additional information

Funding

The research was supported by the Norwegian Research Council grant number 302528.