Abstract
This study contributes to the growing literature on social and ethnic boundaries by tracing the impact of symbolic language boundaries in 20 European countries. Using data from the 2002–03 wave of the European Social Survey, we show that the language boundaries drawn in different host societies affect the subjective well-being of first- and second-generation immigrants. Confirming theoretical expectations, we find that symbolic boundaries only influence those individuals with limited proficiency in the majority language. Moreover, applying a distinction from theories of ethnic boundary-making, we show that it is primarily the contestedness of language boundaries within a society, rather than their strength, that influences subjective well-being. This finding provides indirect evidence for a major hypothesis: that disagreement over the social location of ethnic boundaries will make them politically salient; this, in turn, affects the well-being of individuals whose status is at stake.
Notes
[1] While we base our argument regarding boundary contestedness on Wimmer's (Citation2008b) theory, there are also alternative mechanisms—for example, highly contested boundaries might send ambiguous signals about what is expected of immigrants, thereby complicating their lives and hampering assimilation processes.
[2] Other definitions of immigrants yield substantially similar results.
[3] No information is available on respondents’ skin colour. The religious boundary measure distinguishes only between Christians and non-Christians. It therefore groups together non-believing immigrants and immigrants of a non-Christian denomination (e.g. Muslims), although the boundaries drawn against these groups are likely to differ. Moreover, the small number of non-Christian immigrants in the sample prevents the estimation of cross-level interactions. Finally, the cultural boundary dimension seems too diffuse to construct an individual-level measure, as the meaning of culture is in itself likely to be contested, both within and between different countries.
[4] Simulation-based studies show that low sample size at the higher level (n < 50) can lead to biased results in multilevel models (LaHuis and Ferguson Citation2009; Maas and Hox Citation2005). However, in contrast to the variance of random coefficients, estimates of fixed regression coefficients and their standard errors are comparatively accurate. As our analysis does not rely on random coefficients, its results should be sustainable despite the small number of countries.
[5] We include a linear and squared term for education in order to indirectly account for the curvilinear relationship between income and happiness (see Diener et al. Citation1993). As education is strongly related to income, it will partly capture effects of income that are not included in our models.
[6] In the analysis sample, 53 per cent of the respondents are female, 66 per cent belong to the first generation, 53 per cent are Christians, 7 per cent are Muslims, 2 per cent name another religious denomination and 38 per cent classify themselves as atheists/agnostics. Mean age is 44.6 years (standard deviation: 17.4) and mean years of education are 12.2 (standard deviation: 4.2).
[7] Theoretically, consensus on a language boundary against immigrants requires not only that natives agree on the importance of language (low contestedness) but also that they deem it important (strength). While there is not enough power to test for this interaction, the second requirement is fulfilled empirically, since all country medians except Sweden's lie in the upper half of the importance scale (see ).