National minorities and their representation in Swiss surveys (II): Which practices make a difference?

This paper presents a systematic study of survey mechanisms that produce or reduce minority bias in social surveys. It extends the work of Lipps et al. (2011) who have demonstrated that in the Swiss context, the more an ethno-national minority community differs socio-culturally and socio-economically from the national majority, the less it is likely to be represented in its proper proportion in national surveys (the Swiss Labour Force Survey, the Swiss Household Panel, and the Swiss sample of the European Social Survey). Minority bias furthermore has a vertical dimension: socio-economic bias against individuals from the most deprived backgrounds is stronger within ethnonational minorities. Using data from those surveys, in the present work we empirically assess the impact of three types of survey practices to cope with bias: 1) strategies to increase overall response rates of the whole population indiscriminately from its minority status, 2) the use of pre- and post-stratification measures that take into account the overall share of foreigners in the national population, and 3) the conduct of interviews in a wider range of languages, in order to facilitate survey response among certain (linguistic) minorities. Our findings show that efforts to increase overall response rates can, paradoxically, create even more minority bias. On the other hand, they suggest that a combination of stratified sampling and a wider range of survey languages can have a positive effect in reducing survey bias, both between and within national categories. We conclude that measures that take into account and adapt to the social and cultural heterogeneity of surveyed populations do make a difference, whereas additional efforts that only replicate existing routine practices can even be counter-productive.