Comparing Unit Nonresponse Patterns Among Immigrant-Origin and Native Voters in a Register-Based Local Survey

Abstract

Survey response rates are often lower among immigrant-origin populations than among natives. Using data from the longitudinal IMGES II, conducted in the western German metropolis of Duisburg during the 2021 national election campaign, I compare patterns of unit nonresponse among immigrant-origin and native voters. The probability-based two-stage design survey includes a postal screening in which target persons were asked to participate in a three-wave CATI panel. In the analyses, I use data from the sampling frame and survey response as well as supplementary information in order to model response behavior throughout the survey mobilization. The results show that, except for binary gender on the individual level and the share of foreigners on the neighborhood level, coefficients point in the same directions and reveal similar levels of precision. Higher COVID-19 incidence rates in neighborhoods are associated with lower response propensities (up to 11%, min. to max.). Higher local turnout (up to 6%, min. to max.) and formally higher educated neighbors (up to 30%, min. to max.) are positively associated with survey participation rates for both natives and immigrants. From the screening into the first CATI wave, political interest (up to 43%, min. to max.) is a positive predictor of survey participation. This demonstrates how individual characteristics and local contexts are associated with survey participation among subpopulations alike. This may help future survey projects to anticipate the extent of and processes leading to unit nonresponse in similar designs.