ACSPRI Conferences, RC33 Eighth International Conference on Social Science Methodology

Font Size:  Small  Medium  Large

Less Selectivity by Mixing Modes? Empirical Results from Two Germany Surveys

Corinna Kleinert, Michael Ruland

Building: Law Building
Room: Breakout 8 - Law Building, Room 100
Date: 2012-07-12 03:30 PM – 05:00 PM
Last modified: 2012-06-12


Telephone surveys, on the one hand, offer the advantage of being relatively cheap, but they are increasingly plagued by non-response and selectivity, in particular if they are based on address and telephone registry data. The main reasons for these problems are the declining usage of landline phones and increasing telephone advertising, often under the cover of market research. Face-to-face surveys, on the other hand, are less prone to sampling bias, at least if they use official registry address data as sampling frame, but they are often too expensive for realizing large and today often longitudinal surveys. Mixing these two modes may be an adequate answer to reduce the sources of non-response typical for telephone data collection and thus may help to reduce sampling bias. This hypothesis is empirically tested by comparing coverage and selectivity of two recent large-scale surveys in Germany, ALWA and NEPS stage 8. The IAB survey ALWA was conducted in 2007/08 in telephone mode and covered a sample of 10,000 adults born in between 1956 and 1988. These adults were followed up two years later in line with the National Educational Panel Study (NEPS). Additionally, the NEPS sample was refreshed by a sample of 2,000 new participants from the same population, but this time telephone and face-to-face interviews were mixed. We analyzed the selectivity of the two samples of first-time respondents by pursuing two strategies: first, we compared central characteristics of the groups with official reference data, the Germany Mikrozensus. Second, we estimated the probability of participation in the two groups by using information provided by the registry offices for survey participants and non-participants. The results show that selectivity in terms of age, migration background, and education is reduced considerably by mixing modes.