The Impact of Mixed Modes of Survey Administration in the Generations and Gender Survey round II: Mode Effects and the Role of Mode Assignment

Martijn Schoenmakers , NIDI

Surveys provide an invaluable resource to gain information on a wide range of participant characteristics. However, the mode these surveys are administered in may affect both which respondents answer the survey (selection effects) and how respondents answer the survey questions (measurement effects). As the Generations and Gender Survey (GGS) incorporates data from web-based interviews (CAWI) along with more traditional interviewer-led modes (CATI, CAPI), an understanding of these measurement effects is crucial to ensure data comparability. This study therefore aims to explore measurement effects in the GGS round 2. We hypothesize measurement effects will be minimal for non-sensitive demographic questions, but substantial for sensitive demographic questions such as income and education. Additionally, we expect measurement effects to be largest for subjective scales concerning sensitive topics such as depression or loneliness. We expect that the way participants are assigned to a mode of administration will influence the magnitude of the measurement effect. First, we will provide an overview of how differing modes of administration are implemented in the GGS-II. Second, we will provide a focused analysis in regions such as Taiwan, Uruguay, and France, to separate measurement and cultural effects. Analyses will be conducted using ordinal and linear regression, and tests for measurement invariance will be performed using multigroup structural equation modelling. An exploratory analysis will provide a first look at the effect of mode assignment on the size of the mode effect. This study will provide crucial insights into mode effects and inform best practices for future waves of the GGS.

See extended abstract

 Presented in Session P3. Families, Fertility, and the Life Course 3