Filling in the Gaps: Designing and Imputing Missing Data in Split Questionnaire Surveys

In light of challenges such as declining response rates and rising data collection costs, approaches like split questionnaire designs or planned missing data designs offer a promising strategy for survey projects to reduce respondent burden while still collecting data on a broad range of topics. They achieve this by administering only randomly selected parts of the full questionnaire to each respondent, effectively reducing questionnaire length for each respondent. This comes at the cost of large amounts of missing data, which must be imputed to make the data analyzable.

In this talk, I will present findings from my previous research, which explores how to effectively implement split questionnaire designs and impute the resulting data in the context of social surveys. Using Monte Carlo simulations grounded in real-world data from the German Internet Panel and the European Social Survey, I evaluate how different design and imputation choices affect the accuracy of estimates. The presentation will address key methodological questions, including how to construct questionnaire modules, how planned missingness interacts with traditional item nonresponse, and how general-purpose versus analysis-specific imputation strategies influence results. The goal is to provide practical insights and evidence-based recommendations for researchers considering split questionnaire designs in their own survey work.

Date

Venue

Registration