Surveys have become an indispensable source of information on social and political circumstances in modern societies, with high demands on data quality.

Unfortunately, together with decreasing response rates and increasing fieldwork efforts, the higher requirements regarding data quality lead to surging survey costs. In past years, this has led to a fast increase of online surveys. However, such surveys have an upper limit regarding the questionnaire length that is markedly lower than in face-to-face surveys. Thus, approaches are needed which allow for compressing longer questionnaires into shorter online survey formats. Such an approach is the modular questionnaire design by which not all questions of a questionnaire are presented to all respondents. Instead, respondents only receive parts of the complete questionnaire, the so-called modules. Missing values resulting from questions which are not received by the respondents are imputed.

The aim of this project is to investigate whether modular questionnaires and imputation methods leading to shorter online surveys could be an alternative to large-scale, face-to-face surveys in social science. In particular, it asks whether such an alternative approach leads to high-quality data (in terms of reliability and validity) using appropriate imputation methods and models at reduced costs. In this project, we develop and evaluate modularization and imputation strategies based on different criteria relevant for social surveys. In particular, the strategies are compared by applying a comprehensive Monte Carlo simulation study. This Monte Carlo simulation study is conducted on the basis of existing data from an online survey, the German Internet Panel (GIP).