Project A8:

Evaluating Data Sources for Research into Political Reforms: (Non)probability Online Surveys and Big Data


Prof. Dr. Annelies Blom / Jun.-Prof. Dr. Florian Keusch / Prof. Dr. Frauke Kreuter


Carina Cornesse / Ruben Bach


Daniela Ackermann-Piek / Dr. Joseph Sakshaug / Prof. Daniel Stegmüller, Ph.D.

 Duration:01.01.2018 - 31.12.2021


The A8 project evaluates different types of data sources in terms of their suitability to research political reform preferences and attitudes in Germany. Our objective is to understand how available data sources may be utilized for research on citizens’ acceptance of and preferences for political reform making in Germany, in comparison to other countries and over time, and where their limitations lie. Quantitative social research may be based on a diversity of data sources, such as traditional probability face-to-face survey data, more recent probability and non-probability online survey data and Big Data (by itself or linked to survey data). However, not all data sources are equal in their suitability to research political reform preferences or may differ in their suitability to research certain aspects of this goal. In particular, we aim to leverage some unique properties of the German Internet Panel (GIP) to address four sets of data-quality issues of interest to researchers at large. First, how do issues of representativeness affect the analyses of reform attitudes and preferences conducted with the GIP? Second, how can we best capture respondent attitudes and preferences through the GIP and what are the consequences of respondents using different types of devices (desktop computers, tablets and smartphones) for filling in the surveys? And third, how might GIP panel data be leveraged to better understand the dynamics of how opinion change occurs? The fourth research question constitutes an extension of the project brought about by developments in the research community during the past four years: To what extent can alternative data sources, in particular non-probability online surveys and linked Big Data, add to current insights? In particular, can research based on such alternative data sources be used to infer to the general population in Germany, how does measurement error affect inference and can alternative data sources measure change in opinions over time?