Differences between real and falsified data

Abstract

In social science research, face-to-face interviews are a widely used mode of data collection. Interviewers generally have a positive influence on data quality. But there is also the risk that they depart from their interviewer guidelines and thereby negatively affect the data quality. Interviewers even may decide to falsify interviews. Therefore, it is important to know which effects falsified data have on the results of social science research and thus on data quality. The research question this work aims to answer is how real and falsified survey data differ with respect to the results of theory-driven analyses: Which influence do falsifications have on findings of substantial social science research?For the analyses, data of the research project IFiS Identification of Falsifications in Surveys are used. The database consists of three datasets. First, 78 interviewers conducted 710 real face-to-face interviews. Second, the same interviewers falsified survey data in the lab; to each real interview a corresponding falsified interview was collected. For that purpose, the falsifying interviewers received a short sociodemographic description of the real survey respondents. Third, the interviewers filled in the survey questionnaire for themselves. Theory-driven models of statistical correlations were tested with real and falsified data separately to examine how the results differ with respect to the consistency of the models as well as the effect sizes. Furthermore, interviewer effects on real and on falsified data were analyzed.