The Use of Interviewer Estimations At The Door-Step Interaction: Examinations From The European Social Survey, 2018

Abstract

Social research data is enriched day by day with new studies and new techniques. It is important to develop new methods in order to maximize the efficiency of research, improve data quality, reduce non-response and lay a strong foundation for future studies. Collecting and analyzing interviewer observations and predictions is one of the popular research methods applied in recent studies. The aim of this thesis is to analyze the interviewer estimation data, called paradata, in the light of the data set of European Social Survey (ESS) 2018 and to observe the consistency of the estimations. In the ESS 2018, interviewers were asked to predict how the second visits would turn out for interviews that were not completed after the first visit. Starting from the first interview, interviewers were asked to make predictions about the outcome of the next interview after each session. Interviewers made predictions based on their own impressions and these data were recorded. In this way, the data set of the study includes not only the actual results of the interviews but also the estimated results that the interviewers marked as a result of their predictions. In this thesis, whether the interviewer's estimations coincide with the actual results is analyzed in terms of the age and gender of the interviewers and the type and physical condition of the dwelling where the interview took place. In this study, descriptive analyses yielded more explanatory results than multivariate analyses, particularly due to data deficiencies that rendered the logistic regression outcomes less significant. While the likelihood of consistency in participants’ predictions appeared to be influenced by the interviewer’s age, gender did not produce a significant difference. Similarly, the type of housing and the country in which the interview took place were found to be significant factors, whereas the neighborhood conditions of the residence did not have a meaningful impact on the prediction data.