ESRA (European Survey Research Association) 2023 Conference, Milan, İtalya, 17 - 21 Temmuz 2023
The willingness to survey participation and statement of accurate answers are as important as high interviewer performance in surveys. The impact of respondents in surveys is usually explored for questions designed to measure attitudes, values, and beliefs rather than demographic characteristics, daily practices, and behaviors. This study aims to investigate the effect of respondent motivation on item nonresponse for a set of questions designed with the split-ballot technique. To reach this aim, five different countries (France, United Kingdom, Norway, Netherlands, Portugal) in the European Social Survey (Round 9) were selected. Within the survey, different versions of questions designed to reveal gender differences were asked to randomly selected sub-groups. In this study, “don’t know” and “no answer” response options were accepted within the item nonresponse. On the other side, respondent motivation is evaluated through interviewer evaluations per respondent. Complex sample design adjusted estimates were produced within the multivariate part of the study. Descriptive findings showed that there’s a significant moderate relationship between the motivation and item nonresponse in a negative way (Pearson’s coefficient: -0.24, p<0.01). Multivariate analyses controlling for interviewer and respondent characteristics found that a point increase in respondent motivation leads to a reduced level of item nonresponse (about 3 items in France, almost 2 items in the United Kingdom, Norway, Netherlands, and Portugal). Findings refer to the need for keeping respondent motivation at a high level during the interview, and the value of interviewer assessments when explaining item nonresponse. Moreover, it seems that questions designed with the split-ballot technique are prone to item nonresponse and thus, increasing alerts for such questions would be useful. When the switching mode propensity of the European Social Survey in recent times is considered, these alerts should be designed according to data collection methods.