Understanding ‘Don't know’ answers to survey questions - an International comparative analysis using interview paradata

Abstract

If ‘Don’t know’ answers to survey questions are not being accurately captured, understood and analysed then the validity of the estimates of public knowledge, values and attitudes can be questioned. In this article survey answers and interviewer observation paradata about how respondents answer factual, value and attitude-based survey questions were analysed. There was considerable variation in the likelihood of a ‘Don't know’ response at both the interviewer and country levels. Simply discouraging ‘Don’t know’ answers or coding them all as non-directional responses or missing data is likely to lead to measurement error. Methodological innovations should include follow up questions in surveys to explore the reasons for a ‘Don’t know’ response. Additional support for those respondents who find answering certain questions a challenge could also be introduced. It is clear a more robust and in-depth understanding of ‘Don’t know’ responses is important for the accuracy of social surveys.