It is well-known that sample members who find the survey topic more important are more likely to respond. What is less known is the effect of different topic compositions. To reduce respondent burden, some social science surveys split long multi-topic questionnaires into varying components and administer only subsets to sample members. The rest of the questionnaire is asked in a second (follow-up) part. Whether respondents who answered more questions of importance to them in the first part are more likely to respond to the second part has never been tested.
In this paper, we examine this question using data from a multi-topic push to web survey with six questionnaire subsets, which are assigned at random to sample members in the first part. At the beginning of the first part respondents are asked the importance the different topics have in their lives and at the end whether the survey was interesting to complete. This design allows assessing effects of topic distribution and topic importance on interest in the survey and on response to the follow-up.
We first found that the subsets of the questionnaire are jointly associated with how interesting the respondent found the survey. Yet, we do not find support for our idea that respondents who answered more questions on topics important for them found the survey more interesting. Neither were these respondents more likely to participate in the second part of the survey. This shows that respondents to the first part do not self-select into the follow-up depending on topic importance.