The Longitudinal Item Count Technique: a new technique for asking sensitive questions in surveys

Abstract

Asking respondents sensitive questions directly may lead to socially desirable responding. As alternative, some have proposed using the Item Count Technique (ICT). The problem with ICT methods is that these can have low statistical efficiency, but also do not provide an indicator of the behavior at the respondent level. We propose a new variant of the ICT to overcome these issues: the Longitudinal Item Count Technique (LICT). Instead of administering different lists (one including the sensitive item and one without) to two random groups in a single survey, the LICT administers both lists to each respondent, but at different survey waves. The sensitive attribute can be estimated as the difference within individuals across waves. Like the ICT, the LICT can be extended to a two-list version. In this paper we discuss the assumptions, implementation, limitations, and ethical implications of this novel technique, and present application of the method in the Understanding Society Innovation Panel, estimating the prevalence of the gay, lesbian, and bisexual population in the United Kingdom. In this first application, the LICT in some ways appeared to provide better estimates than the traditional ICT, but also provided some inconsistency in estimates. We discuss the implications of these results and point to routes for further research.