Determinants of item nonresponse in a large national sex survey

The study aim was to describe patterns of item nonresponse in a sex survey and identify factors associated with high nonresponse. A random sample of 4548 respondents to the National Survey of Sexual Attitudes and Lifestyles in Britain, 1990-1991 was assessed. Both respondent-wise and variable-wise patterns of incomplete response are described in terms of distinctive characteristics of nonrespondents and questions with a high nonresponse rate. Logistic regression was used to model the determinants of extreme nonresponse. For at least 90% of respondents, the item nonresponse was ignorable for most practical purposes and recall difficulties were indicated as its main cause, while the sensitivity of the questions had a large effect mainly among those in the top 5% of the overall nonresponse distribution. The latter were also distinguishable in terms of refusals to answer some face-to-face questions. Several overlapping indicators of poor educational and social background were associated with elevated overall nonresponse, particularly among the top 5% of the distribution. Thus recall accuracy rather than sensitivity of some sex survey questions was the key difficulty for a majority of the respondents whose overall nonresponse was satisfactorily low. The sensitivity of the topic had a large effect on item nonresponse only for a small group of participants.