Previous research has demonstrated the considerable benefits of a "topic-based" interview structure for demographic surveys employing a household respondent (Moore and Moyer, 1998a, 1998b). As distinct from traditional person-based interviewing, which completes all questions about person 1, then recycles through the questions for person 2, and etc. for person 3 and all other household members, the topic-based approach is organized by topic: "Has John ever been divorced?" "How about Mary?" "And Tom...?" etc. Moore and Moyer (op cit) report that this structure increases interview efficiency, is preferred by both interviewers and respondents, reduces unit nonresponse, and in general reduces item nonresponse – with income items the notable exception. Loomis (1999) identifies the elevated tendency of topic-based interview respondents to produce blanket refusals for all income questions for all household members as the mechanism for the increase in income item nonresponse. This paper describes an experimental attempt to reduce item nonresponse to income questions in a topic-based interview, using a very brief statement immediately preceding the income questions to try to overcome respondents' reluctance to report by (a) acknowledging that such reluctance is not uncommon; (b) suggesting a cause of the reluctance (unfamiliarity) that avoids suggesting privacy concerns to those not troubled by them; and (c) emphasizing the valid, important, non-personal, statistical (i.e., nonthreatening) uses of the data. We find that use of this statement eliminates the income item nonresponse disadvantage for the topic-based interview treatment – and in some cases even reverses the effect. We also find positive opinions among interviewers about the topic-based design in general, and the introductory income statement in particular.
