As directed by the Department of Urban Development, the Statistical Research Division (SRD) conducted cognitive testing of proposed revisions to the housing subsidy question series in the American Housing Survey (AHS). Revisions had been proposed by ORC Macro (ORCM) based primarily on ethnographic testing of key concepts in the series. That research had identified potential sources of errors in the self-reporting of housing subsidy status and type. Based on their findings, ORCM suggested new words, concepts, and questions to improve the series. The research reported here describes SRD's test of whether the proposed questions 1)reduced problems identified by ORCM and 2) met the original measurement objectives. To accomplish this, SRD administered the actual housing subsidy question sequence in whole, also including the AHS rent control question. By testing the entire series, it was also possible to observe the effect of question context (i.e., neighboring questions) on the interpretation of ambiguous or technical terms associated with housing subsidy concepts. Results showed that with a few exceptions, the revised housing subsidy series outperformed the current AHS questions. The new questions tapped different aspects of housing support and, in theory, should help analysts better classify respondents among program types. The failure of some questions (largely explained by vague or unknown terms) however, indicates that self-identification of subsidy type will remain difficult for participants in some particular programs. Where cognitive difficulties persist, it is largely because persons in traditional Section 8 housing programs use imprecise language to identify their program type. Multiple terms (certificate, voucher, subsidy, Section 8 etc.) compete for the same concept. Also key concepts are understood and used differently between types of program participants, between regions, and among people with varying tenure in the programs. This report suggests that the final specifications for the subsidy questions will benefit from the use of clearer terms, explicitly expressed definitions where ambiguous terms are unavoidable, and modified question skip logic to avoid erroneous inclusions and misreporting for key items.
