Project 1

The primary focus of this investigation is the measurement of persons with disabilities. The project seeks to assess the effect of alternative questionnaire wording on reliability (simple response variance) and the effects of self and proxy reporting on the reliability of responses.

The six disability questions included in the long form of the 2000 decennial census are the basis for the experimental design. Alternative versions of the six questions have been designed to investigate the impact of alterations on response reliability. Alterations include:

· separation of the means by which an activity is performed from the difficulty associated with performance of the activity.

The study will consist of two successive interviews in 800 households. In order to increase the incidence of identifying persons with disabilities, only those households in which at least two people ages 40 and older reside will be eligible for the interview. Each household will be randomized to a question treatment for a sensory impairment question and separately randomized to a version of the activity limitations questions. In the first interview, a randomly chosen respondent will be asked to respond to the questions concerning him- or herself and then to report about one additional person in the household age 40 and older (‘other’). Approximately two weeks later, the household will be recontacted. In half of the households we will interview the same person who responded in the initial interview; in the remaining cases we will interview the ‘other’ person for whom data was collected in the first interview.

One major concern in designing questions about disabilities is that we have little understanding of respondents’ perception of what constitutes a disability and how that does or does not correspond to theoretical models used by researchers. In an attempt to understand respondents’ perceptions of disabilities, each respondent will be read a vignette describing a person and asked as to whether he or she perceives the person described in the vignette to have a disability. The vignettes vary on a number of characteristics including the nature of the disability (e.g. physical, emotional, addiction), the person’s gender, whether or not the person is currently employed, the use of devices or medication, and other individual awareness of the persons’ condition or impairment.

The design permits the assessment of simple response variance across various self-proxy treatments (self-self response; proxy-proxy response; and self-proxy response). In both waves of the study, respondents will be randomized to a vignette (different vignettes in each wave) and asked to respond to the perception questions with respect to the vignette description.

Tables 1 and 2 present the factorial design for the experiment. Due to the nature of the question concerning sensory impairments, we have treated that question differently than the remaining five questions which concern functional limitations and participation. Table 1 presents the design related to the single question in the decennial Census concerning vision or hearing impairments; Table 2 presents design features for questions on physical limitations, learning, remembering or concentrating limitations, activities of daily living, going outside the home alone, or working at a job or business.

Table 1. Experimental versions for vision and hearing (sensory impairment) question

Factor	Version
Factor	E	F	G	H	I
Response scale	Y/N	5 pt.	Y/N	5 pt.	5 pt.
Shorter, simpler q	No	No	Yes	Yes	Yes
Task performance	None	None	None	None	Yes

The design of the Wave I and Wave II questionnaires was completed in early July. The Gallup Organization collected data via telephone and computer assisted interviewing. A small pretest was followed by an interviewer debriefing, interviewer training for the main study, and data collection started in August, 2000. Data collection was completed in the fall of 2000 and included initial interviews and reinterviews with respondents in 802 households. To increase the incidence of identifying persons with disabilities, only those households with at least two people ages 40 and older were eligible for the interview. Each household was randomly assigned to get a version of the sensory impairment questions and was separately randomized to a version of the activity limitations questions. In the first interview, the respondent answered the questions about themselves and another member of the household who was 40 or older. Approximately two weeks following the initial interview, the household was recontacted. In half of the cases, the respondent to the first interview was interviewed again. In the other half, the respondent for the second interview was the the member of the household for who data were collected in the first interview., Thus, we can assess the reliability of the disability data when they came from a single respondent in both interviews or two different respondents.