Appendix B: Detailed Methology AARP Multicultural Survey
total of 429), and 304 Asian Americans (for a total of 351). The margin of sampling error for a sample of 1,500 is ± 2.5 percentage points at the 95% level of confidence. The margin of sample error for the race and ethnicity sub-samples of 404 and 429 is ± 4.9 percentage points. The margin of sample error for the sub-sample of 351 is ± 5.2 percentage points. A more detailed explanation of the sample frames is provided below.
The survey was conducted from March 8 to April 4, 2001. Each interview lasted approximately 20 minutes and contained 97 questions.
The national probability sample was stratified by the four major Census regions: Northeast; South; Midwest; and West. Within each region, a set of telephone numbers was generated proportionate to population with a total pool of about 10,000 numbers after cleaning for disconnected, government, business, fax and computer tones, and non-working numbers. A combination of random and fixed-digit sampling procedures were used to introduce both listed and unlisted numbers as well as new households into the available sample. The response and refusal rates produced by utilizing these procedures are provided elsewhere in this appendix.
The telephone numbers were organized by region into three or four “replicates” of about 1,000 numbers each, depending on the region. Each number in the first replicate was dialed three times before interviewers moved to a second replicate. But, redialing the first replicate numbers was continued up to as often as ten times in an attempt to obtain an interview before being replaced totally by a number from the second replicate.
When contact was established, the interviewer screened for an appropriate respondent based on age. The person in the household between 45 and 55 with the most recent birthday was selected. The application of these procedures produced a calculable probability of being included in the survey sample for each member of the target population.
Three sampling techniques were used for the four samples. The national sample design was a simple random-digit dial (RDD) telephone number sample frame. It was stratified by geographic region according to each state’s proportional share of 45-55 year olds and gender. It produced a representative probability sample of the adult American population age 45 to 55.
The research design called for large enough samples of African- Americans and Hispanics to make accurate projections for these groups as a whole. Budgetary and time constraints required a more efficient method of contacting members of these groups than through a RDD sample frame where the incidence of each is relatively small. So, a second sample design was developed to produce targeted RDD telephone sample frames for each group.
Telephone exchange areas were overlaid with Census data on race and ethnicity. Those exchanges from areas where the density of the African-American or Hispanic population was at least 50 percent were selected. This increased significantly the likelihood of reaching an African-American or Hispanic household. Random dialing telephone procedures were employed