Appendix B: Detailed Methology AARP Multicultural Survey
Based on the codes in the accompanying table, RR5 = I/((I + P) + (R+NC+O)). Using the same codes, REF3 = R/((I + P) + (R+NC+))). The calculated outcome distribution rates that result from these formulae for each of the samples are:
Again, because the sample frames are different for each sample (the national sample is an RDD sample of all adult households regardless of race or ethnicity, the African-American and Hispanic over samples are targeted RDD samples, and the Asian-American over sample is a list sample) the response and refusal rates cannot be compared from sample to sample. Also, while each of the three over samples targeted a particular ethnic group, the actual incidence of the sampled telephone number producing a member of that ethnic group was far short of 100 percent. In fact, some of the numbers for one targeted group produced respondents of the other targeted ethnic groups (e.g., telephone numbers in the Asian-American over sample produced many respondents who were Hispanic or African- American). Therefore, no absolute conclusions can be drawn about the response rates or refusal rates for any particular group.
It is possible that the response rates are artificially low. The formula used to calculate them includes large numbers of initial refusals and contacts where communications were impossible due to language barriers as eligible households when, in fact, it was impossible to determine whether or not the households contained potential respondents aged 45 to 55. Where the age of household members was determined, over 30 to over 50 percent were found to be ineligible for reasons of age or race. If this ineligible rate is applied to reduce the hard refusal rate and the rate where language barriers inhibited determining eligibility, the pool of eligible households would have been reduced significantly and the response rate would increase for each sample frame.