Appendix B: Detailed Methology AARP Multicultural Survey
with these sample frames, and respondent age and race were screened at the beginning of the interview.
Because of the extremely low incidence of Asian Americans in the population (less than 4 percent), another sample frame had to be developed to reach these respondents in a cost-effective and timely manner. A national list of telephone numbers with Asian (Japanese, Korean, Chinese, Filipino, Vietnamese, and Indian) names was generated with the names that were also likely to come from other ethnic groups (e.g., Lee, Park) excluded. Random dialing telephone procedures were employed with the numbers on the list, and respondent age and race were screened at the beginning of the interview.
Respondents who met the age and race screen requirements were interviewed regardless of the sample frame in which their telephone number occurred. For example, if an African American between 45 and 55 was contacted through the Asian-American or Hispanic sample frames, he or she was interviewed and the interview was applied to the African-American over sample. This enhanced the random nature of the over samples and permitted contacts with individuals who lived in areas with lower densities of the targeted population.
The racial/ethnic over samples created disproportionate numbers of these groups in the whole sample. It became necessary to weight by race and ethnicity to align the sample on a more proportionate basis. The over samples also produced some misalignment in the gender of each of the groups. While the overall U.S. adult population is about 52 percent female and 48 percent male, among 45-55 year olds these proportions vary within the racial/ethnic samples. Males compose nearly 50 percent of the white population, 46 percent of the African-American population, 51 percent of the Hispanic population, and 47 percent of the Asian-American population. Therefore, it became necessary to weight the sample for gender as well.
The combined weight factors, then, are based on race/ethnicity and gender. Using July 1998 Census data for these factors among 45-55 year olds, weights were created for the study so that it would represent the U.S. population accurately. The proportions for determining the weights are presented in the following table. It accurately reflects the U.S. 45-55 year old population for race/ethnicity, gender, and geographic location.