Appendix B: Detailed Methology AARP Multicultural Survey
race/ethnicity on all the analyses. For example, the tables for the whole sample reveal that there is a significant difference between non-Hispanic whites and minorities on the question of caring for parents, in-laws, or other older people. Fewer non-Hispanic whites claim to give such care.
One might wonder whether the difference between whites and African-Americans on this question is driven by economics or not. By looking at the bivariate tables run for whites only and African- Americans only, one can quickly determine this. In this case, the proportion claiming to give care increases at the highest income levels for both groups. However, low- income African-Americans are still more likely than low-income whites to care for the elderly and upper-income African Americans are even more likely than upper-income whites to provide such care. This suggests that the income of the respondent is not a factor of the difference between whites and blacks on this question. The income of the care recipient may have an influence, something that cannot be determined from these data, as more African-American elderly need assistance from the younger generation regardless of their income status than older whites, the difference may be cultural, or it may be a little of both.
This methodological example serves to indicate that bivariate tables have been generated for the data that permit analyses that often require multivariate analyses and should be referred to for this purpose when the concern is about differences or similarities among the racial/ethnic groups under study.
Factor analysis is an appropriate technique to examine large numbers of variables by grouping them according to how they correlate with each other. Variables in one group or factor are more correlated with each other than they are with variables in other factors. This summarizes many variables into a few factors, essentially reducing the number of items necessary to analyze. It also creates the opportunity to interpret each factor according to the meaning of the variables making up the factor. In essence, it permits underlying attitudinal dimensions to be identified to help unravel and understand the structure of the variables measured in the survey.
For this study, factor analyses were performed for several sections of the survey:
Q9-Q19, dealing with major events that people may experience during their lifetimes;
Q22-Q32, concerning agreement with a number of statements about family relationships;
Q50-Q61, about the frequency of various types of care and assistance that can be provided the elderly;
Q62-Q68, measuring how caring for the elderly affects one’s other decisions in life;
Q69-Q73, indicating what things one may have done to make caring for the elderly easier;
Q74-Q77, how helping older relatives impacts one’s relationship with them and one’s own sense of aging;
Q78-Q83, the amount of stress caring for older relatives causes on one’s relationship with family and friends; and