Appendix A: Detailed Methodology
Telephone interviews were conducted from October 13 to 26, 2010, with 1,013 people aged 50 and older over two waves of the twice-weekly EXCEL omnibus survey. Each wave used a fully replicated, stratified, single-stage, random-digit-dial sample of telephone households. Within each sample household, one adult respondent was randomly selected using a computerized procedure based on the last-birthday method of respondent selection. If the household respondent was aged 50 years or older, he or she was asked to respond to the survey questions. No other criteria were used to include or exclude respondents. The survey response rate was 11 percent. At least three attempts were made to a telephone number on various days, including weekends, and at different time periods. The number of attempts is limited because each wave is in the field for only 5 days. The attained response rate is typical for quick-turnaround surveys conducted by random-digit-dial telephone interviewing.
Data were weighted to account for disproportionate probabilities of household selection due to the number of separate telephone lines and the probability associated with random selection of an individual household member. Data were then stratified and balanced based on the national distribution of people 50 years and older by age, sex, race and ethnicity, educational attainment, and U.S. Census region to ensure reliable and accurate representation of persons in the target population. Before weighting, the sample was somewhat older, more female, lower income, better educated, and white than the U.S. population aged 50 and older.
The methodology for the 2010 survey largely replicates the methodology for the survey conducted in 2006 with a total of 1,559 respondents aged 50 and older. The 2006 response rate was 30 percent and the margin of error was plus or minus 2.48 percentage points at the 95 percent confidence level. Before weighting, the 2006 sample was somewhat older, better educated, somewhat wealthier, and more non-Hispanic white than the U.S. population aged 50 and older.
Limitations of Method
As with all telephone surveys, people who live in nonresidential settings or in households without landline telephones were excluded, as were individuals who do not speak English well enough to participate. The extent to which the views of these and other nonrespondents (people in households that were called who were not reached or refused to participate) differ from those of respondents is unknown. In addition to sampling error, question wording and practical difficulties in administering surveys may have introduced error or bias into the findings of this, or any, poll of public opinion.
Differences Between 2010 and 2006 Questionnaires
The 2010 and 2006 surveys were carried out using the same methodology and covered the same topics and in many instances contained questions that are materially similar. However, substantial differences in the wording of some questions and response categories limit the extent to which data from the two time periods can be compared. Appendix B contains the 2010 survey instrument. Where possible, statistical testing was performed to identify significant differences in responses between the 2 years; z-tests were used to compare column proportions and identify differences significant at p < .05 (using a two-sided test of equality and adjusted for pairwise comparisons using the Bonferroni correction).