The Quality of Survey Data as Related to Age of Respondent

This research examines how the quality of survey measures varies with the age of the respondent. A structural modeling approach is applied to data from six surveys (five national surveys of American adults and one survey in a Canadian corporation—7,706 respondents in all) to generate measurement quality estimates for 106 survey measures. Results show that as respondent age increases (a) the percentage of true score variance in survey measures tends to decline, (b) the percentage of both random and correlated error variance tends to increase, and (c) people tend to have more interrelated—or less differentiated—views about their worlds. Most of these trends are highly replicable from survey to survey and are—except for the level of differentiation in views—not attributable to older respondents tending to have less education than younger people. The results suggest that data from older respondents tend to provide a somewhat less precise indication of the attitudes, behaviors, or other characteristics being measured than do data from younger respondents, but that relationships among survey measures will not necessarily be weaker.