ABSTRACT
Measurement error has potentially serious consequences in the study of social behavior using sample surveys, regardless of the characteristics of the subject population. Because of the cognitive demands of some survey questions (see Sudman, Bradburn, & Schwarz, 1996), and given well-documented laboratory and survey results regarding declines associated with aging in some domains of cognitive functioning (see Herzog & Rodgers, 1989; Park et al., 1996; Salthouse, 1991, 1996; Schaie, 1990, 1993, 1996), it has been suggested that there may be identifiable differences in errors of measurement associated with the age of the respondent in actual surveys. Given the impact of levels of measurement error on the robustness of statistical analysis in social and behavioral research (Bohrnstedt, 1983; Bohrnstedt and Carter, 1971; Cleary, Linn, & Walster, 1970), if there are differences in measurement error across age groupings, it is important to take these differences into account in the analysis of data.
