Errors of Experience: Response Errors in Reports about Child Support and Their Implications for Questionnaire Design

Abstract

Reporting errors in surveys are partly shaped by the structure of respondents’ experiences. This happens because respondents often use estimation strategies to supplement memory, and these strategies use information about the frequency and other characteristics of events (e.g., Blair & Burton, 1987; Burton & Blair, 1991; Menon & Sudman, 1989). When events have a simple structure, estimation strategies may give accurate results; but when events are complex, estimation is more difficult and may fail. Consider the example of reporting annual earnings, a task commonly requested of respondents. Some respondents will know their previous years’ earnings. Others, however, will estimate them, using information about the frequency and other characteristics of the event “getting paid.” A person who is continuously employed at a constant monthly salary can estimate earnings very accurately by multiplying the amount of the most recent paycheck by 12. In contrast, estimation is more difficult, and probably less accurate, for someone who is employed erratically. Such a respondent must use a decomposition strategy and may recall different employment spells and perhaps different wage rates to calculate total earnings for the previous 12 months. Thus, these two employment experiences may lead to different reporting errors.