Dykema, Jennifer, and Nora Cate Schaeffer
Working paper no. 1999-27
Abstract
A substantial proportion of the research conducted by social scientists involves measuring, describing, and analyzing phenomena that can be broadly classified as events or behaviors. In order to measure experience systematically and in a manner that permits generalization across individuals and to larger groups, retrospective self-reports about events are often obtained within the context of a survey interview. Responses to survey questions are treated as reflecting some truth about the individual’s past. A critical problem with this formulation, however, is that not only do we usually lack methods of measuring the accuracy of survey responses, but we do not have a generalized framework for understanding or systematically analyzing which characteristics of experience, if any, influence errors in reports about those events. Furthermore, we do not fully understand how to design survey instruments to aid respondents in reporting more accurately.
The analysis presented in this paper combines findings from the literature on response effects in survey research with principles derived from the social cognition paradigm within social psychology to both expose and reduce reporting errors for a real-world phenomenon, child support. The central research question addressed is: To what extent do the structural features or characteristics of receiving and making child support payments predict errors in reports about those events in a validation study? In answering this question the concept of a personal event schema is developed. Results show that dimensions of the schema, such as the complexity, clarity, and affective intensity associated with the individual events predict reporting errors. This paper also explores how the analysis of these errors can inform questionnaire design in order to reduce reporting errors.
The data are from self-reports obtained during two different telephone interviews conducted in 1987 (Parent Survey 1) and 1989 (Parent Survey 2) with parallel samples of divorced Wisconsin parents. The analysis compares the court’s records about child support with answers obtained in the telephone surveys. While child support is used as an example in this study, the intention is to produce a methodology that is generalizable to similar phenomena such as reports of income, AFDC receipt, sexual behaviors, and physicians’ visits.