Autobiographical Memory and the Validity of Retrospective Reports.

• Author: Starr-McCluer, Martha

When economists use data from household surveys, we rarely think about the cognitive process by which respondents answer questions. A growing body of research documents why we should. Much of this research investigates how people interpret questions in survey situations, how they retrieve relevant information from memory, and how information is formatted into an appropriate response. A surprising finding, repeated across studies, is that responses are quite sensitive to seemingly minor changes in the questions asked. For example, if a question about the quality of life is asked after a question about the frequency of TV viewing, respondents rate life less satisfying than when the order is reversed - apparently because invoking negative feelings has a persistent effect on judgment. Even a brief perusal of this literature illuminates whole new ways to distrust data.

This book brings together some fine examples of recent advances in cognitive research on survey methodology. It is a collection of 20 papers from a conference at the University of Illinois at Urbana-Champaign on autobiographical memory, that is, people's recollections of events and behavior from their own lives. One of the editors, Seymour Sudman, has a long and distinguished career in the field of survey research. The other, Norbert Schwarz, is a prolific scholar whose innovative work is revitalizing interest in the field. Other conference participants have varied backgrounds, including statistics, psychology, sociology, education and marketing research.

Part I of the book contains theoretical perspectives on autobiographical memory. This section is perhaps the least satisfying for economists, since there are no rigorous or widely-accepted theories in this area, and the ideas under discussion seem rather speculative. Nonetheless, the articles raise some important issues. For example, frequency questions (for example, "how many times did you use an ATM in the past month") ideally expect respondents to recall and count all instances of the target behavior during the reference period. Experimental studies suggest that other strategies are often used to formulate answers, especially when the behavior in question is too mundane for individual episodes to be stored in memory. The papers in this section discuss implications of such issues for understanding how memory works, though without major new insights.

The articles in Part II are interesting case studies of retrospective...