Reducing Underreports of Behaviors in Retrospective Surveys: The Effects of Three Different Strategies

Abstract

Longitudinal social science surveys typically collect data at regular intervals. In most ongoing panel surveys, the time between two consecutive waves of measurement is 1 year. This interval is often chosen because year-to-year changes often suffice to answer the research questions of interest. This article focuses on the accuracy of data that are collected retrospectively on events that occur between two interviews: the use of the medical services of a Family Physician (FP). Surveys like the Survey of Health and Retirement in Europe (Börsch-Supan et al., 2013) and the Health and Retirement Survey (Wallace & Herzog, 1995) annually ask the survey question: “During the last 12 months, about how many times in total have you seen or talked to a medical doctor about your health?” Earlier studies on the quality of such survey reports have found substantial inaccuracies in retrospective reports of behavioral frequencies that were caused by underreporting, overreporting, or a combination of the two.