Reactivity in Panel Studies and its Consequences for Testing Causal Hypotheses

Abstract

The procedure of standardized repeated measurement, as used in panel studies, may hamper the quality of the data, due to the potential “reactivity” of survey interviewing on respondents' attitudes and behavior. In case respondents are interviewed in subsequent waves by different interviewers, differential interviewer effects may occur. These threats to data quality are illustrated with data from a longitudinal study among 2,819 older adults, conducted in the Netherlands. From an analysis of 100 interview protocols, it appears that the behavior of the interviewers has a significant impact on the data obtained. On one hand, interviewers seem to adjust their interviewing strategy to a norm regarding a “normal” personal network and, on the other hand, to a norm about the appropriate interviewing time. Suggestions are formulated to prevent misestimating actual change within respondents over time, leading to incorrect conclusions about causal relationships.