Panel Conditioning in Longitudinal Social Science Surveys

Abstract

Social scientists usually assume that the attitudes, behaviors, and statuses of respondents to longitudinal surveys are not altered by the act of measuring them. If this assumption is false—or even if the quality of survey participants’ responses change because of measurement—then social scientists risk mischaracterizing the existence, magnitude, and correlates of changes across survey waves in respondents’ characteristics. In this article, we make the case that social scientists ought to worry more about panel conditioning biases. We also describe and demonstrate empirical strategies for estimating the magnitude of such biases in longitudinal surveys, and we provide illustrative empirical results that are germane to social science research. We end by outlining a research agenda that would generate specific information about the nature and degree of panel conditioning in specific longitudinal surveys as well as a broader understanding of the circumstances in which panel conditioning is most likely to occur.