The Implications of Respondent Loss in Panel Studies of Deviant Behavior

Abstract

While the panel design offers many advantages to social scientists, especially those studying deviant behavior, a major source of potential bias and error of this design is the loss of respondents over time. The present investigation draws upon a secondary analysis of data from seven panel studies which followed adolescent populations through time to examine the impact of such attrition on univariate, bivariate, and multivariate estimates obtained from individuals who remain in panels. Those who remain over time in panels are found to provide relatively accurate estimates of bivariate and multivariate relationships. Univariate estimates such as incidence or prevalence measures of deviance are found to be slightly but systematically biased, this tendency being more clear when the panel is constructed from a broadly representative population base.