Misleading statistical studies

Abstract

Large statistical studies in the social sciences, including one-off or repeated cross-sectional surveys, time-series surveys and cohort longitudinal research, offer important numeric evidence for policy making. Although single studies rarely occasion dramatic policy shifts, statistical research findings can affect policy debate, even if not always directly or openly. At best, these studies reveal shapes and patterns in the social fabric relevant to health, safety, education and other social goals. Numerical measures of many social phenomena, such as unreported crime, illicit drug use, child-rearing practices or family composition, enter into a policy-making milieu crowded with competing numbers and qualitative information, as well as non-evidential values and power-based influences.