Research Methods Brief: Attrition Happens (and What to Do About It)

Abstract

Attrition (participant "dropout") is the loss of participants from a program/initiative or longitudinal (e.g., pre/post) data collection.  If participants dropout for non-random, systematic reasons, those factors bias the sample and limit the study or evaluation’s generalizability.  The importance of statistically diagnosing participant attrition can scarcely be overstated, given that P/CVE research and evaluations are commonly concerned, not merely with the results from a given sample of participants, but whether, how, or to what extent the results might generalize to other, perhaps much broader samples.  Therefore, the threat to generalizability, posed by non-random participant attrition, threatens the very reason for conducting many, if not most, P/CVE-related research and evaluations.

Non-random attrition prevents research and evaluations from making valid claims or inferences about their target populations, and to know whether attrition likely threatens the validity of a project’s findings, one must test for it.  The present article includes step-by-step guidance on how to diagnose participant attrition, including discussion of the implications: implications that potentially can salvage a P/CVE-related program from seemingly problematic participant attrition.