Reliability and social work research

The conventional practice in social work research is to recommend a level of reliability close to .80 as a minimum standard. This article contends that needed reliability varies by situation. In some situations, particularly ones in which important decisions about individuals are being made, .90 provides a better standard. In other situations, for example, in descriptive survey research with large samples, levels of reliability much lower than .80 suffice. In experimental designs and single-subject designs, needed reliability varies according to many factors, although the .80 standard does hold some utility. Increased availability of brief measures would greatly benefit all parties— practitioners, researchers, agencies, and clients— involved in social work research.