Mediation and moderation in social work research

Abstract

Aruna Gogineni , ACSW, is a PhD candidate, Ruth Alsup , MSW, is a PhD candidate, and David F. Gillespie , PhD, is professor, George Warren Brown School of Social Work , Washington University, One Brookings Drive, St. Louis, MO 63130. An earlier version of this article was presented at the Annual Program Meeting of the Council on Social Work Education , March 1993, New York. Social ciplines. cial drawn by methodological work's work from Much is research sociology a multidisciplinary of the developments methodology and theory psychology, in field social in incorporates influenced other work and dissois by methodological developments in other disciplines. Much of the theory in social work is drawn from sociology and psychology, and socia work' researc e hodol gy corp r te and adapts techniques from each discipline. However, statistical applications have developed differendy in psychology and sociology, creating disjunctives and inconsistencies in applications to social work research. Different definitions for the same terms and different terms used to convey the same concepts create confusion in the interpretation of empirical results and the comparison of results across studies. Useful relationships between variables in the social sciences rarely involve only two variables. A third the test variable, is introduced to elaborate the meaning of the original two-variable relationship. The process of elaboration developed by Kendall and Lazarsfeld (1950) and Lazarsfeld (1955) and discussed in detail by Rosenberg (1968) is the classical source of key terms in sociology for describing multivariate relationships. Psychologists developed different terms for these processes, in part because of the greater use of experimental design and focus on microtheory (MacCorquodale & Meehl, 1948; Royce, 1963; Rozeboom, 1956). Although methodological approaches have been converging across the social sciences since the mid-1960s, there still exists considerable confusion regarding the interpretation of multivariate results. One author may define a variable as intervening, another may call it mediating, and still another may call it moderating. As a consequence of this confusion, although the terms mediation and moderation refer to distinct processes, many examples of their inappropriate use are found in the literature. Baron and Kenny (1986) described studies where a moderator interpretation is given to variables mediate the relation between locus of control and academic achievement. Sometimes the terms are used interchangeably. For example, Harkins first observed identifiability moderates social loafing, then proposed that identifiability is