Selection and use of mediation testing methods; application in management sciences

Abstract

This research article presents a comparative analysis of three major methods of mediation analysis i.e., Baron and Kenny, Sobel, Hayes indirect effect with bootstrap using PROCESS macro and Structural Equation Modeling (SEM). It discusses common issues associated with mediation analysis and common mistakes committed in the selection and application of method and in the interpretation of results. In this article we have developed and tested a variety of models including simple mediation, parallel mediation, and serial/sequential mediation models. We have used a research model drawn from management field which includes two independent variables (i.e., organization justice and corporate social responsibility), two dependent variables (i.e., employee wellbeing and organization performance) and two mediating variables (i.e., organization trust and organization culture). The article illustrates that bootstrapping has an advantage over Baron and Kenny method or with Sobel test. Besides, it has high statistical power and better control on type-I error. It produces better results even when data lacks the property of normal distribution. The article has many practical implications. More particularly for management researchers, it provides an in-depth understanding of how to correctly conduct mediation analysis.