C’ Test Helps or Hinders? -- Clarifying the Communication Fallacy of “Effect To Be Mediated” In Marketing Research

Baron and Kenny’s (1986) classic procedure for testing mediation required that the simple correlation between independent variable X and dependent Y must be significant, which we call c’ test. Many authors, including Kenny, later recommended suspending c’ test under some conditions. A couple recent articles recommended to completely repeal the test. All authors focused on one or two types of mediations at a time. This article takes a more comprehensive view to clarify this communication fallacy for marketing researchers. Following the mediation type framework recently developed by Zhao, Lynch and Chen (2010), we laid out a roadmap of all possible scenarios of three-variable non-recursive models. We grouped the 51 resulted scenarios into three types of mediations and two types of non-mediations. We then examined each type to see if c’ test helps or hinders in establishing mediations. We found that c’ test hinders for establishing two types of mediations, and never helps in any type. We concluded by supporting the emerging view that c’ test should be completely repealed for establishing any type of mediations.