Experimental Design for Mediation Tests

Abstract

Testing mediation for organizational studies is challenging as that requires establishment of two or more causal paths as the causal chain underlying mediation. Many existing studies testing mediation utilize a single experiment that manipulates the independent variable and measures the mediator and the dependent variable, likely because it is unethical or unfeasible to manipulate the mediator. Such studies often receive critics that the causal effect of the mediator on the dependent variable is not established and thus the causal chain not complete. In this paper, we theorize and identify three remedies a single experiment can use to examine the causal chain underlying mediation. When the three remedies are not applicable, we classify parsimonious experimental designs that make experiments testing mediation more cost-efficient and feasible.