The Marginal Decomposition Approach Quantifying Direct and Indirect Effects in Causal Models
Year
2021
Author(s)
Sang June Park, Youjae Yi
Journal
Seoul Journal of Business
Volume
27:1
Pages
1-22
Many researchers have analyzed causal mediation with the measures for direct and indirect effects in a system of regression models. The direct effect indicates the effect of a focal predictor not through a mediator, whereas the indirect effect indicates the effect of the focal predictor through a mediator. Various versions of two approaches (product approach and potential outcomes approach) have been used to find the measures indicating the quantified direct and indirect effects in a system of regression models. However, it may not be easy to identify the measures with the two approaches, because they do not provide a general formula for identifying the measures in various systems of regression models. Thus, this paper proposes a new approach providing a general formula for identifying the measures intuitively and clearly. The new approach decomposes the effect of a focal variable on a dependent variable into five additive components in view of moderation and mediation.