Meditating and moderating variables are essential for building accurate, theory-driven research models. According to Baron and Kenny (1986), mediators explain how or why an independent variable (IV) influences a dependent variable (DV), whereas moderators explain when or for whom that relationship occurs. MacKinnon (2008) expanded this framework by clarifying how indirect effects can be tested and emphasizing the importance of accurately estimating the indirect pathway. 

A mediator represents the process that transmits an effect. It lies between the IV and DV (IV – Mediator – DV). Baron and Kenny (1986) describe three principles for mediation: (1) the IV predicts the mediator, (2) the mediator predicts the DV, and (3) the IV – DV relationship weakens when the mediator is included. If the relationship becomes nonsignificant, full mediation exists if it weakens but remains significant partial mediation is present. As the Mediation and Modern Guide explains, this is typically tested with hierarchical regression or path analysis. 

A moderator changes the strength or direction of an IV – DV relationship. It interacts with the IV (IV x Moderator – DV) to determine when or for whom an effect is strongest. Moderators can be categorical or continuous. Testing moderation often involves creating a mean-centered interaction term and entering it into a regression model to see if it significantly predicts the DV (Mediation and Moderation Guide, n.d.). Mediators clarify mechanisms, while moderations clarify conditions. Understanding the difference is important because misidentifying them can lead to incorrect conclusions about causality or generalizability.

Huang et al. (2022) tested self-worth as both a mediator and moderator in the link between teacher-student relationships (TSR) and student engagement in rural Chinese schools. Using structural equation modeling, they found that TSR affected engagement partly through self-worth (mediation). Through multigroup analysis, they also showed that the TSR engagement effect was stronger among students with lower-self-worth (moderation). This dual – role model demonstrates how a single variable can operate in both capacities and emphasizes why correctly identifying mediators vs moderators is vital to empirical research. 

References

Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51(6), 1173–1182. https://doi.org/10.1037/0022-3514.51.6.1173

Huang, J., Sang, G., & Chao, T. (2022). Self-Worth as a mediator and moderator between Teacher-Student Relationships and Student Engagement in rural schools. Frontiers in Psychology, 12. https://doi.org/10.3389/fpsyg.2021.777937

MacKinnon, D. P. (2008). Applications of the Mediation Model. In Introduction to Statistical Mediation Analysis.

Mediation and Moderation Guide. (n.d.).