Difference between Mediator and Moderator

TL;DR

A mediator explains the mechanism by which an independent variable affects a dependent variable through an intermediate pathway.

Briefing Cornell Notes

Briefing

Mediator and moderator variables both sit in quantitative research models, but they play fundamentally different roles in how independent variables relate to dependent variables. A mediator acts as a bridge: the effect of the independent variable (IV) on the dependent variable (DV) runs through the mediating variable. In plain terms, the IV influences the mediator, and the mediator then influences the DV—so the IV–DV relationship is explained by that intermediate pathway.

A moderator works differently. The IV and DV are already connected, and the moderating variable changes that existing relationship. Depending on the theory, a moderator can strengthen or weaken the effect of the IV on the DV, and in some cases it can even alter the direction of the relationship. Rather than explaining “why” the IV affects the DV (as mediation does), moderation explains “when” or “under what conditions” the IV’s effect on the DV is stronger or weaker.

The transcript uses the same example to make the distinction concrete. For mediation, smartphone usage (IV) affects health (DV) through smartphone addiction (mediator). The logic is sequential: smartphone usage leads to addiction, and addiction leads to health outcomes. If smartphone addiction is the mechanism, then removing it should eliminate the IV’s direct impact on health.

For moderation, smartphone usage (IV) still targets health (DV), but exercise (moderator) changes the strength of that relationship. Under higher levels of exercise, the negative (or otherwise described) impact of smartphone usage on health becomes smaller—meaning the IV–DV link is present, but its magnitude depends on the moderator.

The transcript also distinguishes full mediation from partial mediation. Full mediation occurs when, after accounting for (or removing) the mediator, the direct effect of the IV on the DV drops to zero. Partial mediation occurs when a direct effect remains even after the mediator is removed, indicating that the mediator explains only part of the IV’s influence.

Finally, deciding whether a variable should be treated as a mediator or a moderator comes down to theory. If the variable fits a causal chain—an IV relates to the mediator, and the mediator relates to the DV—it belongs in a mediation framework. If the variable is proposed to alter an already-existing IV–DV relationship—by affecting its strength or direction—it belongs in a moderation framework. Understanding this distinction matters because mediation and moderation require different analytic steps in quantitative research.

Cornell Notes

Mediator and moderator variables differ in how they connect an independent variable (IV) to a dependent variable (DV). A mediator explains the mechanism: the IV affects the mediator, and the mediator affects the DV, so the IV–DV relationship runs through the mediating variable. A moderator changes the relationship itself: the IV and DV are already linked, and the moderator alters the strength (and sometimes direction) of that link depending on conditions. The transcript also distinguishes full mediation (IV’s direct effect becomes 0 after accounting for the mediator) from partial mediation (a direct effect remains). Choosing mediator vs. moderator should follow theory about whether the variable provides a causal pathway or modifies an existing effect.

How does a mediator differ from a moderator in a causal model?

A mediator sits between the IV and DV and explains the pathway: IV → mediator → DV. That means the IV’s effect on the DV operates through the mediating variable. A moderator does not replace the IV–DV relationship; instead, it changes how strong or in what direction the IV affects the DV. In short: mediation targets the mechanism, moderation targets the conditions.

What would full mediation look like after removing the mediator?

Full mediation means the IV’s direct effect on the DV becomes 0 once the mediator is accounted for. In other words, the IV influences the DV only through the mediator. If the direct IV → DV link disappears, the mediator fully explains the relationship.

What would partial mediation look like?

Partial mediation occurs when the IV still has a remaining direct effect on the DV even after the mediator is removed or accounted for. That indicates the mediator explains part of the IV’s influence, but other pathways also contribute to the IV–DV relationship.

Using the smartphone example, why is addiction a mediator?

Smartphone usage (IV) leads to smartphone addiction (mediator), and addiction then leads to health outcomes (DV). The logic is sequential and mechanistic: the effect of smartphone usage on health runs through addiction. Because the mediator sits in the causal chain, addiction functions as a mediator.

Using the smartphone example, why is exercise a moderator?

Smartphone usage (IV) affects health (DV), but exercise (moderator) changes the strength of that effect. The relationship is present regardless, yet it becomes smaller under the moderator’s conditions. That pattern—altering the magnitude of an existing IV–DV link—fits moderation.

How should theory guide whether a variable is treated as mediator or moderator?

If the proposed variable fits a causal sequence (IV relates to the variable, and the variable relates to the DV), it should be modeled as a mediator. If the variable is proposed to affect an existing IV–DV relationship—by strengthening, weakening, or changing its direction—it should be modeled as a moderator.

Review Questions

In your own words, what distinguishes a mediating variable from a moderating variable in terms of the IV–DV relationship?
Give an example of a full mediation scenario and explain what happens to the direct IV → DV effect.
What theoretical criteria would make you classify a variable as a moderator rather than a mediator?

Key Points

1
A mediator explains the mechanism by which an independent variable affects a dependent variable through an intermediate pathway.
2
A moderator changes the strength or direction of an existing independent-variable-to-dependent-variable relationship.
3
Full mediation occurs when the direct effect of the IV on the DV becomes zero after accounting for the mediator.
4
Partial mediation occurs when a direct IV-to-DV effect remains even after accounting for the mediator.
5
Whether a variable is a mediator or moderator should be determined by theory: causal chain supports mediation; conditional influence supports moderation.
6
Mediation and moderation require different analytic approaches because they represent different kinds of relationships.

Highlights

Mediation is a bridge (IV → mediator → DV); moderation is a lever that changes an existing link (IV ↔ DV) under certain conditions.

Full mediation means the IV’s direct effect on the DV drops to 0 once the mediator is included.

In the smartphone example, addiction functions as a mediator because it provides the causal pathway to health outcomes.

In the same example, exercise functions as a moderator because it reduces the impact of smartphone usage on health depending on conditions.

Topics

Mediation vs Moderation
Quantitative Research
Full vs Partial Mediation
Independent and Dependent Variables
Causal Pathways and Interaction Effects