How to Run Mediation Analysis with PROCESS Macro in SPSS

Mediation analysis answers one of the most important questions in research: why does X affect Y? Instead of just showing that a relationship exists, mediation tests whether a third variable (the mediator) explains the mechanism through which X influences Y. The PROCESS macro by Andrew Hayes is the most widely used tool for this analysis in SPSS.

What Is Mediation?

In simple mediation, you have three variables:

• X (Independent Variable): The predictor
• M (Mediator): The variable that transmits the effect of X to Y
• Y (Dependent Variable): The outcome

The key question is: Does X affect Y through M?

Example: Does transformational leadership (X) affect employee performance (Y) through job satisfaction (M)?

The Paths in Mediation

• Path a: X → M (does X predict the mediator?)
• Path b: M → Y (does the mediator predict Y, controlling for X?)
• Path c' (c-prime): X → Y controlling for M (the direct effect)
• Path c: X → Y without controlling for M (the total effect)
• Indirect effect (a × b): The mediated pathway — the product of path a and path b

Mediation exists when the indirect effect (a × b) is statistically significant.

Installing the PROCESS Macro

If PROCESS is not already installed:

1. Download it from Andrew Hayes' website (processmacro.org)
2. In SPSS, go to Extensions → Utilities → Install Custom Dialog
3. Select the PROCESS .spd file
4. Restart SPSS
5. PROCESS now appears under Analyze → Regression → PROCESS

Running Simple Mediation (Model 4)

Step-by-Step

1. Go to Analyze → Regression → PROCESS v4 (or wherever it's installed)
2. Set the following:
   • Y Variable: Your outcome variable
   • X Variable: Your predictor variable
   • Mediator(s) M: Your mediator variable
3. Set Model Number to 4 (simple mediation)
4. Under Options, check:
   • Show total effect model (to see path c)
   • Standardized coefficients (for reporting)
5. Under Bootstrap, set:
   • Number of bootstrap samples: 5000 (standard recommendation)
   • Confidence level: 95%
6. Click OK

Interpreting the Output

PROCESS produces several sections of output. Here is what to focus on:

1. Path a (X → M)

Look at the regression table labeled Outcome Variable: M

• The coefficient for X is path a
• Check the p-value — if significant (p < .05), X predicts M

2. Path b (M → Y, controlling for X) and Path c' (Direct Effect)

Look at the regression table labeled Outcome Variable: Y

• The coefficient for M is path b
• The coefficient for X is path c' (the direct effect)
• If path c' is not significant but the indirect effect is, you have full mediation
• If both path c' and the indirect effect are significant, you have partial mediation

3. Total Effect (Path c)

This is the effect of X on Y without the mediator in the model. It should equal c' + (a × b).

4. The Indirect Effect (a × b)

This is the most critical output. Look for the section labeled Indirect effect(s) of X on Y:

• Effect: The indirect effect value (a × b)
• BootSE: Bootstrap standard error
• BootLLCI and BootULCI: Bootstrap confidence interval

Key decision rule: If the 95% bootstrap confidence interval does NOT include zero, the indirect effect is statistically significant. This means mediation exists.

Why Bootstrapping?

The indirect effect (a × b) typically does not follow a normal distribution, so traditional significance tests (like the Sobel test) are unreliable. Bootstrapping:

• Resamples your data thousands of times
• Computes the indirect effect for each resample
• Builds an empirical confidence interval
• Does not assume normality of the indirect effect

The standard recommendation is 5,000 bootstrap samples with bias-corrected confidence intervals.

Interpreting Different Mediation Outcomes

Full Mediation

• Indirect effect significant (CI does not include zero)
• Direct effect (c') is NOT significant
• Interpretation: X affects Y entirely through M

Partial Mediation

• Indirect effect significant (CI does not include zero)
• Direct effect (c') IS also significant
• Interpretation: X affects Y through M, but also has a remaining direct effect

No Mediation

• Indirect effect NOT significant (CI includes zero)
• Interpretation: M does not mediate the X → Y relationship

Inconsistent Mediation

• The indirect and direct effects have opposite signs
• This is rare but valid — the total effect can be small even when indirect and direct effects are individually significant

APA Reporting

Here is a template for reporting mediation results:

A simple mediation analysis was conducted using PROCESS Model 4 (Hayes, 2022) with 5,000 bootstrap samples. The results indicated that transformational leadership (X) significantly predicted job satisfaction (M), b = 0.45, SE = 0.08, p < .001. Job satisfaction (M) also significantly predicted employee performance (Y), controlling for transformational leadership, b = 0.38, SE = 0.07, p < .001. The direct effect of transformational leadership on performance was reduced but remained significant, b = 0.22, SE = 0.09, p = .014, suggesting partial mediation. The indirect effect was significant, ab = 0.17, 95% CI [0.09, 0.27], indicating that job satisfaction partially mediated the relationship between transformational leadership and employee performance.

Beyond Simple Mediation

PROCESS supports many advanced models:

• Model 4 with multiple mediators: Test parallel mediation (multiple M variables simultaneously)
• Model 7: Moderated mediation (first stage moderation — the a path is moderated)
• Model 14: Moderated mediation (second stage moderation — the b path is moderated)
• Model 8 / 15: Combined moderated mediation
• Model 6: Serial mediation (M1 → M2 chain)

Common Mistakes

1. Claiming causation without experimental design — Mediation with cross-sectional survey data suggests a mechanism but does not prove causation. True causal mediation requires experimental manipulation
2. Using the Sobel test — The Sobel test assumes normality of the indirect effect, which is almost never true. Use bootstrap confidence intervals instead
3. Ignoring effect sizes — Report standardized coefficients or the ratio of the indirect to total effect (PM = ab/c) to quantify the proportion mediated
4. Not reporting the bootstrap CI — The CI is the evidence for mediation. Always report it alongside the indirect effect value
5. Confusing moderators and mediators — A mediator explains how/why X affects Y. A moderator explains when/for whom the effect differs

Recommended References

• Hayes, A. F. (2022). Introduction to Mediation, Moderation, and Conditional Process Analysis (3rd edition). Guilford Press.
• Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models.

Need help running mediation or moderation analysis with PROCESS? Our experts handle everything from model selection to APA-formatted results. Get your free quote.