12. SPSS Hayes Process Macro - Model 14 - Moderated Mediation
Based on Research With Fawad's video on YouTube. If you like this content, support the original creators by watching, liking and subscribing to their content.
Use PROCESS Model 14 when the goal is to test whether a moderator changes the strength of an indirect effect (not just a direct effect).
Briefing
Moderated mediation in SPSS PROCESS Model 14 shows that role ambiguity doesn’t just affect the direct link between collaborative culture and organizational performance—it also changes how strongly organizational commitment translates into performance through collaborative culture. In this setup, organizational commitment (OC) predicts collaborative culture (CC), CC predicts organizational performance (OP), and role ambiguity (RA) moderates the CC→OP path, thereby conditioning the indirect effect OC→CC→OP.
The analysis starts with the mediator model for CC. Organizational commitment has a statistically significant effect on collaborative culture (p < .05), with confidence intervals for the OC coefficient not crossing zero—evidence that OC reliably shifts CC.
Next comes the outcome model for OP, where both main effects and the interaction matter. Organizational commitment significantly predicts organizational performance (p < .05). Collaborative culture also significantly predicts organizational performance (p < .05). Role ambiguity independently has a significant negative effect on organizational performance (t > 1.96, p < .05), again with confidence intervals that exclude zero.
Crucially, the interaction between collaborative culture and role ambiguity is significant (t > 1.96, p < .05, and the confidence interval for the interaction does not include zero). The negative sign on the interaction indicates that role ambiguity dampens the positive relationship between collaborative culture and organizational performance. The interaction also produces a significant change in explained variance (R-square change significant; p < .05), reinforcing that RA meaningfully alters the strength of the CC→OP link.
Johnson–Neyman probing clarifies where the moderation holds. The conditional effect of CC on OP is significant across low to moderate levels of role ambiguity, but moderation fades beyond a threshold: as RA increases past that point, the p-value rises above .05 and the confidence interval includes zero. In plain terms, higher role ambiguity weakens the payoff of collaborative culture for organizational performance.
The moderated mediation results connect this moderation to the indirect pathway. The indirect effect of OC on OP through CC is significant at the mean level of RA (reported conditional indirect effect around .1673, with bootstrap confidence intervals excluding zero). Probing at one standard deviation below and above the mean shows the indirect effect remains significant at low RA (about .236) but shrinks at average and high RA—roughly halving or more as RA increases.
Finally, the index of moderated mediation confirms that the indirect effect varies significantly with role ambiguity (confidence interval excludes zero). The reported index is negative (index = −0.056), meaning role ambiguity negatively moderates the indirect effect—OC’s influence on OP via CC becomes weaker as RA rises. The write-up concludes that hypothesis H1 is supported: the indirect effect OC→CC→OP is moderated by role ambiguity, and the moderation works in the direction of attenuation rather than amplification.
Cornell Notes
The analysis uses SPSS PROCESS Model 14 to test moderated mediation: organizational commitment (OC) affects organizational performance (OP) indirectly through collaborative culture (CC), while role ambiguity (RA) changes the strength of that indirect effect. OC significantly predicts CC, and CC significantly predicts OP. The CC→OP relationship is moderated by RA: the interaction is significant and negative, indicating RA weakens the positive effect of CC on OP. Johnson–Neyman probing shows the moderation is significant at lower-to-mid RA levels but becomes non-significant beyond a threshold. The indirect effect OC→CC→OP is significant at low, mean, and high RA, yet it decreases as RA increases; the index of moderated mediation is significant (index = −0.056), confirming that RA significantly conditions the indirect effect.
What must be true for moderated mediation in this model, and which results support those conditions?
How does role ambiguity change the CC→OP relationship?
What does the Johnson–Neyman output add beyond a simple interaction p-value?
Why is the mediation described as partial rather than full?
How do conditional indirect effects demonstrate moderated mediation?
What does the index of moderated mediation (−0.056) mean in this context?
Review Questions
- In this Model 14 setup, which specific coefficient(s) indicate that role ambiguity moderates the indirect effect rather than only the direct effect?
- How would you use Johnson–Neyman results to report the range of role ambiguity values where collaborative culture significantly predicts organizational performance?
- What evidence distinguishes partial mediation from full mediation in these results?
Key Points
- 1
Use PROCESS Model 14 when the goal is to test whether a moderator changes the strength of an indirect effect (not just a direct effect).
- 2
Confirm the model inputs: OC as X, CC as mediator, OP as Y, and RA as the moderator on the indirect-path-relevant relationship (CC→OP).
- 3
OC significantly predicts CC, establishing the first leg of the mediation pathway (p < .05; confidence interval excludes zero).
- 4
The CC→OP relationship is moderated by RA: the CC×RA interaction is significant and negative, meaning RA dampens the CC effect on OP.
- 5
Johnson–Neyman identifies the RA range where the CC→OP slope is significant; moderation disappears beyond a threshold where p > .05 and confidence intervals include zero.
- 6
The indirect effect OC→OP through CC is significant at low, mean, and high RA but decreases as RA increases, indicating moderated mediation.
- 7
Report the index of moderated mediation (here, −0.056) and its confidence interval excluding zero to support the claim that RA significantly moderates the indirect effect.