Moderation Analysis with Categorical Variables using SmartPLS3

The analysis tests whether “type of Bank” (public vs. private) changes how two predictors relate to “collaborative culture” in SmartPLS. The core finding is that bank type significantly moderates both relationships: (1) perceived organizational support (POS) → collaborative culture and (2) organizational commitment (OC) → collaborative culture. Adding these moderation effects increases the model’s explanatory power, indicating that bank type meaningfully changes the strength of the underlying links.

Before running moderation, the workflow starts with creating dummy variables for the categorical moderator. “Type” is coded so that one category is the reference group, and only one dummy variable is used in the model. In the example, the dummy coding is arranged such that the moderator value 0 represents private sector banks and 1 represents public sector banks (with the alternative dummy treated as the reference). This setup is essential because SmartPLS moderation with categorical variables relies on interaction terms built from the dummy-coded moderator.

The moderation analysis is implemented using the product indicator approach. That choice depends on how the constructs are measured: because the independent variables are treated as reflective, the product indicator approach is selected. For the interaction term generation, the analysis uses the unstandardized option, and the interaction effects are added by selecting the endogenous variable that each predictor moderates. Two separate moderating effects are created: one for OC × type of bank and another for POS × type of bank. Each moderating effect is renamed to keep results interpretable.

Model evaluation begins with a baseline run (without moderation). The R² value is 0.542. After adding the interaction effects, the model is re-run to obtain the updated R². With moderation included, R² rises to 0.570. The difference between included and excluded R² is about 0.028—small in magnitude, but reported as statistically significant. Bootstrapping is used to test significance, with 500 resamples in the demonstration and a two-tailed test because no directional hypothesis is imposed.

To interpret moderation beyond significance, the analysis uses interaction plots and focuses on line steepness by group. For POS → collaborative culture, the public sector group (moderator = 1) shows a steeper positive gradient than the private sector group (moderator = 0). The reported coefficients used for plotting include a POS path coefficient of 0.428, a moderator main effect of −0.012, and an interaction coefficient of 0.345. The steep public-sector slope implies that higher POS is more strongly associated with higher collaborative culture in public sector banks.

For OC → collaborative culture, the moderation pattern reverses. The interaction coefficient is negative (−0.353), meaning the OC → collaborative culture relationship is stronger in private sector banks than in public sector banks. The OC path coefficient is reported as 0.387, and the negative interaction indicates that as the moderator shifts from private (0) to public (1), the OC effect weakens. In practical terms, organizational commitment predicts collaborative culture more strongly in private sector banks, while perceived organizational support predicts it more strongly in public sector banks.

Overall, the procedure demonstrates how to operationalize categorical moderation in SmartPLS using dummy coding, product indicator interaction terms, bootstrapping, and interaction plots to determine which group shows the stronger effect.