Moderation Analysis in SPSS with  Test moderation in SPSS with PROCESS Model 1: interaction term, simple slopes, Johnson-Neyman and interaction plot, with a reporting example.

Moderation asks when or for whom an effect is stronger or weaker. In SPSS, the simplest way to test it is PROCESS Model 1: you enter X, Y and the moderator W, and PROCESS creates the interaction term for you. The key result is the coefficient of X × W. If it is significant, you then probe it with simple slopes and a plot. This guide covers every step.

Moderation in one example

Attitude toward sustainable fashion (X) usually increases purchase intention (Y). But the effect may be weaker for consumers who are very sensitive to price (W). Here, price sensitivity moderates the attitude-intention relationship. The regression model is:

Y = b0 + b1X + b2W + b3(X × W)

Moderation exists if b3 is significant.

Before you start

• PROCESS must be installed. If it is not, follow the installation steps in our guide on mediation with PROCESS Model 4.
• Use composite scores (the mean of the items) for X, W and Y.
• W can be continuous (a scale score) or categorical (for example gender coded 0 and 1). For a categorical moderator with more than two groups, use the Multicategorical option.

Run Model 1 step by step

1. Go to Analyze > Regression > PROCESS.
2. Move PI to Y variable, ATT to X variable and price sensitivity (PS) to Moderator variable W.
3. Set Model number to 1.
4. Click Options:
   • Mean center for construction of products: choose Only continuous variables that define products. Centering makes b1 and b2 easier to interpret; it does not change the interaction test.
   • Conditioning values: choose 16th, 50th, 84th percentiles or -1SD, Mean, +1SD.
   • Tick Johnson-Neyman output.
   • Tick Generate code for visualizing interactions.
5. Click Continue, then OK.

Read the output

Model summary and coefficients

The interaction is negative and significant (b3 = -0.15, p = .004). Below the table, PROCESS reports the R-square increase due to interaction, here ΔR² = .02, F(1, 296) = 8.53, p = .004. Small R² changes are normal for interactions in survey research.

Conditional effects (simple slopes)

Because the interaction is significant, PROCESS shows the effect of X at low, medium and high values of W:

The effect of attitude stays positive at every level, but it shrinks as price sensitivity increases. This is a buffering (weakening) moderation.

Johnson-Neyman technique

The Johnson-Neyman output shows the exact value of W where the effect of X stops being significant. If PROCESS reports no transition point within the observed range of W, the effect is significant for all respondents, as in our example.

Plot the interaction

At the end of the output, PROCESS prints a short block of SPSS syntax (DATA LIST ... GRAPH). Copy it into a new syntax window (File > New > Syntax), run it, and SPSS draws a scatter plot of the predicted values. Double-click the chart, add lines between points with the same W value, and label them "Low", "Medium" and "High" price sensitivity. Reviewers expect this figure.

How to report moderation

Moderation was tested with PROCESS Model 1 (Hayes, 2022), with continuous variables mean-centred. The interaction between attitude and price sensitivity was significant (b = -0.15, SE = 0.05, p = .004, ΔR² = .02). Simple slope analysis showed that the positive effect of attitude on purchase intention was stronger for consumers with low price sensitivity (b = 0.71, 95% CI [0.57, 0.85]) than for those with high price sensitivity (b = 0.41, 95% CI [0.27, 0.55]).

Common mistakes

• Interpreting b1 as a "main effect". With an interaction in the model, b1 is the effect of X when W equals zero (or its mean, if centred). It is a conditional effect.
• Splitting the sample at the median. Median splits lose information and power. Keep W continuous.
• Forgetting the plot. A significant b3 without a plot and simple slopes is hard to interpret.

FAQ

Does centering fix multicollinearity?

Centering reduces the correlation between X and X × W, but it does not change the test of the interaction. It mainly helps interpretation (Dawson, 2014).

My interaction is not significant. Should I still report simple slopes?

No. Report the non-significant interaction and conclude that the moderation hypothesis is not supported.

Can I use gender as a moderator?

Yes. Code it 0 and 1. PROCESS then gives the effect of X for each group.

What is the difference between moderation and mediation?

Mediation explains how X affects Y (through M). Moderation explains when the effect changes (depending on W). Both can be combined in moderated mediation with Model 7.

References

• Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Sage.
• Dawson, J. F. (2014). Moderation in management research: What, why, when, and how. Journal of Business and Psychology, 29(1), 1–19. https://doi.org/10.1007/s10869-013-9308-7
• Hayes, A. F. (2022). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach (3rd ed.). Guilford Press.