Moderation Analysis with the PROCESS Macro in SPSS: A Step-by-Step Guide

Does the effect of one variable on another depend on a third variable? Perhaps study hours raise exam scores — but much more for calm students than for highly anxious ones. Or perhaps a training programme works well for beginners but barely helps experienced staff. When an effect changes depending on a third variable, you have moderation, and it is one of the most common analyses in thesis and dissertation research. In this tutorial, you will learn how to run a moderation analysis in SPSS using the PROCESS macro: click by click, with syntax you can paste, output you can read, and an APA-style write-up you can adapt.

What Is Moderation Analysis?

Moderation analysis tests whether the relationship between an independent variable (X) and a dependent variable (Y) changes across levels of a third variable, the moderator (W). The statistical signature of moderation is an interaction term: the product of X and W (XW). If that interaction term significantly predicts Y, the effect of X on Y depends on W.

In our running example, X is weekly study hours, W is test anxiety, and Y is the exam score. The moderation question is: does the benefit of extra study hours depend on how anxious the student is? If highly anxious students gain less from each extra hour, test anxiety moderates the study-hours effect.

Note the difference from mediation. Mediation explains the mechanism — X works through a middle variable. Moderation explains a condition — X’s effect is stronger, weaker, or even reversed depending on W. In PROCESS, moderation is Model 1; simple mediation is Model 4. They answer different research questions, and many theses test both.

Before You Start: What You Need

Three things before you open SPSS:

  1. Three measured variables. X, W, and Y should each be a single numeric variable (a scale score or a continuous measure). If any of them is a questionnaire scale, compute the scale score first (for example, with Transform > Compute Variable). Moderators can also be categorical — for example, a two-group variable such as gender or treatment group — and PROCESS handles that too, but this tutorial uses a continuous moderator, the most common thesis case.
  2. The PROCESS macro installed. PROCESS is a free macro for SPSS developed by Andrew F. Hayes. Download it from processmacro.org and install it through SPSS’s custom dialog installer (in older versions: Utilities > Custom Dialogs > Install Custom Dialog; in newer versions: the Extensions menu). Once installed, it appears under Analyze > Regression > PROCESS.
  3. A reasonable sample size. Interaction effects are usually smaller than main effects, so moderation analyses need more statistical power than simple regressions. A few hundred cases is a good practical target for a single-moderator model, and larger samples are better when you add covariates.

Also check the usual regression assumptions first: linear relationships between the variables, roughly normal residuals, and no extreme outliers driving the results. Moderation is built on regression, so a violated assumption here is a violated assumption there.

Step 1: Run Model 1, Click by Click

PROCESS Model 1 is the simple moderation model: one X, one moderator, one Y. Here is the exact route:

  1. Go to Analyze > Regression > PROCESS.
  2. In the PROCESS dialog, move your independent variable (for example, StudyHours) into the X variable box.
  3. Move your moderator (for example, TestAnxiety) into the Moderator (W) box.
  4. Move your dependent variable (for example, ExamScore) into the Y variable box.
  5. Set Model number to 1.
  6. Click Options. Tick “Mean center for construction of products” — this centers X and W before forming the interaction term, which reduces multicollinearity and makes the coefficients easier to interpret. Tick “Generate code for visualizing interactions” if you want the plot data, and tick “Johnson-Neyman output” if you want the region-of-significance analysis explained below. Set the number of bootstrap samples to 5000 and the confidence level to 95.
  7. Click OK to run.

If you have control variables — age, gender, or prior grades, for example — add them to the Covariate(s) box. They will be controlled in the model.

SPSS Syntax You Can Paste

If you prefer syntax (and for a thesis, you should keep a syntax record of every analysis), paste this into a syntax window and run it. Replace the variable names with your own:

PROCESS vars = StudyHours TestAnxiety ExamScore
  /y = ExamScore
  /x = StudyHours
  /w = TestAnxiety
  /model = 1
  /center = 1
  /plot = 1
  /jn = 1
  /boot = 5000
  /seed = 12345.

A quick note on the options: /center = 1 mean-centers X and W before computing the interaction, which is the recommended default. /plot = 1 produces the data needed to draw the interaction plot. /jn = 1 requests the Johnson-Neyman analysis. /seed = 12345 fixes the random number seed so your bootstrap results are exactly reproducible — keep the same seed if you ever need to re-run the analysis and match the numbers.

Step 2: Read the Output, Section by Section

PROCESS output is long, so work through it in order:

Model Summary and the Interaction Test

The model summary shows the regression of Y on X, W, and the interaction term XW. Two things matter most here:

  • The coefficient for the interaction term (labelled Int_1) and its p-value. This is the moderation test. If it is significant, the effect of X on Y depends on W, and moderation is supported.
  • The R-square increase due to the interaction. PROCESS reports how much extra variance the interaction term explains beyond the main effects alone. Report this value: it tells the reader how large the moderation effect is, not just whether it exists.

If the interaction is not significant, stop there. There is no moderation to probe, and reporting simple slopes for a non-significant interaction would be misleading.

Conditional Effects: The Simple Slopes

When the interaction is significant, the next table — “Conditional effects of the focal predictor” — shows the effect of X on Y at three values of the moderator: one standard deviation below the mean, at the mean, and one standard deviation above the mean. This is the pick-a-point approach.

In our example, you might find that the effect of study hours on exam scores is strong for low-anxiety students, moderate for average-anxiety students, and weak or non-significant for high-anxiety students. That pattern is the story of your moderation result: describe it in words, not just numbers. Note which slopes are significant and which are not, using the p-values and confidence intervals in the table.

The Johnson-Neyman Region of Significance

The pick-a-point approach only tests three arbitrary moderator values. The Johnson-Neyman technique goes further: it finds the exact range of moderator values for which the effect of X on Y is statistically significant, and the range where it is not. PROCESS reports the transition point (or points) where the effect crosses into significance.

For a thesis, report the Johnson-Neyman finding in one sentence, for example: the effect of study hours on exam scores was significant for test-anxiety scores below a certain value, and non-significant above it. This is more precise than the three-point summary and reviewers like to see it.

Plotting the Interaction

A moderation result should always be plotted. Use the data from the /plot = 1 option (or the values in the conditional-effects table) to draw two or three lines: the relationship between X and Y at low, average, and high levels of W. Lines that fan apart or cross show moderation visually; parallel lines would mean no moderation. In a thesis, this figure usually earns its place in the results chapter, because a reader can grasp the pattern in seconds.

Reporting in APA Style (7th ed.)

Here is a template paragraph you can adapt. Replace the numbers with your own output values:

“A moderation analysis (PROCESS Model 1; Hayes, 2022) examined whether test anxiety moderated the relationship between weekly study hours and exam scores. The overall model was significant, and the interaction between study hours and test anxiety significantly predicted exam scores (b = -0.18, p = .004), adding a further share of explained variance (ΔR² = .04, p = .004). Conditional effects showed that study hours significantly predicted exam scores at low test anxiety (b = 0.52, p < .001) and at average test anxiety (b = 0.34, p < .001), but not at high test anxiety (b = 0.16, p = .089). The Johnson-Neyman technique indicated that the effect of study hours was significant for test-anxiety scores below 28.4 and non-significant above that value.”

A few APA notes: report the interaction coefficient and its p-value, the R-square change for the interaction, the conditional effects at each probed moderator value, and the Johnson-Neyman boundary when you requested it. Always describe the direction of the pattern in words — a reader should understand the finding without decoding the numbers.

Key Takeaways

  • Moderation answers “for whom” or “under what conditions”: the effect of X on Y depends on W.
  • The interaction term (XW) is the moderation test: significant means moderation is supported.
  • PROCESS Model 1 is the standard tool for a single-moderator model in SPSS.
  • Probe a significant interaction with simple slopes at low, average, and high moderator values, plus the Johnson-Neyman region of significance.
  • Always plot the interaction — it is the clearest way to show the pattern.
  • Mean-center X and W so the coefficients are interpretable and multicollinearity stays low.

Frequently Asked Questions

What is the difference between moderation and mediation?

Mediation explains a mechanism: X affects a middle variable, which affects Y. Moderation explains a condition: X’s effect on Y is stronger or weaker depending on W. Moderation is tested with an interaction term; mediation is tested with an indirect effect. They are different questions, and a variable can be a mediator in one study and a moderator in another.

Do I need significant main effects before testing an interaction?

No. A non-significant main effect of X does not rule out moderation. In fact, moderation can hide a main effect: if X helps one group and harms another, the average effect can look like zero. Always test the interaction directly.

Should I center my variables before running moderation?

Yes, mean centering is the recommended default in PROCESS (the /center = 1 option). Centering reduces the correlation between the main effects and the interaction term, and it makes the coefficients interpretable: the coefficient for X becomes the effect of X at the average level of W, rather than at W = 0, which may not be a meaningful value.

My interaction is significant, but one simple slope is not. Is that a problem?

No, that is a normal and interpretable result. It means the effect of X on Y exists at some moderator levels but not others — which is exactly what moderation means. Report all three slopes and describe the pattern, including where the effect disappears.

Can the moderator be a categorical variable?

Yes. PROCESS handles a dichotomous moderator directly: the conditional effects are then reported for each group rather than at standard-deviation points. The interpretation is the same — the effect of X differs between the groups — but you describe it group by group instead of low versus high.

Need this done on your data? Get a free quote → /services/