Categorical Predictor/Dummy Variables in Regression Model in SPSS

Categorical predictors like gender and country can’t be entered directly into a standard linear regression in SPSS because they aren’t metric variables. The practical fix is to convert each categorical variable into dummy variables, then include only enough dummies in the regression to leave one category as a reference point for comparison. In the example, gender has two categories (male, female), so it becomes two dummy variables, but only one dummy is entered into the regression—female is treated as the reference category (coded as 0), while male is the comparison group (coded as 1).

Using SPSS’s Transform → Create Dummy Variables, the workflow generates dummy fields such as gender_1 for male and gender_2 for female (renamed for clarity). In the regression setup (Analyze → Regression → Linear), customer loyalty is the dependent variable, and the independent variable is the dummy-coded gender predictor. With female as the reference category, the regression output indicates whether the male group differs significantly from females. The results show the effect is statistically significant (p < 0.05), and the coefficient is positive (reported as about 0.19), which the analysis interprets as males having higher customer loyalty than females. The write-up also distinguishes statistical significance from practical magnitude: the effect is significant, but described as not substantial because the coefficient value is relatively low.

The same logic scales to categorical variables with three or more categories. Country is treated as a categorical predictor with three groups—China, Pakistan, and Italy. After creating dummy variables for country, the regression includes only two of the three dummies, leaving one category as the reference. Here, China is selected as the reference category, while Pakistan and Italy are entered as comparison predictors. The regression results show that both Pakistan and Italy differ significantly from China in customer loyalty, with negative coefficients indicating lower loyalty scores relative to the reference group. Each difference is evaluated for significance, and the output indicates that the gaps are statistically meaningful (again, p < 0.05).

To make the interpretation intuitive, the analysis also uses mean comparisons: China shows the highest average customer loyalty, while Pakistan and Italy show lower averages. The mean analysis aligns with the regression findings, and the key question—whether the differences are significant—is answered through the regression coefficients and their p-values.

Overall, the core takeaway is a clear rule for regression with categorical predictors in SPSS: create dummy variables, include only k−1 categories in the model for k groups, and interpret each coefficient as the difference between that category and the chosen reference category. With that structure, gender and country both emerge as significant predictors of customer loyalty in the hospitality context described.