Exploratory Factor Analysis in SPSS for Scale Development

Exploratory factor analysis (EFA) in SPSS is used to test whether a newly built multi-dimensional scale actually behaves the way theory predicts—specifically, whether items cluster into the intended factors. In this workflow, the scale targets “university social responsibility,” operationalized through three dimensions: ethical responsibilities, research and development responsibilities, and philanthropic responsibilities. The practical goal is straightforward: run EFA, check whether a three-factor structure emerges, and then prune items that fail to load cleanly.

The scale development process leading into EFA follows a two-source logic. First, existing literature is scanned to collect items used in prior measures of social responsibility, with an explicit effort to address a gap: many existing instruments focus on business organizations rather than universities. Second, focus group discussions and interviews add new candidate statements. Keywords from these qualitative sessions—such as the idea that universities should have a code of conduct—are converted into survey items. At the end of this stage, the construct sits at the top of a hierarchy: construct → dimensions → items, with theory predicting which items belong to which dimension.

EFA in SPSS begins by selecting the items under “Analyze → Dimension Reduction → Factor.” Several diagnostics are used to justify factor analysis. The Bartlett’s test of sphericity is checked for significance, indicating that correlations among variables are sufficiently patterned for factor extraction. The Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy is also reported; here it reaches 0.931, signaling strong suitability. Extraction uses principal component analysis, and the analysis retains three factors based on the eigenvalue rule (factors with I value greater than one) and the cumulative explained variance, which lands at roughly 60% for the three-factor solution.

Rotation is then applied (Varimax rotation) to clarify which items load onto which factor. Initial results show that most ethical responsibility items load well together, and philanthropic responsibility items also form a coherent cluster. However, problems appear in the research and development dimension: one item (RD2) loads weakly or onto the wrong factor, and two items (RD1 and PR1) show no meaningful loading (below the 0.050 threshold). The workflow treats this as a signal to remove problematic items rather than forcing a messy structure. After removing RD2, then RD1, and then PR1—one by one, with re-runs to see how the structure stabilizes—the factor pattern becomes clean and matches the expected three-factor model.

Reporting the results is treated as part of the method, not an afterthought. The write-up should include the extraction method (principal component analysis), rotation (Varimax), the minimum factor loading criterion (0.050), and the Bartlett test and KMO values. Commonalities and model fit are also documented, including the proportion of non-redundant residuals (kept below 50% as an indicator of acceptable fit). Importantly, the process emphasizes that item deletion should be justified: items are removed because they fail to load on the intended factor or fail to load meaningfully at all. The final output is summarized in a factor-loading table with Factor 1, Factor 2, and Factor 3 labeled according to the three dimensions—ethical responsibilities, research and development responsibilities, and philanthropic responsibilities—producing a structure that aligns with the original theoretical expectations.