Psychometric evaluation of gender norm beliefs and attitudes among students in Ghana: evidence from the GNAS and GEM Scales

BACKGROUND: Robust measurement of gender-related beliefs and attitudes is essential for understanding how gender norms operate in higher-education settings. However, little is known about how these constructs are structured among Ghanaian university students or whether widely used instruments retain validity in this context. This study examined the factor structure, reliability, and measurement invariance of adapted versions of the Gender Norm Attitudes Scale (GNAS) and Gender Equitable Men (GEM) Scale among Sociology and Social Work students in Ghana. METHODS: A cross-sectional survey was conducted among 631 students (98.3% response rate) from two public universities, split for Explorative Factor Analysis (EFA) (n = 339) and Confirmatory Factor Analysis (CFA) (n = 364). EFA used principal axis factoring with oblique rotation on polychoric correlation matrices; parallel analysis guided factor retention. CFA was estimated under maximum likelihood and WLSMV estimators. Reliability was assessed using Cronbach's alpha and McDonald's omega. Measurement invariance was tested across sex, university, programme, and year level using the standard nested sequence of configural, metric, and scalar models, and was evaluated by changes in fit indices (ΔCFI, ΔRMSEA, ΔTLI, and ΔSRMR); differential item functioning (DIF) was assessed using ordinal logistic regression. Data were managed and described using SPSS Version 32; all factor-analytic, reliability, invariance, and DIF analyses were conducted in R (psych and lavaan packages). RESULTS: The GNAS yielded a two-factor, 11-item structure reflecting beliefs supporting men's privileges (α = 0.824, ω = 0.868) and equity for girls and women (α = 0.830, ω = 0.898), with strong fit (CFI = 0.960; TLI = 0.949; RMSEA = 0.057). Polychoric EFA confirmed this, indicating ordinal estimation did not alter the solution. After empirically driven item reduction, the adapted GEM Scale emerged as a modified version of the original instrument, with narrower domain coverage and near-complete loss of the Reproductive Health and Disease Prevention domain; the retained 16-item, four-factor structure covered violence, domestic chores and daily life, sexual relationships, and household authority and obedience, explaining 52.6% of variance with acceptable fit (CFI = 0.888; TLI = 0.865; RMSEA = 0.058), though the Household Authority and Obedience factor rested on only two items and the subscales showing weak Cronbach's alpha values. The students generally rejected inequitable norms, though attitudes toward household authority and domestic roles were more variable. For the GNAS, configural and metric invariance held across sex and university (ΔCFI ≤ 0.005); scalar invariance held across university (ΔCFI = 0.006) and was only partial across sex (ΔCFI = 0.022). For the adapted GEM Scale, metric invariance held across university (ΔCFI = 0.006) but not sex (ΔCFI = 0.016), and scalar invariance was not supported in either grouping. CONCLUSION: Although preliminary, the adapted GNAS shows acceptable structural validity, adequate-to-strong reliability, and metric invariance across sex and university and scalar invariance across university. In contrast, the adapted GEM Scale should be interpreted cautiously as a substantially modified measure with reduced domain coverage and less stable invariance properties, for which scalar invariance is not yet established. Further validation, confirmation of these invariance results using ordinal (WLSMV) estimators, and exploration of 5- or 7-point response formats are needed before either adapted scale is used confidently in broader research or programme evaluation.