Psychometric Analysis of High-Stakes Examination Items in West Africa: Evaluating Item Difficulty and Discrimination Across Diverse Socioeconomic and Regional Contexts

This study conducted a psychometric analysis of high-stakes examination items in West Africa, focusing on the relationship between item difficulty and item discrimination across socio-economic, regional, and linguistic backgrounds. Using advanced statistical techniques, including Hierarchical Linear Modeling (HLM) and Structural Equation Modeling (SEM), the study examined how contextual factors influenced test fairness and predictive validity. The analysis utilized a dataset of 12,500 exam items and responses from 250,000 students across five West African countries. Descriptive statistics revealed significant disparities: students from lower socio-economic backgrounds faced higher item difficulty (M = 0.78, SD = 0.12) compared to upper socio-economic groups (M = 0.58, SD = 0.08), while item discrimination was highest among wealthier students (M = 0.60, SD = 0.08). Multilevel regression analysis indicated that socio-economic status significantly predicted item difficulty (β = -0.10, p < 0.01) and item discrimination (β = 0.18, p < 0.01), demonstrating systematic disadvantages for lower-income students. Regional disparities also played a critical role, with rural students encountering more difficult items (M = 0.70, SD = 0.13) than their urban counterparts (M = 0.63, SD = 0.11). Regression analysis confirmed a significant effect of location on item difficulty (β = 0.08, p < 0.01) and discrimination (β = 0.11, p < 0.01), suggesting that rural students faced structural disadvantages in assessment. Linguistic background was another major determinant of assessment fairness. Indigenous language speakers encountered the highest item difficulty (M = 0.75, SD = 0.13) and the lowest discrimination values (M = 0.45, SD = 0.12), compared to English-speaking students (M = 0.60, SD = 0.10) and French-speaking students (M = 0.65, SD = 0.12). Multi-group SEM analysis demonstrated that indigenous language speakers experienced disproportionately higher item difficulty (β = 0.12, p < 0.01) and lower discrimination (β = 0.10, p < 0.01), confirming linguistic bias in standardized testing. The study’s findings highlight structural inequalities embedded within West African high-stakes examinations, raising concerns about fairness and access to educational opportunities. The results underscore the need for equity-driven assessment practices, such as adaptive testing models, differential item functioning (DIF) analysis, and linguistically inclusive test designs. These findings align with global recommendations on assessment fairness and call for urgent reforms in test development and validation.