PSYCHOMETRIC VALIDATION OF A HIGH-STAKES ENGLISH LANGUAGE TEST: A RASCH-BASED ANALYSIS OF KAZAKHSTAN’S UNIFIED NATIONAL TESTING

The Unified National Testing is a high-stakes examination that serves as a critical gatekeeper for university admission. Psychometric evidence is necessary to justify score interpretations in high-stakes testing; however, its psychometric properties remain underexplored. The study aims toanalyze the English section of UNT from a psychometric perspective, including the reliability, construct validity, item difficulty, discrimination indices, and how well the test corresponds to the curriculum and CEFR levels. This study examines whether the test adequately measures actual language proficiency.A pilot quantitative study was conducted with 212 secondary school students (grades 10-11) who selected English language subject. Within the framework of Classical test theory (CCT) discrimination indices were calculated while reliability, fit statistics and item targeting were examined using Rasch measurement analysis. Differential Item Functioning was used to compare results of 10thand 11thgrade performances.Psychometric results reveal that UNT’s English component indicated strong internal consistency (Cronbach’s α = 0.879; ω = 0.898). The majority of items showed acceptable measurement as (person reliability = 0.836; item reliability = 0.972). However, several items were relatively easy in grammar and reading sections. The teacher survey revealed insufficient assessment of productive skills and a strong dominance of receptive skills (M = 4.50).The English section of the UNT is generally at an acceptable level in terms of reliability and structural stability. However, demonstrates limited representation of communicative competence and productive skills. Therefore, the findings emphasize the need for score alignment with standardized tests as IELTS with CEFR descriptions.