Abstract
Educational measurement commonly applies Classical Test Theory (CTT) and Item Response Theory (IRT) to estimate students’ achievement scores. CTT derives observed scores from the total number of correct responses, whereas IRT models students’ latent ability by incorporating item characteristics. The one-parameter logistic (1PL) or Rasch model, in particular, estimates ability based solely on item difficulty under the assumption of equal item discrimination. Although Rasch-based approaches have been increasingly adopted in large-scale assessments in Indonesia, empirical evidence comparing Rasch- and CTT-based scoring in religious subject assessments remains limited. This study aims to examine differences in score estimates produced by CTT and the IRT 1PL (Rasch) model, with specific attention to the role of item difficulty. An evaluative quantitative design was employed involving 387 twelfth-grade students from MAN 1 Surakarta who completed a Fiqh achievement assessment in 2023. The dataset consisted of dichotomous responses to multiple-choice items. Students’ ability scores were estimated using both CTT and the Rasch model, and differences between the two scoring approaches were analyzed using paired-sample t-tests, as both scores were derived from the same examinees. The results reveal a statistically significant difference between CTT- and Rasch-based score estimates. The mean Rasch score (M = 71.61) was significantly higher than the mean CTT score (M = 59.18), with t = −13.298 and p < .001. Further analysis indicates that 78% of students experienced lower scores under CTT, while only 21% benefited relative to Rasch-based estimates. These findings suggest that Rasch-based scoring provides a more equitable and measurement-consistent representation of students’ latent ability than CTT, particularly in assessments containing items with varying levels of difficulty. The study underscores the importance of adopting modern measurement models to enhance fairness and interpretability in educational assessment.