Teachers’ assessment literacy and student science achievement: insights from multilevel analysis of TIMSS data

Abstract This study examined the extent to which school-level variation in science teachers' assessment literacy predicts student science achievement in secondary education, after accounting for student-level differences in socioeconomic status and motivation toward science. Drawing on the U.S. Grade 8 sample of the 2023 Trends in International Mathematics and Science Study (TIMSS), a two-level multilevel modeling approach was employed using the WeMix package, which incorporated complex sampling weights directly into model estimation. School-level predictors; teaching experience, formative assessment frequency, and instructional modification were derived by aggregating science teacher responses within each school. Student science achievement was operationalized using five plausible values pooled via Rubin's rules. A null model indicated that 10.4% of the variance in science achievement was attributable to between-school differences (ICC = 0.104), justifying the multilevel framework. Socioeconomic status (SES) was a statistically significant positive predictor of achievement, while motivation toward science was a statistically significant negative predictor, interpreted in relation to measurement limitations of self-reported motivation scales in large-scale assessments. School-level teaching experience emerged as a statistically significant positive predictor in the full model, while formative assessment frequency and instructional modification did not reach statistical significance. Model fit improved substantially with student-level predictors (ΔAIC = 1,704.6) and modestly with school-level predictors (ΔAIC = 229.4). These findings suggest that student socioeconomic characteristics play a stronger role in predicting science achievement than school-level assessment practices as currently measured in TIMSS. The statistically significant effect of teaching experience highlights accumulated pedagogical expertise as a meaningful school-level resource. The study recommends investment in more nuanced measures of assessment literacy and broader equity-oriented policies addressing socioeconomic disparities in science learning.