Subject and Keywords:
The issue of predictive validity of school grades relative to large-scale standardized tests is described. The extent to which the results from standardized external exams at the end of lower secondary school from 2012–2013 can be predicted on the basis of school grading (both single and average grades) was estimated. Due to the hierarchical structure of analysed data, two-level linear regression models were used. In order to estimate the level of class selectivity, IRT modelling was used. The results indicated that single school grades and their averages are very good predictors for standardized test results, both at individual and class levels. Further, intra-class correlation coefficients demonstrated a high degree of class selectivity. The results are practically applicable to high school recruitment, particularly to student choice of class.