Research has found that grades are the most valid instruments for predicting educational success. Why grades have better predictive validity than, for example, standardized tests is not yet fully understood. One possible explanation is that grades reflect not only subject-specific knowledge and skills but also individual differences in other aspects. The purpose was to investigate the relative importance of knowledge and skills and other aspects encapsulated in grades for the predictive validity of compulsory school grades for educational success in upper secondary school. Structural equation modelling was used. Participants were 9th-grade students from 3 birth cohorts, each comprising full populations of approximately 100,000 students. The results showed that the subject-specific factors and an additional common grade factor contributed to the predictive validity. Effects of gender and parents' education were found in the common grade factor, with girls and students with a lower educational background being advantaged. (Contains 3 tables and 4 figures.)