In many educational systems, grades are the main instrument for measuring student success in school, hence, grade assignment have significant individual and societal consequences. Generally, two grading systems can be identified: norm-referenced grades, which are based on the assumption that individuals’ knowledge and skills vary in relation to the knowledge and skills of other students, and criterion-referenced grades, which are based on the evaluation of student performance in relation to defined goals and criteria. Norm-referenced grades are primarily designed for the purpose of selection, and criterion-referenced grades are intended primarily for providing information about the knowledge and skills that the students have acquired. Nevertheless, grades are used for multiple purposes. In the early 1990s, the Swedish school system was decentralized through major school reforms, leaving the responsibility for primary and secondary education to the local municipalities. Another part of the reform was that a norm-referenced grading system was replaced by a criterion-referenced, which took full effect from 1998. Gustafsson and Yang-Hansen (2009) showed that the variation between schools in final grades from the Swedish compulsory school increased annually over the years 1998-2007, whereas it was almost constant during the period from 1988 to 1992. However, the largest variability took place between 1998 and 2004. One problem in finding explanations for the increasing between-school variation in grade levels during this period, is that it is not clear if the variation is due to differences between schools in the actual achievement of knowledge and skills, or whether it reflects differences between schools in degree of rigor in the assessment of students’ knowledge and skills. Gustafsson and Yang-Hansen (2009) also found an increase in the mean level of the GPA of 0.16 standard deviation units between the years 1998-2007. The increase varied considerably between different subject matter areas, which seemed to be related to whether national tests were available or not. This suggests that the grades are not reliable indicator of the level of students’ knowledge and skills. The changes on the variability between schools and the changes in levels of grades over different subject matter areas need to be analyzed with multivariate, multilevel, methods to be adequately described and understood. The purpose of the present study is to investigate changes in the internal structure of grades at individual and school levels over a 19-year period that includes cohorts from the two different educational and grading systems.

Data used in this study is derived from The Gothenburg Educational Longitudinal Database (GOLD), which contains a rich amount of information about every individual in Sweden born during the years 1972-1992. The analyses were based on marks for specific subjects from the final year of compulsory school, for those who left compulsory school during the years 1989-2007. Each cohort comprised approximately 100 000 students and about 1000 schools. Among the students, 46% have norm-referenced grades, and 54% have criterion-referenced grades. The analyses were performed by the use of two-level confirmatory factor analysis at school and students levels. The 19 cohorts were analyzed one at a time.

Previous one-level analyses of the dimensional structure of the norm-referenced grades (Andersson, 1998) and the criterion-referenced grades (Giota, 2001) indicate that four or five dimensions are needed to describe the internal structure of grades: a general factor and specific dimensions for math-science, language, spatial-practical, and aesthetic subjects. In the two-level analyses we expect that for the norm-referenced grades there will at individual level be a similar pattern as in these previous studies, while at the school level there will only be a general factor. For the criterion-referenced grades we expect a successively more elaborate dimensional structure at the school level, reflecting the fact that interpretation and implementation of the national curricula and the grading criteria differs between schools.