All school systems have to answer the question, how to deal with different student abilities. Mostly, some kind of ability grouping or tracking is applied in schools or classes. It is a challenge to put students in homogeneous learning groups according to their abilities while not penalising some by limiting their opportunities and perspectives. Thus, ability grouping of students is one of the most controversial issues in educational research (Slavin, 1987). The school system in Germany is highly differentiated. While elementary school education is provided for all children in a single track system, students should be allocated to different tracks in secondary schools (Hauptschule, Realschule, Gesamtschule, Gymnasium) according to their achievement in primary school. The secondary tracks are typified on the number of grades, the study programmes and the certificates they offer. One consequence of this differentiation is social segregation as recent research has shown: The average student attainment scores and social background characteristics differ greatly between the tracks. Thus, it is obvious that one finds school composition effects as well as institutional effects of tracking in the German school system. Institutional effect in this context means that the students’ learning results vary between the tracks, even when controlling for individual intake characteristics and the student composition. It is seen as problematic from an educational and equity perspective that the development of student achievement depends on the chosen school track, even for comparable student pre-conditions. So far, most research models school effects on average attainment scores. There are hardly any studies which focus on institutional effects on learning progress as a dependent variable. The research design of our study allows to estimate school level effects for individual learning gains. Our analysis focuses on two main questions: Firstly, we estimate average learning gains for students during a time period of four years in secondary school. Secondly, we can describe influences on those growth rates. Therefore, we model school composition characteristics as well as effects of secondary school tracking.

We use data from a longitudinal achievement study in the federal state Hamburg (Germany) that cover one cohort of students (N = 9,425). This dataset embraces a time period of four school years with three measurement times at the end of grade 4, grade 6 and grade 8. We choose a three-level hierarchical linear regression model (using the software HLM). Therefore, we define the time point of measurement at level 1. The student data are specified at level 2, and the school level data at level 3. This methodological approach allows to identify simultaneously the multilevel structure of the data, overall growth trajectories averaged across the sample (fixed effects) but also variation in learning rates between students and between schools (random effects). Additionally, this method allows a decomposition of the variance into within- and between-school components.

The students’ achievement level at the end of grade 4 depends on sex, migration background and cognitive ability. There is a strong effect of the school level: Students starting their secondary school career at a Gymnasium show average achievement scores which are one standard deviation higher than those of students attending a Hauptschule. The effect of Realschule is smaller but still statistically significant. These results indicate that the allocation to different secondary tracks is mainly based on prior student achievement as it is intended. The growth rates at individual level depend on the students’ cognitive ability (positive effect) as well as their prior knowledge. As prior knowledge shows an unexpected negative effect, methodological issues as regression to the mean have to be discussed at this point. At school level, students attending a Gymnasium show higher growth rates than those attending a Hauptschule, even when controlling for individual and composition effects.