Description: Analysis of PISA 2003 (additional dataset for Switzerland)We first address the question to what extent the socio-economic composition of classes and schools influences individual student performance in the PISA assessment. Under socio-economic composition we understand the share of foreign-speaking students, of non-natives, of low socio-economic students, etc. within classes and schools. Second, we investigate whether the composition affects class and school climate and how the latter are interrelated with performance. Third, we consider features of the education systems such as school tracking, the time of instruction and analyse their impact on the variation in student performance. Up to now many analyses of the PISA data either focused on the student and school perspective or the system level. We do believe that it is crucial to combine the different levels into one single model that aims at explaining variation in student performance. Methodology: Multilevel models (hierarchical linear models) will be estimated that correctly take into account the clustering of the data, i.e. students are nested within classes, classes within schools and schools within cantons. Conclusions: We expect variables of all the different levels (class, school, system) to influence student performance. With regard to public policy and system evaluation we are particularly interested in the impact of the system level variables after controlling for all the other effects.