Although a gap in educational performance of migrant children compared to children without a migration background is to be observed in most industrialized countries, it is particularly big in countries as Belgium, Germany, Austria and the Netherlands, as has been attested by the PISA-data. Social and ethnic segregation, which is particularly high in these educational systems, seems to be one of the important explanatory factors. This project wants to disentangle what are the crucial factors by which this high level of segregation impacts on unequal opportunities for immigrant children. Going beyond the classic composition effect model (looking at peer group effects, i.e. positive or negative influences of pupils on each other), this project wants to also examine the potential impact of differentiated teacher profiles on group performance. The project wishes to test the hypothesis that the link between school composition and educational performance is a (partly) spurious effect, caused by mediating effect of teacher characteristics. We hypothesize that better skilled and more positively oriented teachers are overrepresented in schools with an 'easier' school population, while so-called 'difficult' schools (populated by working-class immigrant children) have difficulty in attracting and - especially - keeping competent and motivated staff. In order to examine this hypothesis a mixed methods approach will be used, combining quantitative statistical analysis (on new and existing data, for instance multi-level analysis of the PISA-data set and other eligible datasets), qualitative case studies and focus groups. Secondary analysis of existing data-sets (PISA, TIMMS, PIRLS) will be undertaken and new data will be collected (taking the Flemish and Francophone educational systems in Belgium as case-studies).