Although many studies have focused on the importance of school composition for student achievement, there is still no consensus on "whether" school composition matters to student achievement, and, if so, "why". Therefore, the present study investigates the association between school composition and mathematics achievement at the end of second grade in Flanders. International research points to the initial ability level, SES, ethnicity and sex composition of the school as potential variables in explaining differences in student achievement. Moreover, some researchers suggest that schools "react" to their student body and for that reason we investigated the possible association between school composition and school processes. Data from the SiBO Project have been analyzed using multilevel regression and multilevel mediation analysis. The results showed no direct school composition effects with respect to prior achievement, SES, ethnicity and sex on math achievement. We found two small differential effects, indicating that mean school prior achievement seems to positively affect initially high achievers, and the proportion of minority students in school seems to negatively affect students speaking a non-European language except for Turkish, Arabic or Berber at home. Furthermore, two small indirect effects were found which suggest that schools with a high mean prior achievement or a high mean SES keep in regular contact with their students' parents and this, in turn, appears to enhance students' math achievement. Overall, our results seem to indicate that school composition in the early years of primary education hardly matters.