This paper examines educational outcomes of pupils selected to secondary school types by different tracking regimes in a German state: The traditional regime of streaming pupils after fourth grade of elementary school is compared to a regime in which pupils are selected into different secondary school tracks after sixth grade. Descriptive evidence demonstrates that the proportion of pupils reaching the highest level of secondary education is relatively small for those who attended later tracking schools. Additionally, the incidence of track modification is relatively frequent for schools with a high proportion of incoming pupils from the later tracking regime. However, less favorable educational outcomes of the later tracking schools are due to self-selection of relative low performers into these schools: The downward bias in estimating tracking regime effects is reduced considerably by controlling for a broad variety of socio-economic background characteristics. Corresponding regression results mainly indicate that there are no negative effects of later tracking on observed educational outcomes measured in the middle of secondary school. Regression analyses for different subgroups suggest that the reading performance of immigrant pupils is better under the later tracking regime compared to the early tracking system.