Coleman (1988, 1990) has formulated the assumption that social capital lies in social structures and facilitates certain actions such as the accumulation of human capital. Close social ties between actors in the educational environment should therefore exert independent effects on a child's academic performance. Considering the early selection into the different types of secondary education in the German education system after fourth grade, not only social relations at this decisive point in time become important, but also changes across time towards this transition. Therefore, using longitudinal data (from the BiKS-8-12 study) is of special advantage in this field of research which is dominated by cross-sectional analyses. In this paper, we apply multinomial logistic regressions in order to investigate whether family-school relations are important for academic achievement: Independent effects of student-student, student-teacher, and parent-school relations could be found. Children with good social relations to their teachers and classmates, children with high performing friends, and parents who engage in school activities have significantly better chances in reaching a high performance level. Analyses with growth curve models show in addition that changes in student-teacher interactions, student-student interactions and changes in parental involvement contribute to a better school performance.

Keywords
social relations, social capital, longitudinal studies, academic achievement