Please join us for our next Population and Health Research group seminar, which is organised jointly between the Centre for Population Change, Scotland and the Population and Health Research group. The seminar is given by:
Juho Härkönen, European University Institute, Florence
Age at parental separation and children's school outcomes in Sweden: A sibling difference analysis
(together with Siddartha Aradhya, Stockholm University)
Parental separation is associated with a host of adverse outcomes, one of which is lower educational performance and attainment. Evidence has accumulated suggesting that part of this association is causal, and recent research has focused on understanding heterogeneity in these effects. The age at which parents separate is one source of such heterogeneity, and previous research has offered inconsistent evidence on whether the effects of parental separation are more or less severe when parents separate when the child is young. In this study, we use a sibling design with population register data from Sweden to analyze the effects of the age of parental separation on children's GPAs at age 15. Non-twin siblings from the same family experience parental separation at different ages, which can be used to identify the effects of variation in age at exposure to parental separation on school outcomes. Our main result is a negative age gradient in the effects of parental separation: the effects of parental separation are the strongest around the ages at which grades are set, and gradually weaken the younger the child was when her parents separated. This result supports the conflict model of parental separation. Additional analyses show that a) children whose parents separated just after the grades were set (placebo test) perform better than those whose parents separated just before the grades were set, but worse than those whose parents separated when the child was young (supporting a "divorce process" perspective), b) regular regression analyses produce the opposite age pattern in the effects (suggesting strong negative selection of parents who separate sooner rather than later), c) sibling fixed effects results are sensitive to error in measuring parental separation, and d) the effects of age at parental separation are heterogeneous. Taken together, our results suggest that in many cases, the children of parents who separate later rather than sooner show worse educational performance.