Estimating Neighbourhood exposure Effects in the Low Countries

Abstract
Recent quasi-experimental studies provide robust evidence that childhood neighbourhood of residence matters for long-run income and schooling in the United States, Australia and Canada. These studies exploit variation in the age of children when families move. We replicate this analysis in the context of a country with a relatively low level of income inequality and a relatively high level of social mobility. Exposure effects appear to be less important in this context. Our baseline estimates of one year of exposure suggest a convergence rate of two percent in the Netherlands versus four percent in the United States. Rich administrative data allow us to perform two novel empirical tests. First, previous studies assume that children move with their parents. Our data show that nearly one out of two parental moves does not include moves of their children. Focusing on children actually moving with their parents reduces the estimates of exposure effects by fifty percent. Second, we use the outcome of a high-stakes cognitive test taken at the age of 12 to investigate exposure effects before and selection effects after this age. We find that children moving after the test on average have a lower coefficient on the predicted test score rank. Selection especially occurs for parental moves in the years around the child’s transition from primary to secondary education. These findings suggest that estimates of childhood exposure effects are sensitive to measurement error in parental moves, and that parental moves appear to be selective at specific child ages. Joint paper with Bas ter Weel and Céline Odding.