Selection response of parents to low school quality, for instance large class sizes, is a major problem when estimating causal class-size effects, also in experimental and quasi-experimental studies. To address this problem a new identification strategy using within-school variation over time in...