Causal effects of foster care: An instrumental-variables approach

► Measuring causal effects of foster care on child outcomes can be confounded by selection bias. ► Instrumental variable estimation is proposed to explore natural experiments. ► Application is the effective randomization of child-protection investigators to cases. ► Estimates suggest foster care causes higher rates of juvenile delinquency among marginal cases. ► Estimates suggest little effect of foster care on emergency medical care among marginal cases.