Our sibling models eliminated confounding from shared unobserved family background characteristics and attempted to restrict the identifying variation to prenatal shocks that induced low birth weight but did not appear to persist into the postnatal environment.