Temporal and geographical variation in the severity of influenza pandemics has been used pervasively in identification strategies for examining the effects of in-utero exposure to influenza on later-life outcomes. However, given the unavailability of individual-level data on influenza exposure prior to birth, previous studies have focused on scrutinising the exogeneity of this variation, when examining reduced-form effects of pandemics themselves. However, to interpret reduced-form effects as causal effects of in-utero exposure to influenza, this variation must implicitly satisfy the instrumental variables (IV) assumptions of excludability and relevance. In this study, we examine the validity of these additional assumptions for the first time using newly-released individual-level data on in-utero exposure to influenza from the 1958 National Child Development Study (NCDS). The NCDS tracks a large cohort of British children who were all exposed to the 1957 Asian Flu pandemic.

Perhaps surprisingly, we find that increases in area-level pandemic severity, proxied by pneumonia prevalence in the local authority of birth, fails to predict increases in the probability of in-utero exposure to influenza at an individual-level, providing evidence against the relevance assumption. The validity of the excludability restriction is also called into question, with theoretical arguments and evidence from literature suggesting that pandemic severity may have direct influences on many later-life outcomes. Results suggest that studies should interpret estimates from only reduced-form regressions relying on geographical variation in pandemic severity with caution.