What determines human health is known only in part. Even where causal pathways have been identified, great uncertainty frequently exists regarding the strength of the relationship. Compelling and precise estimates of these relationships are needed to inform public policies that will improve population health. This project would improve understanding of these relationships, using natural and quasi-experiments. The project highlights projects on the use of Ramadan to study the effects of maternal nutrition, the effects of heating subsidies in China to study the effects of pollution on prenatal development, and the impact of obesity report cards on obesity and health in Arkansas.
This CAREER proposal argues that refutable research hypotheses can be evaluated in an observational setting where identifying variation is plausibly exogenous. Ideally, the research hypothesis has strong implications which can be evaluated in available (i.e. pre-existing) data. The most compelling empirical comparisons do not require sophisticated econometric fixes. Instead, good natural experiments can identify treatment effects with minimal statistical adjustment. The cleanest of experiments, where the treatment is randomly assigned, requires only the comparison of means to estimate causal effects. This controlled experiment should be emulated wherever possible. Operationally, the proposed approach seeks to unearth comparisons in an observational setting where unadjusted impact estimates are quite similar to regression-adjusted impact estimates.
