Research

This paper examines how the complexity of medical discharge instructions influences patient mortality. Using data on 239,878 hospital admissions from a large U.S. academic medical center, I measure textual complexity using standard readability metrics. A one-standard-deviation increase in complexity is associated with a 0.14 percentage point rise in 28-day mortality, representing an 8.7% increase relative to the baseline rate. Associations are highly heterogeneous and concentrated among patients requiring intensive self-management: for heart failure patients, the association is nearly ten times larger. The pattern persists within subsamples of notes containing identical self-care instructions, demonstrating that linguistic framing matters beyond task assignment. The magnitude of the mortality association is comparable to estimates from Medicare eligibility or increased hospital spending, suggesting that interventions to simplify discharge instructions could yield substantial health benefits at low cost.

We propose a test to evaluate whether the causal effect of a conditionally randomly assigned treatment is fully mediated by observed intermediate outcomes, and whether the various causal mechanisms operating through different mediators are identifiable conditional on covariates. We show that if both conditions hold, the treatment is conditionally independent of the outcome given the mediators and covariates. We extend the framework to non-random treatments, where full mediation remains testable but identification of mechanisms is no longer guaranteed. We propose a double machine learning implementation that accommodates high-dimensional covariates and is root-n consistent and asymptotically normal, and present simulations and two empirical applications on maternal mental health and social norms.

Hosting mega-sports events generates optimistic projections of economic benefits, yet empirical evidence on actual local returns remains mixed. Focusing on the UEFA EURO 2024 in Germany, this paper provides causal evidence on the short-term local consumption effects of hosting mega-events. We leverage anonymized daily card spending data at the postcode level to measure changes in consumer spending. Using a difference-in-differences and local projection framework, we document a statistically significant and economically meaningful 3% increase in consumer spending in host cities during the tournament. This effect is driven entirely by international visitors, whose spending increases by more than 6%, and is concentrated in the group stage and on match days. Domestic spending does not change on aggregate, but exhibits spatial displacement. The gains are highly concentrated in city centers and tourist-facing sectors.