Research and Publications

Pharmacy benefit managers (PBMs) influence where Medicare Part D beneficiaries fill prescriptions through both preferred pharmacy network design and differences in out-of-pocket costs across pharmacies. This paper studies whether these two instruments are strategically interdependent. I examine how local copay differentials change when a nonpreferred pharmacy that may constrain a PBM’s network strategy closes. Using Medicare Part D Prescription Drug Event data from 2019–2022 linked to pharmacy ownership, network status, and closure information, I construct local markets around nonpreferred pharmacies and compare affected plan–pharmacy relationships with comparable relationships that have not yet experienced a closure. The empirical design exploits variation across plans within the same geographic market and pharmacy organization while controlling for pre-event pharmacy importance and plan dependence. I find that 30-day copay differentials increase by around 50% following the loss of a threatening nonpreferred pharmacy, with the response evolving over time. The results show that network structure and cost sharing are jointly determined tools of pharmacy steering, implying that policies affecting one PBM contracting margin can induce adjustment along another.

We study how confidential rebates affect producer surplus across the U.S. pharmaceutical supply chain. Using the July 11, 2019, rescission of a proposed rebate pass-through rule as an unanticipated shock, we estimate how investors value rebate confidentiality. The rescission increased market capitalization for PBMs with rebate aggregators by 6.86% and insurers without PBMs by 6.09%, while reducing market capitalization for brand-name manufacturers by 2.81%. Our estimates imply that PBMs retain only 1.9% to 7.4% of rebates, but that share represents 26% to 62% of PBM profits. Results are consistent with confidential rebates redistributing surplus from manufacturers to intermediaries without creating value and potentially leading to 3.7 fewer drug approvals annually.

This paper examines whether the relationship between unemployment and mortality varies over the business cycle and with the strength of social welfare systems, using data from 36 OECD countries from 1979–2016. I distinguish short- and long-term unemployment and estimate dynamic panel models using fixed effects, Difference GMM, System GMM, and bias-corrected LSDV. The results suggest that social welfare spending attenuates the association between unemployment fluctuations and mortality. A 1% increase in the expenditure-to-GDP ratio for incapacity and health programs reduces the mortality response associated with a 1% decline in unemployment by approximately 0.14% and 0.04%, respectively. The attenuation is particularly pronounced for long-term unemployment and becomes roughly four times larger in the long run.