We study a 2008 policy reform in which Medicare revised its hospital payment system to better reflect patients’ severity of illness. We construct a simulated instrument that predicts a hospital’s policy-induced change in reimbursement using pre-reform patients and postreform rules. The reform led to large persistent changes in Medicare payment rates across hospitals. Hospitals that faced larger gains in Medicare reimbursement increased the volume of Medicare patients they treated. The estimates imply a volume elasticity of 1.2. To accommodate greater volume, hospitals increased nurse employment, but also lowered length of stay.

We develop a framework to measure the role of hospital allocation in racial disparities in health care and use it to study Black and white heart attack patients. Black patients receive care at lower-performing hospitals than white patients. However, over two decades, the performance gap between hospitals treating Black and white patients shrank by over two-thirds. This progress is due to more rapid performance improvement at hospitals that tended to treat Black patients rather than reallocation of patients. Hospital improvement is correlated with adoption of a productivity-raising input, beta blockers. Our work highlights reallocation and performance improvement as future disparity-reduction levers.

Despite the continuing US hospital merger wave, it remains unclear how mergers change, or fail to change, hospital behavior and performance. We open the “black box” of hospital practices through a mega-merger between two for-profit chains. Benchmarking the merger's effects against the acquirer's stated aims, we show they achieved some of their goals, harmonizing electronic medical records and sending managers to target hospitals. Post-acquisition managerial processes were similar across the merged chain. However, these interventions failed to drive detectable gains in performance. Our findings demonstrate the importance of organizations for merger research in health care and the economy more generally.