Does Hospital Capacity Strain Lead to Racial Disparities in Health Care?
Singh and Venkataramani (2026) ask whether growing hospital resource scarcity increases racial disparities in patient care and mortality. They analyze 107,224 inpatient admissions from two large teaching hospitals in the southeastern United States between 2015 and 2018, using hourly hospital occupancy to measure capacity strain. They find that as hospitals become crowded, mortality rises for Black patients but not White patients. At the highest strain level, Black patients were 0.7 percentage points more likely to die than White patients, compared with a 0.2-point gap at lower strain. Black patients also waited longer for beds and received less provider effort.
Why This Article Was Selected for The Policy Scientist
This article addresses an important policy question: whether resource scarcity can widen racial disparities in the delivery of essential services. In health care, the consequences are especially significant because unequal treatment under capacity pressure can affect waiting times, clinical attention, and mortality. The topic is timely as hospitals continue to face staffing shortages, high occupancy, and other capacity constraints. The study contributes to the literature by showing that racial disparities may become larger when institutional resources are strained, linking research on discrimination with research on scarcity. The detailed records covering more than 107,000 admissions are a major strength. The authors also use extensive fixed effects and plausibly exogenous variation in hospital strain, producing stronger causal evidence than standard multivariate regression. However, because the analysis covers only two hospitals, replication across health systems and regions is important for assessing generalizability.
Full Citation and Link to Article
Singh, M., & Venkataramani, A. (2026). Rationing by race. American Economic Review, 116(9), 3552–3585. American Economic Association.
Central Research Question
The study asks whether increasing scarcity of hospital resources causes care to be rationed differently by race, and whether those differences affect patient outcomes. The authors focus on a central problem in the economics of scarcity: when resources become constrained, institutions may not allocate them solely according to need. Instead, group identity may begin to influence who receives timely attention, access to beds, and provider effort.
The analysis examines whether hospital capacity strain widens racial disparities between Black and White patients. The authors are especially interested in whether mortality, waiting times, and provider effort diverge as hospitals become increasingly crowded. Their conceptual argument is that resource constraints can amplify existing forms of discrimination by reducing the time, attention, and flexibility available to clinicians. Under low strain, providers may have enough capacity to evaluate patients carefully. Under high strain, however, implicit bias, reliance on heuristics, or institutional practices may exert greater influence over decisions.
The article therefore links two broad literatures that are often studied separately: the economics of scarcity and the economics of discrimination. Its primary contribution is to show that scarcity itself may change the extent to which race affects resource allocation.
Previous Literature
The study builds on longstanding economic research concerning discrimination, scarcity, and resource allocation. Arrow (1998) and Darity and Mason (1998) provide foundational frameworks for understanding racial discrimination in economic settings. The authors also draw on research showing that discriminatory behavior can increase when people compete over scarce resources. Experimental work by Krosch, Tyler, and Amodio (2017), Berkebile-Weinberg, Krosch, and Amodio (2022), Antunes et al. (2023), and Haushofer et al. (2023) suggests that scarcity can intensify discriminatory beliefs or behavior.
Within health care, the study builds on extensive evidence of racial differences in diagnosis and treatment. The Institute of Medicine (2003) documented widespread racial and ethnic disparities in medical care. Schulman et al. (1999) provided influential evidence that patient race can affect physicians’ treatment recommendations. More recent research has identified additional mechanisms, including algorithmic bias, provider-patient social distance, and structural differences in access to care.
Obermeyer et al. (2019), for example, demonstrated that a widely used health care algorithm underestimated the medical needs of Black patients because it relied on health expenditures as a proxy for illness. Alsan, Garrick, and Graziani (2019) showed that racial concordance between physicians and patients can affect medical utilization and preventive care. Other studies have documented racial differences in waiting times, clinical documentation, and assessments of pain or illness severity.
The article also draws on economic research concerning rationing under scarcity. Earlier studies generally examine allocation through prices, willingness to pay, queues, credit constraints, or other economic mechanisms. The present study extends that literature by examining whether social identity can function as an additional criterion for rationing when resources become scarce.
Data
The authors analyze detailed electronic medical records from 107,224 inpatient admissions at two large teaching hospitals in a southeastern U.S. metropolitan area. The data cover June 2015 through June 2017 at one hospital and June 2015 through June 2018 at the second. Both hospitals provide medical, surgical, intensive care, and Level I trauma services.
The data are unusually detailed because they contain timestamps precise to the second. These records allow the researchers to reconstruct each patient’s movement through the hospital, beginning with arrival and continuing through discharge. The data include patient demographics, insurance status, diagnoses, procedures, attending physicians, vital signs, admission pathways, discharge outcomes, and unstructured clinical notes.
Black patients constitute approximately 60 percent of admissions in the sample. The analysis focuses on non-Hispanic Black and non-Hispanic White patients because there were too few patients from other racial and ethnic groups to support sufficiently precise statistical estimates.
The central measure of resource scarcity is hospital capacity strain. The authors calculate the total number of hospitalized patients during every hour of the study period and rank each hour within each hospital’s historical occupancy distribution. These rankings are divided into ten deciles. The highest decile represents the most strained periods, when approximately 91 to 95 percent of beds are occupied.
The authors also measure medical need using the Elixhauser mortality index, which predicts mortality risk from patients’ underlying medical conditions.
Methods
The primary empirical strategy examines whether racial differences in mortality change as hospital capacity strain increases. The authors estimate regression models interacting patient race with the hospital’s occupancy level at the time the patient arrives.
Their main specification compares mortality during the highest decile of hospital strain with mortality during the lower nine deciles. The coefficient on the interaction between Black race and high strain identifies whether mortality changes more for Black patients than for White patients when hospitals are operating near capacity.
The statistical models incorporate an extensive set of controls. These include age, sex, insurance coverage, measures of comorbidity and medical need, physician fixed effects, and almost 10,000 time-related fixed effects. The time controls account for hospital, year, month, day of the week, hour of the day, and major holidays. This allows the researchers to compare patients who arrive under unusually different levels of capacity strain while holding typical temporal patterns constant.
The authors perform numerous robustness tests. They estimate alternative definitions of high strain, add diagnosis-specific controls, examine abnormal vital signs, use alternative functional forms, and estimate models with hospital-by-date-by-hour fixed effects.
They also conduct extensive tests for patient selection. These analyses examine whether the characteristics of Black and White patients change differently as hospital strain rises. The evidence does not indicate that the mortality findings are driven by increasingly sick Black patients being admitted during crowded periods or by selective discharge patterns.
Finally, the researchers investigate mechanisms using two measures of resource allocation: waiting time for an inpatient bed and provider effort. Provider effort is approximated from the content of clinical notes using measures such as note length, word length, subjectivity, polarity, and adjective use.
Findings/Size Effects
The principal finding is that racial disparities in mortality increase substantially when hospitals become highly strained. At lower levels of strain, Black patients are approximately 0.2 percentage points more likely to die in the hospital than White patients. When hospitals reach the highest decile of capacity strain, the mortality gap rises to approximately 0.7 percentage points.
The estimated increase in the Black-White mortality gap associated with moving into the highest level of hospital strain is approximately 0.53 percentage points. Given an overall mortality rate of about 1.6 percent, this represents a substantial change. In practical terms, the racial mortality gap more than triples under severe capacity strain.
The increase is concentrated among patients with more complex medical needs. For patients with relatively straightforward conditions or highly protocolized treatment pathways, the racial difference in mortality does not widen substantially with strain. This pattern is consistent with the idea that disparities become larger when clinical decisions require greater judgment, attention, and diagnostic effort.
Waiting times provide additional evidence of unequal resource allocation. Black patients wait longer for inpatient beds than White patients at every level of capacity strain. At lower strain, Black patients wait approximately 0.95 hours longer. At the highest strain level, the difference increases to about 1.3 hours, representing a 37 percent increase in the racial waiting-time disparity.
Medical need also becomes less predictive of bed allocation as strain increases. At nearly every capacity level, high-need Black patients wait longer for beds than low-need White patients. At the highest levels of strain, differences based on medical need largely disappear while racial differences remain.
The provider-effort analysis produces a similar pattern. At lower levels of strain, the estimated provider-effort index is approximately 0.10 units lower for Black patients than for White patients. At the highest strain level, the difference increases to roughly 0.21 units. Thus, the disparity approximately doubles when hospitals are most crowded.
The researchers find no corresponding evidence that changing patient composition explains these results. Patient characteristics, admission pathways, predicted mortality, presenting complaints, and discharge patterns do not change differentially by race as hospital strain increases.
Conclusion
The study concludes that resource scarcity can alter not only how much care patients receive but also how that care is distributed across racial groups. When hospitals approach capacity, racial differences emerge more strongly in mortality, waiting times, and provider effort.
The results suggest that discrimination may be conditional rather than constant. Racial disparities may remain relatively modest when institutions have sufficient resources but become substantially larger when time, personnel, beds, and provider attention are constrained. Scarcity therefore appears to interact with existing institutional and behavioral processes rather than simply reducing care uniformly for all patients.
The article also broadens the economic understanding of rationing. Traditional models generally emphasize prices, queues, willingness to pay, or formal measures of need. These findings indicate that group identity can also influence allocation when resources become scarce.
Because the analysis is based on two hospitals within a single health system, the extent to which the results generalize to other hospitals and regions remains uncertain. Nevertheless, the highly detailed data, extensive robustness tests, and use of plausibly exogenous hour-to-hour variation in capacity strain provide substantial evidence that hospital crowding can magnify racial disparities in care and patient outcomes.
Comments