Posts by Collection

Disparities between Black and White Americans in health care coverage and health outcomes are pervasive in the United States. In this paper, we describe the evolution of the market-based approach to health insurance and health care delivery in the United States and its implications for racial disparities. First, we discuss the history of the United States’ predominantly private health insurance system. Second, we illustrate the persistence and pervasiveness of disparities through three present-day epidemics: maternal mortality, opioid use, and Covid-19. Through the epidemic case studies, we highlight the systemic roots of racial inequality in health care. Finally, we conclude with a brief discussion of potential policy approaches for reducing disparities in the health care system.

Racial disparities in access to medical care are pervasive in the United States. We combine a simple model of healthcare utilization with empirical methods for estimating causal place effects to study the role of geography in driving these disparities. First, we show that the national Black-white access disparity in a given year can be decomposed into person and place components. We present two such decompositions, one where we assume that place effects are homogeneous and one where we allow for race-specific place effects (place-by-race effects). We then estimate these two decompositions using Medicare claims data from 2008-2018 and a mover design that leverages beneficiary migration across areas to estimate causal place effects. When place effects are assumed to be homogeneous, place matters very little for disparities. However, when place effects can vary by race, place matters enormously. Crucially, this place-by-race component is driven almost entirely by the fact that Black and white beneficiaries face very different, largely uncorrelated place effects in a given area, not because of differences in geographic sorting by race. Using a series of empirical exercises, we demonstrate the importance of these different place effects for access to medical care and the potential of different classes of policies to close disparities. We also show that our results are not driven by differential noise in our Black place effect estimates and hold for various levels of geographic granularity. Ultimately, our results suggest that while place-based policies are unlikely to close access disparities, more-targeted place-by-race-based policies are a promising path toward improving racial equity in utilization of healthcare services.

Rising life expectancy in the U.S. has focused considerable attention on the possibility of widespread increases in the population with Alzheimer’s disease and related dementias (ADRD). At the same time, however, the age-sex-adjusted prevalence of ADRD fell by more than 30% from 2000-2016. These patterns have competing implications for the evolution of years lived with dementia. In this paper, we combine data on dementia prevalence from the Health and Retirement Study (HRS) with life table data from the National Center for Health Statistics (NCHS) to estimate changes in dementia and dementia-free life expectancy at 65. From 2000-2016, total life expectancy at age 65 for the average American increased by approximately 1.6 years. Nearly all of this increase was driven by increases in dementia-free life years; we find substantively small and statistically insignificant changes in life expectancy with dementia at 65 since 2000. Both patterns persist across all racial and ethnic subgroups that we study. Our results indicate that as Americans are living longer, the additional years are cognitively healthy.

New medical technologies are increasingly expensive. These high-cost innovations make generous health insurance coverage more valuable for individuals at risk of needing new therapies. However, if those individuals are also costlier to insure, innovation may generate adverse selection. I develop a conceptual framework to study this trade-off and examine it empirically using data from Medicare Part D, the prescription drug insurance program for the elderly. I first show that an innovation shock driven by high-cost new drug approvals in the mid-2010s generated substantial adverse selection against Part D plans with generous coverage for those drugs, increasing those plans’ average costs by 35%. In the years following the shock, the market exhibits hallmark patterns of dynamic adverse selection: switchers into generous coverage are high-cost and more likely to use the new drugs; premiums rise by 52%; and price sensitive low-cost enrollees switch out of generous plans. Ultimately, the market significantly unravels, as the market share for the generous plans falls by 49%. Using a structural model of plan choice, I show that this unraveling leads to inefficiently low equilibrium enrollment in the generous plans and raises prices for those who remain enrolled, substantially reducing the insurance value of generous coverage and decreasing ex-ante social surplus. More robust reinsurance and risk adjustment policies would limit the losses from selection.

The age-sex adjusted prevalence of dementia in the United States has declined by more than 30% since 2000. However, the underlying sources of this cognitive improvement remain unclear, in part because the etiology of cognitive decline is not well understood. In this paper, we study these patterns, focusing on two commonly hypothesized contributors to cognitive function: air pollution and education. Using the Health and Retirement Study (HRS) and quasi-experimental designs, we assess the importance of air pollution and education for improvements in cognitive function since 2000. To form exogenous measures of exposure to small particulate matter (PM2.5), we leverage plausibly exogenous changes in power plant emissions. We instrument for educational attainment by digitizing historical data on the timing of university openings across states. We find that both pollution and education affect cognition. A 1 standard deviation increase in same-year PM2.5 increases the probability of dementia by approximately 0.05 standard deviations (17% of the sample mean). An additional year of education decreases the probability of dementia by approximately 0.1 standard deviations (26% of the sample mean). These effects grow monotonically with age but show little systematic pattern by individuals’ genetic predisposition for Alzheimer’s disease. All else equal, we find that dementia rates are 25% lower than they would have been without these changes.