When Industry Growth Becomes Geographic Concentration
The U.S. hospital establishment is not only a national workforce; it is a set of places distributed across counties. The question here is what happened to that local distribution while employment in included hospitals was rising. Reading county presence alongside the jobs total helps readers judge concentration and potential access loss without treating a county label as a direct measure of any patient’s care.
A county can contain several included hospitals, one, or none. Those three states create a simple map of local hospital structure. They also expose what a national average hides: employment can expand in the industry while the number of counties with multiple nearby establishments shrinks.
Three county states describe different local markets
The analysis places counties into three categories: counties with no hospital meeting the study’s employment threshold, counties with one such hospital, and counties with two or more. The categories are about establishments, not hospital systems. They also describe county shares, not the number of residents in each category.
That distinction is important because the categories imply different kinds of local dependence. A county with two or more included hospitals has at least the possibility of local choice or competition. A county with one has a single included establishment as its local institutional presence, even if other facilities are available elsewhere. A county with none has no included hospital within its boundary, so distance and neighboring markets become central to the access question.
The categories do not tell us whether the hospitals are equal in size or capability. Two establishments might offer different services, and a single large hospital might employ more people than several smaller ones. The classification is therefore a structural starting point: it says how many included locations are present, not what those locations can do.
The county distribution moved toward fewer multi-hospital places
In 1987–1997, 20.3% of U.S. counties had no included hospital, 46.1% had one, and 33.6% had two or more. In 2009–2019, the corresponding shares were 23.9%, 49.8%, and 26.4%.
The change in the two-or-more category is the clearest signal. Its share fell by 7.2 percentage points. Over the same comparison, the no-hospital share rose by 3.6 points, while the one-hospital share increased from 46.1% to 49.8%. The country therefore moved toward a landscape in which a larger share of counties had either one included hospital or none, and a smaller share had multiple included establishments.
This is not just another way of stating that the national establishment count declined. A national count could fall because a few locations disappeared while the overall distribution remained stable. The county pattern suggests a change in local market structure as well. Some counties lost their final included hospital; others appear to have moved away from a multi-hospital arrangement. Those changes carry different immediate implications, but both point to less local plurality than before.
The evidence is consistent with consolidation or reduced local competition. It does not identify which process produced each county’s change. A hospital may have closed, merged, changed classification, or been joined by growth elsewhere. The county categories show the resulting distribution, not a transaction history. They are strongest as evidence that the geography of establishments changed and weaker as evidence about the exact mechanism behind every change.
The diagram separates a count from the questions that count prompts. Two or more establishments may create room for local choice, but the data do not measure effective competition. One establishment may be a robust regional center or a fragile institution. No establishment may mean a substantial access burden in one place and a manageable cross-county commute in another.
Why fewer multi-hospital counties matter
A decline in counties with two or more included hospitals changes the baseline from which local decisions are made. When several hospitals operate in the same county, patients may have alternatives, and institutions may compete or coordinate within a denser care market. If one closes, the county can move into the one-hospital category even though it still has a local facility. The practical consequence may be less choice, greater dependence on one employer, or more exposure to the financial condition of that remaining hospital.
That is a different concern from the loss of the final hospital. In a zero-hospital county, the question is no longer only whether local choice narrowed. It is whether residents must travel to another county for services that were previously available within their own. The county measure cannot determine the size of that burden, but it identifies the institutional transition that makes the burden possible.
It is also possible for concentration to bring benefits inside the remaining institutions. A larger or more centralized hospital may support specialized staff, equipment, or services that smaller sites could not maintain. This is why the county result should not be read as proof that every move from two hospitals to one is harmful. The point is more precise: the distribution of establishments changed, and the trade-off between scale and local plurality deserves separate evaluation.
The national employment increase makes this distinction especially important. Suppose expanding hospitals in a few markets add enough employees to raise the national total while several counties lose a location. The workforce statistic records growth; the county distribution records a change in where care institutions are present. Neither statistic cancels the other. They describe different consequences of the same reorganization.
Consider the two transitions hidden inside the county categories. Moving from two or more hospitals to one removes local plurality but leaves a facility inside the county. That may reduce choice, make the remaining employer more important to the local economy, and increase the consequences of a future financial shock at that institution. Moving from one hospital to none removes the county’s included establishment altogether. It shifts more services into neighboring markets and makes the distance to the next hospital part of ordinary access, not merely an issue of competition.
These transitions need not have the same cause. A multi-hospital county might lose one site because activity is centralized in a larger neighboring facility, because an establishment closes, or because its classification changes. A zero-hospital county might be close to a dense regional network or separated from alternatives by long roads. The table cannot tell us which story applies to a particular county. It tells us that the number of counties in each structural position changed, which is enough to make location a necessary part of the analysis.
The word “competition” also needs a narrow meaning here. Two or more hospitals in a county create the possibility of local alternatives, but the data do not report market shares, prices, referral relationships, ownership control, or whether the establishments offer substitutable services. A county can contain multiple hospitals that specialize in different forms of care, just as one large hospital can face competition from facilities across a county line. The falling two-or-more share is therefore a signal of reduced local plurality, not a measured estimate of market power.
The county lens is valuable partly because establishment-level data can preserve geographic detail that broader public statistics may suppress in sparsely populated areas or narrow industries. That detail helps reveal changes that a national total would conceal. But precision about location does not remove the need for precision about meaning. A well-located establishment count can still be an imperfect proxy for the services people need, the time it takes to reach them, or the resilience of the surrounding care network.
A county without a hospital is not the same as no access
The phrase “no hospital” can invite an exaggerated conclusion. It means no hospital meeting the study’s definition was located in that county during the relevant classification period. It does not mean that every resident was without medical care, that no clinician practiced there, or that the nearest hospital was unreachable.
The paper gives a metropolitan example: a county in or around the New York City area might lose a hospital while remaining part of a larger labor market with nearby facilities. Residents could still commute across county lines. In such a setting, the county boundary is a poor proxy for the actual care market. A zero in the county column can indicate institutional loss without implying a complete loss of practical access.
Rural geography changes the likely meaning. Rural counties tend to be more sparsely populated and farther from alternative hospitals. When the last local hospital closes, the next facility may be substantially farther away, and transportation options may be limited. The same binary change—one establishment to none—can therefore carry a larger access burden in a rural setting than in a dense metropolitan network.
“Likely” is doing important work here. The source argues that rural loss might be more consequential because of distance and sparse population, but the county data do not calculate travel times, emergency response times, transportation access, or the quality and service mix of nearby facilities. The pattern supports a reason for concern, not a universal estimate of harm.
The denominator changes what the percentages mean
A share of counties is not a share of people. Counties vary enormously in population, land area, density, and relationship to neighboring markets. If a small number of sparsely populated counties move into the no-hospital category, the percentage of affected counties may rise substantially without implying that the same percentage of Americans lost local hospital access. Conversely, a single densely populated county without an included hospital could contain many more people than several rural counties combined.
This does not make the county measure unhelpful. It answers a place-based question that a population total cannot: how widely distributed are hospital establishments across the political and geographic units where communities organize services? But it should be paired with population-weighted measures and distance-sensitive measures before anyone estimates how many people are affected.
The denominator also matters for the one-hospital category. Nearly half of counties were in that category in both broad periods, but “one hospital” is not a stable level of capacity. One county’s hospital may be large and comprehensive; another’s may have a narrow service range or be financially strained. The category shows local singularity, not adequacy.
A careful interpretation therefore uses three layers. The first is presence: is an included establishment inside the county? The second is structure: are there multiple local establishments or only one? The third is access: what can residents reach, how quickly, and with what services? The study supplies strong evidence for the first two layers and flags the third as an important but incompletely measured consequence.
What the county signal makes possible
The county distribution changes the central question from “Did hospital employment grow?” to “Where did the growth occur, and what happened to local institutional choice?” It shows why an industry can be larger in employment terms while becoming more concentrated geographically. It also identifies the comparisons needed to interpret the pattern responsibly: region, rurality, neighboring facilities, service capacity, and the economic role of the local hospital.
The signal is not a diagnosis of market power, quality, or patient outcomes. To establish those claims, we would need evidence on ownership and competition, travel and emergency access, services available at remaining hospitals, population exposure, and what happens after a closure. The source does not supply all of those measures, and absence of those measurements is not evidence that the effects do not exist.
What it does establish is enough to move beyond the national paradox. Fewer included establishments were spread across a county landscape with fewer multi-hospital counties and more counties in the one- or zero-hospital categories. That is geographic concentration, even though its practical meaning varies by place. The next comparison must therefore ask whether the pattern is distributed evenly or whether some regions and types of counties bear more of it than others.