Beyond “care deserts”: a new metric for mapping formal care employment equity in the United States

Abstract

The formal care economy, including paid labor in health care, education, and daily living support, is essential to individual well-being and national productivity. Yet access to these services remains geographically unequal across the United States. Existing approaches commonly measure either within-area inequality (e.g., Gini coefficient) or regional specialization/adequacy (e.g., location quotient (LQ)), but rarely integrate both dimensions in a way that supports consistent classification and comparison across places, sectors, and time. Here, we develop the Care Resource Equity Score (CaRES), which integrates the Gini coefficient and the Location Quotient to quantify the joint distribution of care employment across internal regions and its regional concentration. Using U.S. census tract data aggregated to all counties from 2009–2021, we classify counties into four care-employment typologies using common thresholds: low inequality (Gini < 0.5) vs. high inequality (Gini > 0.5), and below-average concentration (LQ < 1.0) vs. above-average concentration (LQ > 1.0, relative to the national average). The resulting categories are Equitable Coverage (low Gini, high LQ), Concentrated Access (high Gini, high LQ), Even Desert (low Gini, low LQ), and Unequal Scarcity (high Gini, low LQ). Applying CaRES across health care, education, and daily living sectors, we found that most counties fall into either Unequal Scarcity or Concentrated Access, indicating that care employment is typically either under-supplied and unevenly distributed or abundant but spatially concentrated. Nonmetropolitan counties, especially small-town and rural areas, disproportionately exhibit Even Deserts, whereas metropolitan areas more often exhibit Concentrated Access, reflecting localized inaccessibility despite overall resource abundance. Over time, average within-county inequality declines, but this trend is often accompanied by a decline in relative care-employment concentration, suggesting that more even spatial distribution can coincide with growing scarcity. The resulting typology provides an interpretable, scalable tool for identifying distinct types of care-access challenges and for targeting policy responses that differ between places facing scarcity and those facing concentration.

Introduction

Across the globe, demographic, economic, and social transformations are reshaping how societies organize and provide care. Aging populations lead to increasing demand for health and eldercare (United Nations, 2024). Simultaneously, rising female labor force participation from 50% in 1980 to over 57% in the United States (U.S.) today has reduced the availability of unpaid caregivers within households (U.S. Bureau of Labor Statistics, 2023). The growing prevalence of dual-earner and single-parent households has further intensified reliance on formal care infrastructure for a wide range of care activities, ranging from cooking and cleaning to child and elder care (Folbre, 2006; Hartley et al., 2022). These shifts have made the formal care economy, defined as all economic activities related to providing care for another human for money, a focal point of both academic and policy concern (Berik & Kongar, 2021; M. L. Heggeness, 2023; Soldo et al., 2024).

The formal care economy encompasses paid labor that supports individuals’ health, development, and daily living needs. This sector includes not only teachers, nurses, and home health aides but also food service workers and childcare providers whose labor enables households and the broader workforce to function. Nearly 48 million adults in the U.S., approximately 18% of the labor force, are employed in care-related occupations, reflecting the sector’s increasing importance to economic and social well-being (M. Heggeness et al., 2025).

Despite its growing size and importance, the care economy remains under-recognized, undervalued, and unevenly accessible across communities (Blofield, 2018; Razavi, 2007). As society pivots to rely more heavily on paid care activities, important geographic questions need to be asked about where the economic resources dedicated to caregiving are located, focusing on whether they are spread out equally, giving many people easy access to them, or if they are geographically concentrated, creating potential accessibility concerns for certain groups. As part of this effort, scholars have often defined areas lacking sufficient services of any kind as “deserts” (Emch et al., 2017). However, the lack of a standardized definition for “care deserts” complicates efforts to identify and compare these areas systematically. In the absence of a universal methodology, scholars have employed a range of tools to delineate care-scarce regions(Erdenee et al., 2017; Li et al., 2023; Malik et al., 2018; Wittenauer et al., 2024). Among the most commonly used are the Gini coefficient, which measures internal inequality (Gini, 1921), and the Location Quotient (LQ), which captures the regional concentration of services (Isserman, 1977). Some studies also used geographic proximity, such as distance to the nearest facility, to assess access (Wittenauer et al., 2024). However, each approach has notable limitations. The Gini coefficient reflects disparities in distribution, but does not account for overall adequacy (Cowell, 2000). The LQ indicates whether services are concentrated or underrepresented, but fails to capture how equitably they are distributed within a region (Alam & Ria, 2018; Markusen, 2017). Distance-based measures, while intuitive, are difficult to generalize across large study areas and require context-specific definitions of acceptable travel distances, which may vary by service type, infrastructure, and population needs (Bitto et al., 2003a; Wittenauer et al., 2024).

Moreover, existing approaches often focus on either the adequacy or the equity of care resources, but rarely both, thereby constraining our ability to fully characterize and compare care deserts. The conceptual issue with analyzing these separately comes from the question of whether it is better to live in a resource-rich community that is highly unequal or a resource-poor community that is highly equal? The answer to this question is almost certain to depend on the levels of resource abundance and inequality within the community and how they relate. Studies that analyze these concepts separately are thus limited in their ability to directly compare the normative welfare of communities. Some prior studies have applied multiple indices to mitigate individual limitations (Alam & Ria, 2018; Xu et al., 2014), but no existing approach integrates these dimensions into a single, unified framework. Additionally, while a substantial body of research has applied these tools to identify underserved areas, such classifications may not always distinguish between different types of scarcity, whether due to a lack of overall resources or to their uneven distribution within communities. As a conceptual example, if a household faces significant difficulty in accessing child care, it could be because their community has fewer than needed childcare resources available or because the resources dedicated to serving their community are geographically concentrated within the community but not nearby to the person in need, or both. This distinction is important for designing effective interventions tailored to specific local needs and for researchers seeking to understand the distribution of economic care-giving resources. Additionally, few studies have examined how they evolve over time, particularly across multiple sectors. Most existing research remains cross-sectional, utilizing data available at the current time point, leaving important questions about the persistence, mitigation, or expansion of care scarcity unanswered.

This study adds to the existing literature and advances econometric research on existing data by introducing the Care Resource Equity Score (CaRES), a new framework that jointly captures the supply and spatial distribution of formal care employment as both a continuous measure and a typology. The CaRES provides a useful single metric for practitioners and researchers to understand the spatial distribution of caregiving resources throughout the country. Combining the Gini coefficient and LQ into a standardized metric, CaRES enables consistent comparisons across geographies, time, and sectors. Using data from 2009 to 2021 across all U.S. counties, we assess equity in three care domains: health care, education, and daily living, and identify persistent clusters of inequality using spatial analysis. By providing a standardized typology of care access, CaRES offers a new lens for understanding regional disparities in the care economy. In this study, we detail the development of the CaRES framework, analyze spatial and temporal trends in care provision, and identify regional clusters using Local Indicators of Spatial Association (LISA). We used CaRES to address the following research questions: (1) which U.S. counties face the greatest formal care-access challenges due to low supply, unequal distribution, or both; (2) how these patterns differ between urban and rural counties; (3) how care resource equity changes over time (2009–2021); and (4) whether inequities exhibit regional clustering.

Methods

This paper calculates yearly CaRES measurements at the county level for 2009–2021. We rely on two specific data measures at the census tract level to create this data: 1) employment data in the care economy, and 2) population data.

We used the Workplace Area Characteristics (WAC) dataset from the Longitudinal Employer-Household Dynamics Origin-Destination Employment Statistics program to obtain employment data developed by the U.S. Census (US Census Bureau Center for Economic Studies, 2025). This dataset utilizes both census and administrative data to provide yearly employment statistics at the census block level by North American Industry Classification System (NAICS) industry codes. The data itself comes from state unemployment insurance wage records, the Quarterly Census of Employment and Wages, federal employment data from the Office of Personnel Management, various Census Bureau surveys, and administrative data. Data from WAC covers most U.S. states from 2002 to 2021(US Census Bureau Center for Economic Studies, 2025). Some counties periodically drop out of the sample as states change their reporting, preventing the Census Bureau from providing the necessary data to compute the score. During the years of this study, data for Mississippi end in 2018, for Alaska in 2016, and for Massachusetts in 2011. We excluded the District of Columbia from the analysis because it comprises a single census tract, which is insufficient for calculating the Gini coefficient required by this study’s spatial inequality framework. We also exclude counties with population counts too small to be included in a single census tract, as these counties lack sufficient subunit variation to calculate the Gini coefficient. We provided the score only for counties with full data availability within a year (see Supplementary Table S2 for the number of counties included in each year).

We divided this dataset into three categories of the care economy based on industry codes. The WAC data utilizes 2-digit NAICS codes: Educational Services (61), Health Care and Social Assistance (62), and Accommodation and Food Services (72), which we label as Educational Services, Health Care, and Daily Living, respectively. These three groups represent the three distinct sectors of the Care Economy, as defined by the Care Board (M. Heggeness et al., 2025). Tract population denominators are derived from the American Community Survey (ACS) (United States Census Bureau, 2021) five-year estimates, which pool data across five consecutive years to improve reliability at small geographies such as census tracts. To align with the annual WAC employment data, we assigned tract populations by matching each analysis year t (2009–2021) to the ACS 5-year file ending in year t (e.g., 2005–2009 ACS for the 2009 population denominator; 2017–2021 ACS for 2021). We used ACS data from 2009 to 2021 at the tract level to improve reliability and to match the WAC data (US Census Bureau Center for Economic Studies, 2025). We computed separate CaRES for these three sectors using these datasets.

The urban-rural spectrum is associated with resource allocation and employment (Bitto et al., 2003b; Wittenauer et al., 2024). Therefore, we applied urban and rural classifications from the U.S. Department of Agriculture (USDA)’s 2013 Rural-Urban Continuum Codes (RUCC), which provide county-level categories based on metropolitan influence and population size (USDA, 2013). RUCC 2013 was selected to maintain consistency across our 2009–2021 study period and to match the geographic scale of Gini coefficient and LQ metrics. RUCC codes classify U.S. counties on a 9-point scale based on metropolitan influence and population size. A detailed description of the urban and rural classification scheme is provided in Table S1.

Gini coefficient

The Gini coefficient measures inequality in resource distribution, ranging from 0 (perfect equality) to 1 (maximum inequality) (Gini, 1921). Although often applied to income, we used it to assess the distribution of care-related employment across census tracts within each county. The normative assumption behind this score is that it is preferable for paid care work to be located proximally to where people live. The care economy is something many people require every day of their lives, whether it’s dropping their children off at school or picking up groceries, and excess travel time or lack of nearby availability can be a potential hindrance. The Gini coefficient measures how equally spaced the care economy jobs are compared to where the people live in a county. Higher values indicate care jobs are concentrated in a few tracts separated from where the population lives, while lower values suggest a more even overlap between the location of care jobs and the population. We calculated the annual Gini coefficient (2009–2021) for three sectors (Health Care, Educational Services, and Daily Living) using population-weighted tract-level employment data. See Supplementary Section 1 for calculation details.

Location quotients (LQs)

The Location Quotient (LQ) identifies whether a region is specialized in an industry relative to the national average (Isserman, 1977). This measure is based on the normative assumption that counties with more care resources are better off than those with fewer resources. We used LQ to determine if care-related jobs are disproportionately concentrated in certain counties. An LQ of 1.0 indicates average representation; values above 1.0 suggest specialization, while values below 1.0 indicate underrepresentation. In a society where care jobs are perfectly split relative to the population (i.e. a county that has 10% of the population has 10% of the care jobs), all counties would have a score of 1. This metric highlights local surpluses or shortages in care employment, supporting our analysis of resource equity (see Supplementary Section 2 for calculation details).

Gini coefficient and LQ typology

Although both the Gini coefficient and LQ measure aspects of community resources, they capture different dimensions. LQ reflects the concentration of care jobs relative to the national average, while the Gini coefficient measures how evenly those jobs are distributed within a county. High Gini values (above 0.5) indicate significant inequality (Conceição, 2019), whereas LQ values above 1 suggest a surplus of care jobs. Using thresholds of 0.5 for Gini and 1.0 for LQ, we developed a 2 × 2 typology to classify counties (Table 1). This framework identifies four categories: Equitable Coverage (high LQ, low Gini), where care jobs are both plentiful and accessible; Unequal Scarcity (low LQ, high Gini), where care jobs are few and unequally distributed; Even Desert (low LQ, low Gini), where jobs are scarce but access is equally limited; and Concentrated Access (high LQ, high Gini), where care jobs are concentrated in specific areas, limiting reach. This typology helps clarify the distinct policy challenges facing different types of communities.

Table 1.

Gini–Location Quotient (LQ) Typology.

Low Gini–High LQ	High Gini–High LQ
(Equitable Coverage)	(Concentrated Access)
Low Gini–Low LQ	High Gini–Low LQ
(Even Desert)	(Unequal Scarcity)

Care resource equity score (CaRES)

We created county-level estimates of the Care Resource Equity Score (CaRES) because the data we used at the census tract level provide the necessary sub-county variation to calculate measures such as the Gini coefficient. While census blocks offer finer geographic detail and would allow us to create CaRES at the census tract level, they are often too small to analyze resource distribution because of the ease of access across block boundaries. Census tracts, typically covering 8000 to 12,000 people, strike a balance between geographic resolution and data availability, making them a common proxy for neighborhoods in research on spatial and demographic inequality (Emch et al., 2017).

To evaluate spatial equity in care-related employment, we calculated CaRES separately for three sectors: health care, educational services, and daily living services. CaRES combines two standardized metrics. The Gini coefficient and the Location Quotient are used to place them on the same scale using z-scores. The score is calculated by subtracting standardized LQ from standardized Gini, so higher CaRES values indicate counties with both fewer care jobs and greater inequality in their distribution (see Supplementary Section 3 CaRES Calculation for details). Lower scores reflect more equitable, resource-rich environments. This continuous measure highlights counties facing care access challenges due to both scarcity and uneven distribution.

Spatial analysis

We used spatial analysis to examine clustering and outliers in care-related employment equity across U.S. counties from 2009 to 2021. A Queen contiguity spatial weights matrix was defined to identify neighboring counties. We calculated global Moran’s I to assess overall spatial autocorrelation in CaRES, and Local Moran’s I (LISA) to identify localized clusters (Anselin, 1995). LISA detects whether a county’s CaRES is significantly similar to its neighbors, categorizing results as High–High (HH), Low–Low (LL), High–Low (HL), or Low–High (LH). HH indicates clusters of neighboring counties with high CaRES consistent with widespread inequity and relatively scarce care employment (e.g., Unequal Scarcity–type conditions). LL indicates clusters of neighboring counties with low CaRES, consistent with more equitable coverage and relatively greater care-resource availability. HL and LH represent spatial mismatches (often termed spatial outliers): HL identifies counties with high CaRES surrounded by neighbors with low CaRES, whereas LH identifies counties with low CaRES surrounded by neighbors with high CaRES. Counties with non-significant Local Moran’s I show no detectable local spatial association. To assess trends, we aggregated annual LISA results for each care sector, scaled cluster counts for visual comparison, and excluded non-significant cases. We also analyzed clustering across the rural–urban spectrum by linking LISA results with RUCC codes and calculating proportions of cluster types per RUCC category, scaled for standardized comparison.

Results

Using the typology classifications discussed above, we calculate the GINI and LQ for each U.S. county based on available data and plot them on a map. Supplementary Table S3 reports the number of states assigned to each typology (by sector), while Supplementary Table S4 reports the distribution of typology classifications across all county observations in the study period. As shown in Fig.1, the bivariate maps illustrate distinct spatial typologies in care employment across U.S. counties. The four typologies are represented as follows: Even Desert (white), Concentrated Access (dark green), Unequal Scarcity (dark blue), and Equitable Coverage (light green), with intermediate shades indicating counties near typology boundaries.

Fig. 1.

Multivariate choropleth maps displaying the spatial distribution of Gini coefficient and location quotients (LQ) at the county level across the U.S. for three service sectors: a Health Care, b Educational Services, and c Daily Living Services. The color scheme represents the intersection of inequality (Gini coefficient) and sectoral concentration (LQ), with darker shades indicating higher values of both metrics.

We used Fig. 1 to evaluate whether there are distinct spatial patterns in the CaRES typology across the U.S. In the healthcare sector, 74.6% of New England counties are classified as Concentrated Access. Moving south along the east coast (Delaware through West Virginia down to Florida), the dominant typology shifts: 49. 9% of counties are classified as Unequal Scarcity, indicating high inequality and below-average care employment. In the educational services sector, patterns differ considerably from those in healthcare. New England again leans toward Concentrated Access (60. 7%), but the two largest states diverge sharply: 85. 5% of New York counties fall into Concentrated Access, whereas 58. 2% of Pennsylvania counties are classified as Unequal Scarcity. Florida shows a further departure from the regional pattern: although the South is mainly characterized by Concentrated Access for educational services (57. 0%), 53. 7% of Florida counties fall into Unequal Scarcity, indicating both high inequality and below-average educational employment concentration. These within-region differences represent the potential impact of differing state policies on the paid care environment. Many states in the Midwest (Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota, and South Dakota) stand out for educational services, with 43. 7% of counties are classified as Equitable Coverage, the highest share of any region/sector combination, suggesting a more balanced distribution of educational resources relative to local need. Finally, in the daily living sector, the Central North West exhibits a distinctive pattern, with 44.7% of counties are classified as Even Desert, indicating low inequality and low service concentration across these often low-population states. Florida, which showed high Unequal Scarcity in educational services, reverses course entirely in daily living: 62.7% of its counties fall within Concentrated Access, suggesting sector-specific variation in care employment patterns.

Figure 1 clearly demonstrates the conceptual advantage of the CaRES typology over a single Gini coefficient or LQ metric. In the educational services sector, counties in Florida, Texas, and New England all exhibit high Gini coefficients, signaling unequal resource distribution across the population. However, the accompanying LQ patterns vary significantly. In Texas, 59.0% of counties fall into Concentrated Access, indicating that despite high internal inequality, these counties have above-average educational employment concentrations. Florida tells a different story 53.7% of its counties fall into Unequal Scarcity for educational services, combining high inequality with low LQ. This pattern is sector-specific: for healthcare, Florida is divided between Unequal Scarcity (53.7%) and Concentrated Access (46.3%), and for daily living, 62.7% of Florida counties are in Concentrated Access. The difference between Concentrated Access and Unequal Scarcity has distinct policy implications. In Texas and New England, where counties show high internal inequality and above-average employment levels in educational services, increasing within-county allocations to underserved populations is the main strategy. In Florida’s educational services sector, counties face a dual challenge: high inequality and below-average employment concentration, suggesting that both expanding the total supply of educational service workers and targeting resources to currently underserved communities should be priorities.

Figure 2 illustrates the relationship between the Gini coefficient and the LQ for formal Care employment across U.S. states and counties from 2009 to 2021. Panels (a–c) plot state-level averages (one point per state; 2009–2021 mean) for health care, educational services, and daily living, while panels (d–f) plot the corresponding county-year observations pooled across 2009–2021. In all panels, the x-axis shows the within-county Gini coefficient (higher values indicate greater spatial concentration of jobs within the county), and the y-axis shows the location quotient (LQ; values above 1 indicate above-average concentration of care employment). The dashed horizontal line marks LQ = 1.0, and the dotted vertical line marks Gini = 0.5; together, they define the four typologies: Equitable Coverage, Concentrated Access, Even Desert, and Unequal Scarcity.

Fig. 2.

Gini–LQ scatterplots and typology composition by sector and urbanicity. a–c state-level averages (2009–2021) and d–f display county-level observations for Health Care, Educational Services, and Daily Living, with typology cutoffs indicated. g reports typology shares across RUCC categories (See Figs. S2–4 for enlarged (a–c)).

Figure 2a–c presents the state-level average typology composition (one point per state, averaged over 2009–2021) for Health Care, Educational Services, and Daily Living. At this scale, Unequal Scarcity is most common in Health Care (28 states; 56%) and Daily Living (29 states; 58%), whereas Concentrated Access dominates Educational Services (28 states; 56%). Equitable Coverage is comparatively rare across sectors (Health Care: 2 states; 4%; Daily Living: 2 states; 4%), though it appears more frequently in Educational Services (11 states; 22%) (Detailed information can be found in Supplementary Table S3).

Panels (d–f) show the corresponding county-level typology distribution and reveal substantial within-state heterogeneity. In Health Care (Fig. 2d), the largest share of counties is classified as Unequal Scarcity (8644; 43.7%), followed by Concentrated Access (6876; 34.8%), with smaller shares in Even Desert (2542; 12.8%) and Equitable Coverage (1724; 8.7%). In Educational Services (Fig. 2e), Concentrated Access is most prevalent (8495; 43.9%), followed by Unequal Scarcity (5662; 29.3%), Equitable Coverage (3747; 19.4%), and Even Desert (1438; 7.4%). In Daily Living (Fig. 2f), Unequal Scarcity again dominates (8690; 44.1%), with substantial shares of Concentrated Access (5716; 29.0%) and Even Desert (4133; 21.0%), and fewer counties in Equitable Coverage (1176; 6.0%). Panel (g) shows systematic variation by urbanicity: metropolitan counties more frequently fall into Concentrated Access and Unequal Scarcity, whereas rural categories have higher shares of Even Desert (and occasionally Equitable Coverage), consistent with relatively even but limited care employment in many rural settings (detailed information can be found in the supplementary Table S4).

Figure 2g illustrates the distribution of Gini–LQ typologies pooled across the three care sectors (Health Care, Educational Services, and Daily Living) across the urban–rural continuum (sector-specific temporal dynamics across urbanicity categories are shown in Supplementary Fig. S1). In metropolitan counties ( ≥ 1 M, 250K–1 M, and <250 K), the most common classifications are concentrated access and unequal scarcity, indicating a polarized distribution of care resources. In contrast, concentrated access dominates small town areas both adjacent and non-adjacent to metro regions. Notably, equitable coverage is most prevalent in rural areas, particularly in rural adjacent and rural remote counties, suggesting a more balanced distribution of care services in these regions.

Temporal change of Gini coefficient and LQ

Figure3 a-c presents temporal trends in the distribution and concentration of service sectors across U.S. counties from 2009 to 2021. Average Gini values decline across all three sectors, with the steepest drop occurring after 2019–2020, particularly in Daily Living (Fig. 3a). LQ patterns diverge by sector: Educational Services remains consistently above the national benchmark (LQ > 1), Health Care declines from near/above average to below average by 2021, and Daily Living rises in the final years and exceeds 1.0 by 2021 (Fig. 3b). CaRES values are relatively stable through most of the 2010s but shift sharply after 2019, most notably for Daily Living, indicating a substantial change in the combined inequality–concentration pattern in the final years of the study period (Fig. 3c).

Fig. 3.

Temporal trends in care-employment inequality, concentration, and CaRES (2009–2021). a–c annual national averages across all U.S. counties for the three care sectors (Daily Living, Educational Services, and Health Care): a average Gini coefficient (within-county inequality in the spatial distribution of care jobs), b average location quotient (LQ; relative concentration compared with the national average), and c average CaRES. d–f the corresponding state-specific time series (each small subplot represents one state, labeled by its abbreviation), calculated as annual averages across counties within each state for d Gini, e LQ, and f CaRES. Colored lines indicate sectors.

Figure 3d, e provides a state-level view of the temporal dynamics of LQ and the Gini coefficient. Figure 3d shows that most states exhibit declining Gini coefficients over time across all three service categories, reinforcing the national-level trend toward reduced spatial inequality in care-related employment. Figure 3e highlights substantial variation in LQ trends by state and service category. For example, Educational Services remain highly concentrated in states such as Massachusetts and New York, while Health Care displays more moderate and, in some cases, declining LQ trends across states. The daily living category shows increasing divergence among states in recent years, with several states experiencing sharp rises in LQ post-2020.

Figure 3f shows the state-level trends for the CaRES in its entirety. Because higher CaRES values reflect greater within-state inequality and a lower relative concentration of care employment, declining CaRES in many states suggests a shift toward a more even within-county distribution of care jobs and/or a higher relative concentration of care employment across the study period. Interestingly, this measure also shows a variety of differences across states. As an example, in some states, Educational Services is the most vulnerable category, while in others, it is the best. There appeared to be regional patterns in these scores across the three categories, and future research can examine how state and regional policy trends affect a state’s formal care economy. A striking feature of the CaRES is the widespread change in scores across numerous states in 2020, reflecting the magnitude of that year’s disruptions (e.g., COVID-19). Due to current data availability, our post-pandemic analysis is limited to 2021, preventing a full assessment of how states have recovered. Nevertheless, our results indicate an increased reliance on the formal care economy in many states, most notably in health care and daily living, where care employment grew as a share of states’ overall economic activity.

Local indicators of spatial association analysis

Based on the statistically significant Moran’s I results (see Supplementary Table S5), we conducted LISA analysis. As shown in Fig. 4, the LISA results illustrate the spatial distribution of clusters for the CaRES across three rolling five-year periods (2009–2013 to 2017–2021) within the three care economy sectors. Across time, a clear spatial pattern was found. High-High (HH) clusters, representing significant vulnerabilities in care employment, remain persistent in parts of the Southeast and other historically underserved regions, while Low-Low (LL) clusters, representing areas with adequate care employment, gradually expand across the Midwest, Northeast, and Western U.S. This shift points to a growing presence of spatial equity in certain regions, even as other areas remain concentrated in scarcity. These maps also show important differences by sector of care: the first row represents healthcare, the second represents educational services, and the third represents daily living.

Fig. 4. Local Indicators of Spatial Association (LISA) cluster maps for Care Resource Equity Score (CaRES) in U.S. counties.

Panels show spatial clustering of CaRES for three formal care sectors (a, d, g: row 1) health care, (b, e, h: row 2) educational services, and (c, f, i) Daily living—across three time periods: 2009–2013 (a–c: column 1), 2014–2018 (d–f: column 2), and 2019–2021 (g–i: column 3).

Figure 5a complements these spatial trends found in Fig. 4 by summarizing the temporal dynamics of LISA clusters from 2009 to 2021 across three service categories. Each tile in these figures represents approximately 10 counties classified into one of the four significant LISA categories. The chart reveals that while HH clusters (dark red) were dominant in early years, particularly in Health and Education, LL clusters (dark blue) steadily increased over time, especially in the Education and Daily Living sectors. Together, these figures illustrate the changes of the geography of care equity in the U.S., with some sectors and regions progressing toward greater balance while others remain entrenched in inequity.

Fig. 5. Local Indicators of Spatial Association (LISA) clusters by time and urbanicity from 2009 to 2021.

a (top rows) Temporal change in LISA clusters. Each tile represents approximately 10 U.S. counties within a given year and service category. b (bottom rows) LISA clusters by urbanicity level based on RUCC classification. Each tile represents ~1% of counties within a given RUCC group (Note: counties not significant at the 0.05 level are excluded).

Figure 5b illustrates the spatial clustering patterns of care-related employment equity across different levels of urbanicity, as defined by the RUCC. For each RUCC level, the total number of counties is scaled to 100 tiles to enable proportional comparison across service categories and cluster types. Thus, a tile represents 1% of the counties in that RUCC group.

A clear spatial trend was found across service categories: as urbanicity decreases, the prevalence of LL clusters increases relative to HH clusters. In the case of Educational Services, HH clusters dominate the most urbanized counties (Metro ≥1 M, Metro 250K–1 M, and Metro <250 K). However, a shift toward LL dominance becomes apparent starting with Urban Non-Adjacent counties. It continues across Small Town and Rural designations, particularly in Small Town Non-Adjacent and Rural Remote areas. Health Care showed a similar transition. While HH clusters remain prevalent in metro regions, LL clusters become more common beginning in Rural Adjacent counties and remain dominant across Rural Remote, Small Town Adjacent, and Small Town Non-Adjacent areas. For Daily living, HH clusters again dominate metropolitan and Urban Adjacent counties, but LL clusters exceeded HH clusters starting in Urban Non-Adjacent areas and strongly dominate in Small Town Non-Adjacent and Rural Remote regions. This trend indicates that the Formal Care Economy is a more significant share of the economies of non-urban areas and that these areas tend to have easier geographic access to the existing formal care infrastructure. It is important to remember that the CaRES is a relative measure and is able to account for within-unit population sizes, meaning that these Rural communities might have far fewer formal jobs than their urban counterparts, but see a higher share of their employment in these positions as well as a more equitable distribution among the population. We discuss why this might be the case in the discussion section.

Discussion

This study introduced the CaRES and care resource typology, a novel metric designed to evaluate the geographic distribution of formal care labor in the U.S. By combining the Gini coefficient and the LQ, CaRES captures both the intensity of care-related employment relative to population size and its spatial concentration across counties. This dual-lens approach provides a comprehensive framework for assessing not only whether care jobs are abundant or scarce, but also whether they are equitably distributed within counties. Unlike prior methodologies that rely on singular inequality or specialization metrics, CaRES enables systematic comparisons across time, place, and care sectors, and yields a typology that supports both academic inquiry and policy intervention. This contribution fills a critical gap in empirical research on the care economy, which often lacks interpretable, analytically comprehensive tools.

Our results revealed pronounced and persistent spatial inequities in the distribution of care-related employment. Across all three sectors examined, typologies of Unequal Scarcity and Concentrated Access dominate, reflecting structural imbalances in both job availability and spatial allocation. In the Health Care sector, over 78% of counties fall into one of these two categories, suggesting widespread issues with both access and equity. Educational Services showed a similarly skewed pattern, with nearly 73% of counties classified as either over-concentrated or insufficiently served. The Daily Living sector, encompassing support and ancillary care roles, exhibited the highest share of Unequal Scarcity counties (44.1%), indicating significant gaps in foundational care infrastructure, particularly in economically vulnerable regions.

Our analysis also indicated interesting temporal trends in these measures. Nationally, the Gini coefficient is decreasing, indicating better national distribution of care resources. At face value, this suggests a positive trend, indicating that care resources within a community are becoming more evenly distributed across the population. However, in many states and sectors, the LQ score is decreasing, meaning that while these communities might have more equal distribution of resources, they have fewer resources to distribute. This reflects an increasing pattern of scarcity, rather than equal access. This pattern is partially reflected in our LISA results: while LL clusters have grown over time in many regions, HH clusters remain persistent in historically underserved areas, particularly in the Southeast, suggesting that gains in spatial equity have not been uniform across the country. Additionally, given the relative nature of the LQ score, we see signs that much of the growth in the Care Economy is concentrated in specific geographies, leaving other areas behind. Nationally, this appears to lead to an improvement in the data; however, at subnational level, there is significant variation in outcomes. This trend aligns with prior literature showing a decrease in the supply of teachers and nurses in specific geographies, particularly rural areas (Ingersoll & Tran, 2023; Sablik, 2022; Sutcher et al., 2019; Terry et al., 2021).

Large shifts in CaRES correlate with major economic events, notably the 2008 recession and the COVID-19 pandemic. In 2020, CaRES rose sharply. One plausible explanation is that care jobs, widely deemed essential during the pandemic, represented a larger share of total employment as other sectors contracted, though this interpretation is observational and requires further investigation with post-2021 data. While this temporarily boosted LQ scores, it remains unclear whether this shift reflects lasting structural change or a pandemic-specific anomaly. Future data will help determine if CaRES returns to pre-pandemic levels or continues to rise.

Our results also provided valuable insights into the urban-rural spectrum. In more rural RUCC categories, especially remote rural and small-town counties, care employment was more often classified as Even Desert. One interpretation is that in many rural counties, care employment functions as an anchor sector (e.g., hospitals, long-term care facilities, school districts) that constitutes a relatively large share of local employment (raising LQ) while being less spatially segregated within the county (lower Gini), given smaller geographies and fewer strongly differentiated tracts (Smith, 2013). At the same time, rural counties often have fewer and less sharply differentiated subcounty employment centers, which can yield lower within-county inequality (Low Gini) relative to larger, more spatially segmented metropolitan counties. In contrast, metropolitan RUCC categories more frequently exhibited Concentrated Access, in which care employment is relatively abundant but spatially clustered within counties, thereby limiting access for many residents. Importantly, Equitable Coverage reflects relative concentration and internal distribution, not necessarily absolute service capacity; rural counties can appear “equitable” while still facing long travel times, limited specialty services, or workforce shortages. These findings echo prior studies on pharmacy, schools, and child care deserts and reflect broader patterns of suburbanization and infrastructure disparity (Malik et al., 2018; Sutcher et al., 2019; Terry et al., 2021; Wittenauer et al., 2024).

We believe that patterns of urban development may have influenced the spatial distribution of resources observed here. Prior research suggests that shifts in U.S. metropolitan structure, namely, the decentralization of employment toward suburban sub-centers have been associated with regional employment growth. Greenfield development in suburban or rural areas is often less constrained by zoning, cost, and density than infill construction in urban cores, which may partly explain why some rural areas appear more feasible for care facility expansion. These explanations are exploratory; this study does not directly test causal mechanisms (Medical Construction & Design, 2018). As a result, rural areas, while seemingly more isolated, may offer more feasible opportunities for care facility expansion than inner-city areas constrained by cost, space, or zoning. This discussion is echoed in recent research by Pandey et al. (2025), who demonstrate that urbanization and economic development are frequently accompanied by rising infrastructure inequalities, particularly in rapidly growing or fragmented urban regions. Their work suggests that growth alone does not guarantee equitable access to infrastructure, reinforcing the importance of spatial justice frameworks for evaluating care systems (Pandey et al., 2025).

These results have significant policy implications. Within a community, individuals may have difficulty accessing formal care resources. However, these results suggest that not all causes of this difficulty are the same. For instance, in an Even Desert, one potential strategy is expanding care resources in existing locations, as the primary challenge appears to be overall scarcity rather than distributional inequality. However, these policy directions are illustrative; the CaRES framework identifies spatial patterns but cannot establish causal mechanisms, and effective interventions will require local context and further analysis. For areas with Concentrated Access, simply increasing employment will not solve the problem. Concerned policymakers in these areas need to focus on investments to reach new communities. This could involve creating new infrastructure, such as building new schools, incentivizing businesses to open in underserved communities by providing subsidies to grocery stores (Bastian et al., 2017), or increasing resources in remote operations, including home health aides or telehealth services (Butzner & Cuffee, 2021). Future research and policy considerations should specifically focus on correlations between areas with a high CaRES and other measures of socio-economic vulnerability. Finally, areas in the unequal scarcity type should see the most attention as they need immediate action. Policymakers should prioritize these areas for improvement, as they are both under-resourced and unequal. For each of these reasons, it will likely be crucial to consult with locals and examine the local economy to understand why these problems exist and develop community-specific plans. In this paper, we hesitate to hypothesize about the cause of these differences, but they might be due to a wide range of factors, from regional economic development patterns to political preferences, as these vary across regions. Future work can analyze these differences to look for economic, cultural, political or institutional reasons for them. Future work can also use these responses, coupled with data from time-use surveys such as the American Time Use Survey, to examine how individuals’ daily caregiving patterns vary by community type.

Together, these patterns suggest that while some regions serve as hubs of care employment, large swaths of the country remain either underserved or unequally provisioned. The typological framework of CaRES thus offers insights into the geography of care inequality, identifying not only where disparities exist but also the spatial dimensions and patterns of those disparities. Future research on CaRES will involve continuing to produce these statistics for the coming years to understand how the Care Economy has recovered in the years following the pandemic. Research should also examine the application of this methodology to other care economy resources. This study has focused on employment as an important proxy for caregiving in the formal economy, but there are other opportunities as well. For example, this methodology could be used to understand the distribution of available childcare or eldercare slots in a society or the distribution of any other caregiving resource. The CaRES framework provides a method for comparing different resources.

While the CaRES framework offers numerous advantages, it also has limitations that should be considered in future research. First, data availability is not uniform across states and years in the LEHD WAC series, resulting in some counties being included in and excluded from the analytic sample (e.g., gaps or discontinuities for specific states and periods). For example, Mississippi and Alaska, which exit the sample in 2018 and 2016, respectively, are predominantly rural states; their exclusion from later years may cause post-dropout averages to underrepresent rural care-employment conditions, potentially biasing observed national trends. Second, the analysis relies on broad 2-digit NAICS codes to define care-related employment. Although these codes offer consistent national coverage, they may overstate care-economy employment because not all jobs within NAICS 62 (Health Care and Social Assistance) or NAICS 72 (Accommodation and Food Services) directly constitute caregiving or care support. This classification choice can inflate employment counts used to compute the Gini coefficient and LQ, and therefore the typology and CaRES, particularly in counties where these industries include substantial non-care or general service employment. Future work could refine sector definitions using more granular NAICS (3–6 digit), occupation-based measures, or crosswalks that isolate direct care functions. Third, we were unable to calculate CaRES for single-unit geographies, such as the District of Columbia, because the Gini coefficient requires variation across subunits. Fourth, a select number of counties, most notably in rural and frontier areas, had too small a population to obtain reliable estimates of the Gini coefficient and its distribution. Moreover, our analysis is based on administrative boundaries, which limit consideration of service flows across these boundaries. Counties also vary substantially in geographic size and population density, which can influence both the spatial analyses and interpretation of CaRES. In large, sparsely populated counties, care jobs may appear more “evenly” distributed within the unit while still implying long travel distances, whereas in small, dense counties, clustering can be more apparent even when access may be relatively high. For example, many large health care facilities, such as major hospitals, are located outside of a city boundary, which causes unequal distribution of employment. Future research can consider utilizing other boundaries, such as commuting zones, to analyze CaRES. Furthermore, the Gini coefficient in this framework assumes that a distribution of care employment proportional to the local population across census tracts is the normative ideal. However, for certain care services such as regional hospitals, specialized clinics, or higher education institutions geographic concentration may be appropriate and efficient. Future work should consider whether sector-specific normative benchmarks better capture access equity for different types of care. Finally, because the LQ is a relative measure, shifts in national industrial composition, such as the broad contraction of non-care sectors during the COVID-19 pandemic, can produce apparent changes in county LQ scores even when absolute care employment levels remain stable. Interpreting LQ trends as indicators of care-employment growth or decline should therefore be approached cautiously, particularly during periods of broad economic disruption, such as the 2008 recession and the 2020 pandemic.

Author contributions

YA and JB jointly developed the research design. YA led the data analysis and drafted the introduction, methodology, visualization, and results sections. JB was responsible for data compilation and drafted the discussion sections. YS and JB reviewed and revised the final version of the manuscript.

Data availability

This study uses publicly available datasets. The Workplace Area Characteristics (WAC) dataset, a component of the Longitudinal Employer-Household Dynamics (LEHD) Origin-Destination Employment Statistics (LODES) program, is accessible via the U.S. Census Bureau at https://lehd.ces.census.gov/data/. American Community Survey (ACS) microdata were obtained from the Integrated Public Use Microdata Series (IPUMS) at the University of Minnesota, available at [https://ipums.org]. The rural and urban classifications from the United States Department of Agriculture (USDA) are publicly available from the USDA Economic Research Service at [https://www.ers.usda.gov/data-products/rural-urban-continuum-codes/]. The code used in this study is openly available at: [https://github.com/YoonjungAhn/Care-Resource-Equity-Score-CaRES].

Competing interests

The authors declare no competing interests.

Ethical approval

This article does not contain any studies with human participants or animals performed by any of the authors.

Informed consent

This article does not involve any human participants, and therefore, informed consent was not required.

Supplementary information

The online version contains supplementary material available at 10.1057/s41599-026-07491-7.