Child care affordability is a well-documented issue for families nationwide, but even with expanded subsidies that would bring the cost of child care within reach, an insufficient supply of child care slots still fuels lengthy waitlists and leaves families with limited options.
Understanding the availability of licensed child care options is a crucial step toward addressing systemic issues and the lack of public investment in the sector that keep prices high, options for families low, and providers and educators living on subsistence wages. This brief provides an overview of the underlying methodology for the 2025 analysis of licensed child care deserts.
Explore the map
What are licensed child care deserts?
Child care deserts are areas with little local licensed child care capacity, which includes center-based care, home-based care, Head Start, and state preschool programs, relative to the number of local young children, such that there are at least three local children for every one available local licensed slot. Empirically, this measure is limited to licensed care and does not account for informal, unlicensed, or unregistered options such as care provided by family, friends, or neighbors (FFN). Although a comprehensive child care system would rely on all types of care, lack of data on FFN prevents the inclusion of these informal options in feasible localized measures of child care deserts.
Despite this limitation, child care deserts analysis is useful for highlighting significant imbalances in the local supply of licensed care relative to the number of local children. Little supply relative to potential demand means that families may face extensive waitlists, excessive commuting times, high costs, and even a complete lack of viable options, hindering their ability to work and contribute to the local economy.
Furthermore, these gaps can result in missed early education opportunities during a critical developmental period, leading to persistent academic disparities that can strain the K-12 system and negatively affect children’s long-term outcomes.
Where do licensed child care deserts data come from?
Child care deserts data combine data on provider locations and capacities from state child care licensing agencies with U.S. Census Bureau and other data on where young children live. Distances between family locations and provider locations are a key ingredient in assessing relevance and access. A desert is a location where there is little local care capacity per local young child.
Provider location data are derived from multiple sources, including state administrative data on child care provider licensing, federal Head Start administrative data, and U.S. Department of Education Common Core of Data for America’s public schools. Researchers at the W.E. Upjohn Institute for Employment Research and Stanford University collected these data, with support from the Center for American Progress, between December 2024 and September 2025. State administrative data reflect conditions in 2024 and 2025 according to what was most recently available in state records; federal data related to Head Start and Common Core of Data represent the most recently available data.
Center-based and home-based child care provider data were collected by a combination of web scraping and direct outreach to state administrators. Freedom of Information Act (FOIA) requests were submitted for 11 states (Washington, D.C., Delaware, Idaho, Louisiana, Maryland, Mississippi, New Jersey, New Mexico, North Carolina, Oklahoma, and Oregon). Head Start data are derived from the 2024 Program Information Reports. All schools with at least one pre-kindergarten student enrolled, based on the 2022-23 Common Core of Data school year, were counted toward the preschool total. Select states reported child care facility type as well as the ages of the children served. For these states, the authors removed schools that operate exclusively before- or after-school programs, drop-in only care, day camps, or care for school-aged children to limit results to care for children ages 5 and under who are not yet enrolled in school and participate in the licensed child care system.
Data Validation
To assess the completeness and accuracy of the authors’ administrative data collection, the authors conducted a validation exercise using multiple federal data sources. The dataset includes the precise location and licensed capacity of every licensed child care center and family child care home across all 50 states and the District of Columbia (with limited exceptions to data from Kansas),1 primarily based on 2025 administrative records.
To validate these data, the authors compared state-reported counts to business establishment counts from three federal sources, using the most recently available data from each:
- 2024 Quarterly Census of Employment and Wages (QCEW)
- 2023 County Business Patterns (CBP)
- 2023 Nonemployer Statistics (NES)
Comparison by provider type
For child care centers, the authors compared aggregated state-level numbers of establishments counted under NAICS 624410 (child day care services) from both the QCEW and CBP. These figures were then compared to the total number of child care centers reported in the state administrative data. Because the QCEW and CBP include both traditional child care centers and Head Start centers under NAICS 624410, the authors combined these counts in the state administrative data for comparison purposes. It is important to note that federal sources like QCEW and CBP may include both traditional child care providers and Head Start centers under the NAICS 624410 classification, which are typically tracked separately in state systems.
For family child care homes, the authors compared the state administrative counts with the number of nonemployer businesses in NAICS 624410 reported in the NES, which captures self-employed individuals operating without paid employees.
Adjustments and interpretation
The authors recognize that exact alignment between administrative and federal data is not expected due to several factors:
- The administrative data reflect conditions in 2025, while federal datasets represent the most recently available years (2023 or 2024).
- State licensing requirements vary, and some states do not require all family child care homes to be registered or licensed, leading to undercounts in the administrative data relative to NES (which includes all self-employed providers).
To enable a more consistent comparison across sources, the authors calculated state-level shares of national totals (e.g., the share of all child care centers in the United States located in California, based on each data source). This allowed for an assessment of whether the administrative data aligned with expected patterns from federal data sources, even when absolute counts differed. This validation exercise confirmed that the administrative data fall within reasonable bounds of external benchmarks, with expected deviations primarily explained by definitional differences and data timing.
How are licensed child care deserts measured?
The authors measure the location of young children’s residences beginning with 2023 American Community Survey five-year block-group data. To measure the block group’s number of children relevant for infant, toddler, or preschool age care, the authors scaled up the measured number of children under age 5 by one-fifth to represent the number of children under age 6. The country was partitioned into a set of hexagons, with sides 1-mile long in nonmetro areas and 1/4-mile long in metro areas. To account for misalignment between block-group and hexagon boundaries, young children were allocated to hexagons proportionally based on the share of each block group’s area overlapping each hexagon, assuming a uniform distribution of children within block groups.
Geographic coordinates of licensed child care programs were geocoded using ArcGIS, based on the administrative data collected from each state and Washington, D.C. For each hexagon, providers within a 30-mile straight-line distance were identified using a Near Table tool in ArcGIS, and centroid-provider pairs within that candidate set were used to calculate driving distance with OSRMtime command in Stata, which queries the Open Source Routing Machine API to estimate travel distance and time along the road network for each centroid-provider pair within the candidate set.
To visualize children’s residential locations on the map, one synthetic family location was drawn for every 10 young children in each block group. Uninhabitable areas, such as rivers, lakes, and state or national parks, were ruled out. Then, the synthetic families’ locations were assigned a specific location at random within the block group’s inhabitable area. The dots on the map were placed at these synthetic family locations. This assured that the dots were distributed approximately where families with children under the age of 6 live, such that each dot represents 10 children. The child care access value assigned to each family location follows from its hexagon’s supply measure.
The 2020 child care deserts map2 used the same methods, though the current data collection process relied more on web scraping and less on FOIAs. A total of 4,636 providers (approximately 2.4 percent of all geocoded providers) lack specific addresses due to limitations in available data. These cases are largely concentrated among home-based providers whose addresses, for confidentiality reasons, may be redacted in administrative records.
Distance-based methodology
As was the case in the 2020 licensed child care data map, a distance-based approach to measuring supply was applied to the current study to more accurately establish measures of local access. This section provides context for the use of the distance-based methodology and describes the process by which updated local access measures were calculated using 2025 licensed child care provider data.
The distance-based approach was initially developed by Davis, Lee, and Sojourner (2019) and was published in Early Childhood Research Quarterly.3 For the current project, the approach was adapted for use on a sample of licensed providers from all 50 states and Washington, D.C.4 By using this methodology, the current study avoids known analytic limitations to the use of administrative or political boundaries as the unit of measure, which are arbitrarily defined, poor geographical markers for assessing where families tend to access care for their children. Those zone-based measures of access to licensed supply historically have measured the total capacity of licensed providers within that given area—a ZIP code, county, or congressional district, for example—divided by the number of young children in that zone. But that approach introduces the potential for information loss and statistical bias, neglects to account for where families might realistically be searching for care, which, for any number of reasons, might cross zone-based boundaries. Families do not restrict search for child care to their ZIP code or county.
The distance-based methodology instead centers family locations as the unit of analysis, establishing local access measures relative to a certain drive-time radius around the family’s hexagon’s central point, and adjusting nearby provider capacity based on the number of young children who could potentially compete to enroll in that provider’s program. Having more licensed capacity in nearer-by programs increases a family’s measured access. Having fewer other young children nearer by those providers also increases a family’s measured access.
Access measures
For each provider, program capacity was divided by the distance-weighted sum of children under the age of 6 living within a 20-minute drive-time catchment area to calculate a capacity-to-population ratio. These provider-specific ratios were then distance weighted again and summed across all providers within 20 minutes of a child’s location to represent demand-adjusted supply for that family location (hexagon centroid). This produces an enhanced two-stage floating catchment area-based measure of supply for each hexagon.5
CAP has long used the threshold of 0.333 local slots per local child to determine whether a family lives in a child care desert. If local access is below that level, then the observation qualifies as a licensed desert; if the measure falls above that threshold, it does not qualify. This threshold corresponds to the existing definition of a child care desert, for which there are more than three local young children per local licensed child care slot. Aggregates of the synthetic family locations were used to generate average local access measures by state, with thresholds set at 0.333 (child care desert), 0.10 (extremely low supply), and 0 (no licensed supply), acknowledging that measures for extremely low supply and no licensed supply both fall under the threshold for a child care desert. Calculations were performed using R version 4.3.3 (2024-02-29); Copyright© 2024 The R Foundation for Statistical Computing.
Limitations
Like previous work on licensed child care deserts, the present study has some limitations related to the assumptions used to define a desert, to establish access measures, and to obtain data. Child care deserts can currently only account for licensed providers because data on FFN capacity are lacking. As described above, the data were largely derived from state licensing agencies, but there are cases of incomplete data, inadequate data about the ages of children served, and no available data about enrollment nor true program capacity, which accounts for staffing or operational shortages. This family residence-centered measure also assumes parents prefer care that is closer to their home, rather than where they work or where their children attend school, but this may vary significantly by family or by the child’s age and the true availability of local care. Data on the joint residence and work locations of parents of young children are lacking. Additionally, while in the aggregate these analyses provide a snapshot of licensed supply and access, they do not disaggregate by other factors that may influence whether a parent chooses a specific program, including whether they have a child with special needs, available prices and subsidies, nor if they need a program that operates at nonstandard hours.
Appendix 1: Administrative data sources
Appendix 2: Census tables for map overlay