Mapping Childcare Deserts with Public Data

Childcare desert is one of the few terms in commercial real estate analytics with a precise, published, third-party definition: an area where more than three children compete for every licensed childcare slot, or where no licensed provider exists at all. The definition belongs to the Center for American Progress, whose desert database has become the field's reference map, and this article does two things with it. It reads the reference honestly, with its 2026 findings and its stated limits. And it shows how the underlying map is built from public data by anyone, because every layer of it, where the young children are, which parents work, what the industry looks like, what care costs, where the licensed capacity sits, is a federal table or a state roll with a date. For a lender or developer looking at a childcare property, the buildable version matters more than the reference, because deserts are tract-scale phenomena and credit decisions are corner-scale. The reference map answers whether a region is short of care; the buildable map answers whether this site, with this drive-time polygon, these working parents and this roll-verified competitive set, supports these slots at this staffing plan, and no published map can answer that second question because no published map knows the site.

One framing note connects this article to its neighbor in the series. Childcare demand analysis is the senior housing method run at the other end of the age curve: the same single-year-of-age files locate the demand base, the same labor-force tables qualify it, the same payroll census measures the industry, and the same licensure rolls hold the supply numerator. One method, two markets, seventy years apart, and the demographic engines behind them are moving in opposite directions, which makes the pair a controlled experiment in how much of demand analysis is method and how much is tailwind.

The base, counted and shrinking

Start where the cohort math starts. The population under 5 fell from 19,133,927 in July 2020 to 18,491,867 in July 2025, a decline of 3.4%; the 5-to-9 group fell 1.5% and the under-18 population 2.2% (MMCG computation from the Vintage 2025 national single-year-of-age file, 2026). This is the inverse of the senior arithmetic, produced by the same birth series at its other end: fewer children each year, already born or not, with no scenario required to know the direction of the next five years of entrants. A demand analysis for a childcare property that leads with national tailwinds has the sign wrong; the national base is a headwind, mild but persistent, and the honest model says so and then explains why the industry is nonetheless larger than it has ever been, which is where the structure of demand takes over from its size.

The structure: whose children need care

The demand unit is not a child; it is a child whose resident parents all work. The ACS measures exactly that: of 21,170,757 children under 6 in 2024, 9,176,140 lived with two parents both in the labor force, 3,925,052 with a single mother in the labor force and 1,634,958 with a single father in the labor force, a care-relevant base of 14,736,150 children, 69.6% of everyone under 6 (U.S. Census Bureau, ACS 2024 1-year estimates, table B23008, 2025). The table publishes to the tract, which makes the working-parent share the first local multiplier on the child count, and its variation is wide enough to reorder markets: two tracts with identical under-5 populations can differ by 20 points of working-parent share and therefore by a fifth of their formal-care demand. The commuting overlay completes the structure, because parents buy care near home or on the way to work, and the LODES commuting flows show which one a given tract's workers do; a bedroom tract whose workers stream to a distant employment core supports care at both ends of the flow, and the flow is public.

The under-3 subtlety: where scarcity actually lives

The age split inside the base matters more here than in any other family asset class, because regulation prices it. Children under 3 numbered 10,740,251 in 2024 (U.S. Census Bureau, ACS 2024, table B09001, 2025), and infant and toddler rooms carry the tightest staff ratios in every state's rules, which makes them the least profitable and least supplied tier of licensed care; preschool rooms, with wider ratios, carry the economics of most centers. A catchment's under-3 share therefore predicts the shape of its scarcity: a tract stack heavy with infants can show adequate total slots and an acute infant shortage at once, invisible to any metric that pools ages. The state roll usually licenses capacity by age band, and the memo's slot table should keep the bands apart for the same reason the senior article kept 75-and-over apart from 65: the aggregate flatters exactly the tier where the constraint binds.

The industry: smaller base, bigger sector

Against a shrinking child base, the childcare industry has never been larger. County Business Patterns counts 82,162 child day care establishments with paid employees in 2023, employing 1,045,052 people, 36.3% of establishments under five employees (U.S. Census Bureau, County Business Patterns 2023, NAICS 624410, 2025). The employment-insurance series dates the arc: 919,521 employees on annual average in 2019, a pandemic crater to 817,958 in 2021, then recovery through and past the old level to 983,412 in 2024, 6.9% above 2019 across 78,926 reporting units, with average pay up 36.1% to $32,595 (BLS, QCEW annual averages, NAICS 624410, 2020 to 2025). More workers caring for fewer children is not a paradox; it is formalization. Care shifted from unlicensed and family arrangements toward licensed settings, ratios and standards tightened, and the sector's labor intensity deepened, which is precisely why the desert metric is denominated in licensed slots rather than in children per establishment.

The price layer, from a federal index

The price story is measurable without any private survey. The Consumer Price Index for day care and preschool rose from 303.268 in 2019 to 386.417 in 2025, an increase of 27.4% in six years, with the increments accelerating after 2021: 5.6% in 2023, 5.3% in 2024, 5.1% in 2025, and the index at 398.763 by July 2026 (BLS, CPI series CUUR0000SEEB03, 2026). For county-grain price levels the named federal resource is the Department of Labor Women's Bureau's National Database of Childcare Prices, county median prices by age of child and care setting for 2008 through 2022, updated 9 December 2025; its files sit behind access controls that resisted this session's tooling, so this article cites it as the price-level source by name and carries the trend on the CPI, a substitution the memo should state rather than blur (DOL Women's Bureau, NDCP, 2025). Price belongs in a demand analysis for a structural reason the desert literature emphasizes: parents' willingness to pay is capped by wages and by the free alternative of a parent leaving the workforce, so price increases squeeze quantity demanded in a way hotel or storage pricing does not, and a pro forma built on tuition growth above the CPI series' measured trend is making a claim the federal record does not support.

The desert metric, read at the source

The reference map's 2026 update, the first since 2020, reports that 46% of children aged 6 and younger lived in areas with more than three children per available licensed slot, an improvement from just over half in 2018; that 70% of families with young children in the most remote rural areas face shortages; and that the state spread runs from the District of Columbia at 5% and Massachusetts at 21% to Idaho at 83% and Alaska and Hawaii at 96% (Center for American Progress 2026 analysis, as reported by Axios, 29 April 2026, and the CAP desert database). The method's stated limits are as informative as its findings: it counts licensed providers only, so the informal care networks that absorb much of the real load are invisible to it, and it assumes care within roughly 20 minutes of home, which privileges dense areas and understates arrangements near workplaces. Neither limit is a flaw for the metric's advocacy purpose; both are corrections a site-level analysis must make, which is why the buildable map matters.

The improvement between 2018 and 2026 also deserves its honest reading. Part of it is supply, funded by pandemic-era federal support that has since expired; part of it is the denominator, because a desert share falls when the children leave, and the under-5 base shrank 3.4% over the last five measured years. A metric with children in the numerator of its ratio and geography in its frame improves for good and bad reasons alike, and a market analyst quoting the trend should know which reason applies in the counties at hand, a question the county child counts answer directly.

An establishment-density map anyone can rebuild

To show the buildable layer at work, this article computes the simplest supply-side map the federal tables allow: licensed-employer childcare establishments per 1,000 children under 5, by state, County Business Patterns over ACS. The national figure is 4.48; Maine leads at 11.52, followed by Vermont at 10.14, Montana at 9.04, Wyoming at 8.64 and North Dakota at 8.58, with the District of Columbia at 7.67; at the bottom sit Nevada at 2.36, Arizona at 2.45, Tennessee at 3.03, Texas at 3.20 and Utah at 3.25 (MMCG computation from CBP 2023, NAICS 624410, and ACS 2024 table B09001, 2025). Two things make the comparison with the slot-based desert map instructive rather than redundant. The metrics differ in unit, establishments against capacity, so a state of many small providers ranks high here and can still ration slots; and this map counts employer establishments only, while family childcare homes without payroll, a large share of licensed capacity in many states, appear in the Nonemployer Statistics if anywhere. Where the two maps agree, as they broadly do at the extremes, the signal is strong; where they diverge, the divergence names the structure, formal centers versus licensed homes, that a local analysis must resolve from the state roll itself. The District of Columbia illustrates the resolution: near the top of the establishment-density map and, in the reference analysis, the lowest desert share in the country at 5%, a jurisdiction where both metrics agree because a compact, center-based, heavily funded system is exactly the structure both instruments see well.

Building the local map: rolls, gaps and geography

The tract-level analysis that decides a site assembles four public layers. The children: under-5 counts by tract from the ACS five-year file, scaled by the county's Vintage trend, with the working-parent multiplier from B23008. The capacity: every state childcare licensing agency publishes its roll, providers, addresses, license types and, in most states, licensed capacity, which is the honest slot numerator the desert metric standardized; the roll's family-home and center split is the structure variable. The gaps: the County Business Patterns method frames the employer layer by ZIP code, and the reconciliation of roll against census flags the entries each misses. The geography: a 20-minute standard is a modeling choice, and the analysis should run the drive-time polygon of the trade-area method from the site rather than borrow the literature's radius, with the commuting overlay deciding whether home-end or work-end demand governs. The output is the desert arithmetic, children per slot, computed for the actual catchment with a numerator traced to the roll, and it will disagree with the tract-painted reference map exactly as often as a site-grained analysis should.

Who runs childcare: the provider structure

The provider structure explains most of what confuses outsiders about this market. The employer census's 82,162 establishments skew small, 36.3% under five employees, and the census sees only payroll operators; beneath it runs the licensed family childcare tier, homes caring for a handful of children each, which state rolls license and count but federal employer data cannot see, and whose long decline in many states is the quiet supply story behind rising center shares. Above both sit the multi-site operators and the employer-sponsored centers, large enough to appear in the census's upper size classes. The analytical consequence is that no single national series describes childcare supply; the state roll is the only complete frame, and its license-type field, center against family home against group home, is the structure variable every serious catchment analysis tabulates first. The reconciliation habit from the rest of this library applies unchanged: roll against census against parcels, address by address, with each source covering the others' blind spots. One further split hides in plain sight: the payroll files carry ownership codes, and this article's industry figures are the private sector's; school-district preschool rooms and other public programs ride government payrolls in the same files under their own codes, so a market where the public tier is large can look thin in private data while being saturated in fact, one more reason the roll, which is ownership-blind, anchors the count.

The work-end market

The 20-minutes-from-home assumption in the reference method is a modeling convenience, and the commuting data show where it breaks. Parents who commute buy care in one of two geographies, and the choice is visible in the flows: where the LODES matrices show short, dispersed commutes, home-end care dominates and the residential tract stack is the right catchment; where they show long, focused flows into an employment core, a share of demand travels with the parent, and centers at the core's edge, or inside employers' own campuses, intercept it. The work-end market has a further property the home-end lacks: its demand base is countable from the workplace side, jobs by workers' home tracts and by age structure of the workforce, from the same LEHD system. A proposed center beside a hospital campus or a logistics park is not competing in its residential tract at all, and pricing its demand from the neighborhood's child counts mistakes the geography entirely. The blended case is the common one, and the method handles it by running both polygons, home-end from the site and work-end from the employment core, and stating what share of the roll's existing capacity sits in each; where the two polygons barely overlap, the site is choosing its market, and the memo should say which one it chose.

What the lender reads

Childcare centers are core small-balance SBA territory, real estate plus an operating license plus a staffing model, and the credit file's demand half is precisely the map this article built: the working-parent base in the catchment, the roll-verified slot supply, the price context from the federal index, and the structural caution that the national child base is shrinking while formalization pushes the licensed share up. The regulatory layer is the asset's distinguishing risk: ratios, licensure and subsidy programs are state policy, and a demand model should note which of the catchment's slots ride on public subsidy streams, because those streams are appropriations, not demographics. The public SBA 7(a) and 504 datasets, which MMCG Analytics' SBA layer is built on, carry the childcare code's long lending history, subject to the standing rule that no performance rate is shown for any cohort of fewer than ten loans. MMCG Analytics supplies the data and the analysis; the credit decision rests with the lender. And the adjacent question, whether a property is childcare or private early education, changes the demand base enough to have its own article: the daycare against early education piece draws that boundary with the same tables.

The building, briefly

The demand method is the deliverable here, but one paragraph on the real estate keeps the map honest, because licensure writes itself into the building. Licensed capacity is a function of usable square feet per child, outdoor play area, fenced and adjacent, parking and drop-off geometry that survives a fire marshal and a morning rush, and room configurations that match ratio bands; a retail suite that fits ninety children on floor area may license for sixty after the outdoor and egress arithmetic. This is why the parcel layer belongs in the analysis even before any acquisition question: lot configuration screens candidate sites the way zoning does, and a catchment can show demand for two hundred slots with no parcel in it that can legally hold them, which is a finding, not a failure, and one the assessor roll surfaces in an afternoon.

The subsidy geography, from named programs

Because the policy layer moves money through named programs, its geography is public. The federal block grant flows through state subsidy programs whose enrollment and rate schedules are state documents; Head Start and Early Head Start, the direct federal tier, publish their center locations and funded enrollment through the program's own public data, which makes the fully subsidized layer mappable address by address; and state pre-K programs, where they exist, publish theirs. For a private-pay center the mapped subsidy layer is context and competition both: it caps effective demand in some income bands and anchors it in others, and its funding rides appropriations cycles the demand model should name rather than absorb. The practical habit is one column in the catchment inventory, funding type per provider, filled from the rolls and program lists, so the memo can say what share of the polygon's slots are private-pay, subsidy-carried and federally funded, three demand bases with three different risk clocks, and the column takes an hour to fill for a county once the rolls are in hand.

The labor gate, again

Like its senior mirror, childcare sells labor, and the payroll series doubles as the feasibility gate. Ratios are regulatory: an infant room in most states runs near one caregiver per four children, a preschool room wider, so a center's licensed capacity implies its staffing bill within narrow bounds, and the local caregiving wage trend, up 36.1% nationally in five years on the QCEW series, prices it. The gate binds hardest exactly where the demand arithmetic looks best. A high-working-parent tract with thin supply is usually thin on the caregiving workforce too, and the same county QCEW rows that showed the market's employment arc show whether staffed expansion is plausible at the modeled wage. A childcare pro forma whose staffing line grows slower than the measured local wage series is the sector's most common quiet fiction, and the check costs one query.

The two federal series also measure the sector's margin scissors directly. Between 2019 and 2024, average caregiving pay rose 36.1% while the day care CPI rose about 21% over the same five years, reaching 27.4% only by 2025: labor costs outran the price parents paid through the whole recovery, and the gap is the arithmetic behind every operator's complaint and every closure headline, computable from two public indexes without a single interview. For underwriting, the scissors reframe the tuition line: price growth near the CPI trend is not pricing power, it is cost pass-through under pressure, and a pro forma that widens margins while both series say the industry's are narrowing carries the burden of explaining why this operator differs.

The policy layer, named and dated

No asset class in this series leans harder on public policy, and the honest model names its exposure lines. Subsidy: federal block-grant funds flow through state programs whose eligibility and rates are published policy, and the share of a catchment's demand that is subsidy-carried is knowable from the state's own program data. Regulation: ratios and licensure categories set capacity arithmetic directly. And funding cycles: the reference map's own improvement between 2018 and 2026 was funded in part by pandemic-era support that has expired, which the 2026 reporting states plainly. None of this is a forecast; it is a list of the policy variables whose values are public today, with the reminder that a demand model treating subsidy-carried slots as demographically driven has misclassified an appropriations risk as a population fact.

Method: the five numbers a childcare memo should carry

First, the child base: under-5 population for the catchment tracts from the five-year ACS, with the county's Vintage trend and the five-year direction stated. Second, the working-parent multiplier: the share of under-6 children with all resident parents in the labor force from B23008, tract-grained, against the 69.6% national figure. Third, the slot supply: licensed capacity from the state roll by address and license type, reconciled against the employer census, with the center-against-home structure stated. Fourth, the desert arithmetic: children per licensed slot for the drive-time catchment, computed, not borrowed, with the reference map's tract reading noted beside it. Fifth, the operating gates: the caregiving wage trend from QCEW and the subsidy share from the state program data, the two lines that decide whether demonstrated demand is serviceable demand.

Each carries a source and a date; the capture assumption on top is the operator's plan, stated as such; and the boundary between measured and assumed is the same provenance standard the whole pillar enforces.

A worked sequence for one site

Run for a proposed 120-slot center on a suburban arterial, the sequence goes as follows. The tract stack within a 15-minute drive-time holds 3,850 children under 5, down 2% over five years per the county trend, with a working-parent share of 74%, above the national 69.6: a care-relevant base near 2,850. The state roll lists 22 licensed providers in the polygon, 14 centers and 8 family homes, 1,610 licensed slots in total; the computed ratio is 1.77 children per slot on the full base and 2.39 on the care-relevant one, below the desert threshold but tighter than the county average, and the roll shows two centers licensed in the past 18 months. The QCEW county rows price caregiving labor up 7% year over year; the state program data put subsidy-carried enrollment near a fifth of county slots. The memo that results states: a slightly shrinking child base with a strong working-parent multiplier, measured tightness below desert level, fresh competitive entry, a hot staffing market and a known subsidy exposure, and the operator's enrollment plan is then judged against five sourced lines rather than against a national narrative. The protective case is symmetric, and one map query catches the classic error: a site pitched on a desert tract whose actual drive-time catchment crosses into a slot-rich neighbor.

Two more protective checks earn their lines. Age-band the slot table, because a polygon adequate on pooled slots can be an infant desert, and the pro forma's room mix should meet the shortage it claims to serve. And date the roll pulls, because licensing databases turn over: a provider counted in January can be closed by September, and the closure pattern itself, visible by comparing two pulls a year apart, measures the market's churn better than any anecdote about operators leaving the business.

The 2026 read, and the cadence

Read in August 2026, childcare is the series' clearest case of structure beating size. The base is shrinking, 18.5 million under 5 and falling; the demand structure is strengthening, with the working-parent share at 69.6% and formalization still shifting care into licensed settings; the industry is at record employment with pay up a third; prices are compounding above 5% a year on the federal index; and the reference desert share improved to 46% for reasons that are partly supply and partly departed children. Every layer updates on schedule: the estimates each June, the ACS each September and January, the QCEW quarterly, the CPI monthly, the state rolls continuously, the reference map by its own cycle. A childcare position reviewed on that calendar is watched by the instruments that justified it, and the one number this article never used, a national average tuition from a private survey, was never needed.

Set beside its senior mirror, the pair make the series' cleanest demonstration that demography is not destiny but denominator. The old are multiplying and their industry cannot staff itself; the young are thinning and their industry is at record size; and in both cases the investable truth lives not in the direction of the population but in the structure the tables expose underneath it, which is what the next article in this queue, on the boundary between daycare and private early education, takes up with the same instruments.

Frequently asked questions

What is a childcare desert?

By the standard definition from the Center for American Progress, an area with more than three children per available licensed childcare slot, or with no licensed providers. CAP's 2026 update put 46% of children aged 6 and younger in deserts, down from just over half in 2018, with the method counting licensed providers within roughly 20 minutes of home.

What public data builds a childcare demand analysis?

Four layers: under-5 population by tract (ACS and the Vintage estimates), the working-parent multiplier (ACS table B23008; 69.6% of under-6 children have all resident parents in the labor force), licensed capacity from the state licensing roll reconciled with County Business Patterns, and operating gates from QCEW caregiving payrolls and the CPI day care index.

Is childcare demand growing?

The child base is shrinking, 18,491,867 under 5 in 2025, down 3.4% in five years, but formal-care demand grew anyway: childcare employment reached 983,412 in 2024, 6.9% above 2019, as care shifted into licensed settings. Structure, not size, drives the sector, which is why tract-level working-parent shares matter more than national child counts.

How much have childcare prices risen?

The CPI for day care and preschool rose 27.4% between 2019 and 2025, with annual increases above 5% in 2023, 2024 and 2025. County price levels through 2022 are published in the Department of Labor's National Database of Childcare Prices.

Which states have the largest childcare deserts?

In the 2026 reference analysis: Alaska and Hawaii (96% of young children in deserts) and Idaho (83%), against the District of Columbia at 5%, Massachusetts at 21% and New Jersey at 25%. On the simpler establishment-density map computed in this article, Maine leads supply at 11.52 employer establishments per 1,000 under-5 and Nevada trails at 2.36.

Why does staffing belong in a childcare demand model?

Regulated ratios make capacity a direct function of staff, so slots without caregivers are not supply. Caregiving pay rose 36.1% from 2019 to 2024 on the QCEW series, and a catchment's wage trend, from the same county files, is the feasibility gate on any expansion the demand arithmetic appears to justify.

Sources

1. U.S. Census Bureau, Vintage 2025 national population estimates by single year of age (nc-est2025-agesex-res), 2026; child bands computed by MMCG. https://www2.census.gov/programs-surveys/popest/datasets/2020-2025/national/asrh/nc-est2025-agesex-res.csv
2. U.S. Census Bureau, American Community Survey 2024 1-year estimates, table B23008, children by presence of parents and parents' labor-force status, released 11 September 2025. https://www2.census.gov/programs-surveys/acs/summary_file/2024/table-based-SF/data/1YRData/acsdt1y2024-b23008.dat
3. U.S. Census Bureau, American Community Survey 2024 1-year estimates, table B09001, population under 18 by age, 2025; state under-5 denominators. https://www2.census.gov/programs-surveys/acs/summary_file/2024/table-based-SF/data/1YRData/acsdt1y2024-b09001.dat
4. U.S. Census Bureau, County Business Patterns 2023, NAICS 624410, national and state files, released 26 June 2025; density per 1,000 under-5 computed by MMCG. https://www2.census.gov/programs-surveys/cbp/datasets/2023/
5. U.S. Bureau of Labor Statistics, Quarterly Census of Employment and Wages, annual averages 2019, 2021 and 2024, NAICS 624410, private, United States, 2020 to 2025. https://data.bls.gov/cew/data/api/2024/a/industry/624410.csv
6. U.S. Bureau of Labor Statistics, Consumer Price Index, day care and preschool (series CUUR0000SEEB03), annual averages 2018 to 2025 and monthly 2026, 2026. https://api.bls.gov/publicAPI/v1/timeseries/data/CUUR0000SEEB03
7. U.S. Department of Labor, Women's Bureau, National Database of Childcare Prices (county prices 2008 to 2022), updated 9 December 2025. https://catalog.data.gov/dataset/national-database-of-childcare-prices
8. Center for American Progress, U.S. child care deserts database and 2026 analysis (46% of children 6 and younger in deserts; definition and method), 2026. https://www.americanprogress.org/feature/child-care-deserts/
9. Axios, Nearly half of families with young kids live in a child care desert, new report finds (Emily Peck; reporting the CAP 2026 figures, state extremes and method caveats), 29 April 2026. https://www.axios.com/2026/04/29/child-care-daycare-working-parents
10. U.S. Census Bureau, Nonemployer Statistics 2023 (the no-payroll family childcare tier), released 15 May 2025. https://www.census.gov/newsroom/press-releases/2025/2023-nonemployer-statistics.html
11. MMCG Research, SBA 7(a) Performance Series: MMCG analysis of the public SBA 7(a) loan register, 2026. https://mmcganalytics.com/sba-default-rates/