Methodology & sources

How the numbers on this map were built, what they mean and what they don't.

The question

New York City's Child Care Hub map shows roughly 10,000 licensed programs but doesn't summarize them. This map answers: how much licensed child care supply exists relative to the number of young children who could use it, neighborhood by neighborhood?

Data sources

These are the two public datasets covering the two halves of New York City's regulated child care: the state Office of Children and Family Services (OCFS) regulates home-based and school-age care, and the city Department of Health and Mental Hygiene (DOHMH) regulates center-based (Article 47) care. Combined they total roughly 11,000 programs — comparable to the ~10,000 the city's hub map advertises — but the hub draws on the city's live internal center feed, which we could not access, so this is not a byte-for-byte reproduction of the hub.

How programs were counted

• OCFS file is one row per program — used as-is, filtered to the five boroughs (the dataset labels them Bronx, Brooklyn, Manhattan, Queens, Staten Island).
• DOHMH file is one row per inspection (26,397 rows, dated May 2020–April 2023). It was de-duplicated to one row per center by name + ZIP + borough, keeping the largest capacity seen — yielding 2,359 unique centers. The file carries no active/closed status, so a center inspected in that window is counted even if it has since closed.
• Combined universe: 11,020 programs mapped to a neighborhood. Five records are dropped for having no usable New York City ZIP (four blank, one a New Jersey ZIP mislabeled to a borough in the source).
• Post-2010 ZIP fix: the MODZCTA crosswalk is built on 2010 ZIP areas, so newer ZIPs are missing. The largest, 11249 (South Williamsburg — 43 programs, 2,432 slots, 4,700 children under 5), was folded into its parent zone 11211, whose polygon still physically covers that waterfront; two tiny office-district ZIPs (10281, 10166) were likewise reassigned. Without this fix Williamsburg was absent from the map on both the supply and population sides.

Program types

The headline metric — early-childhood slots per 100 children under 5 — deliberately excludes school-age (SACC) capacity, because after-school seats don't serve infants and toddlers. Mixing them would overstate supply for the youngest children. School-age slots are still shown in each neighborhood's detail popup.

Within the DOHMH center file, all three center subtypes — "Child Care - Pre School," "Child Care - Infants/Toddlers," and "School Based Child Care" (school-based pre-K/3-K, which serves 3–5-year-olds) — are counted as early childhood.

The metrics

• Home-based slots per 100 kids under 5 (default view) = (GFDC + FDC licensed capacity) ÷ (population under 5) × 100. This uses only the current data — family and group-family day cares, updated daily — and excludes the 2023 center snapshot entirely. It understates total supply (centers are most of it) but every number in it is current. On this view the Bronx (50 per 100) stands far above every other borough (11–14).
• All early-childhood slots per 100 = adds day-care-center capacity (DCC), i.e. (DCC + GFDC + FDC) ÷ population under 5 × 100. More complete, but the center portion is the April 2023 snapshot — read it as "supply as of 2023."
• Number of programs, total licensed slots (all ages), and children under 5 are also selectable.

The default was set to the home-based-only view specifically so the map's headline number is fully current; the center-inclusive view is one click away for a more complete but older picture.

Limitations

• Capacity is licensed maximum, not enrollment. A program licensed for 60 may serve fewer children. This measures regulatory supply, not utilization.
• Slots aren't age-locked. "Early childhood" program types serve mostly under-5s but the exact age mix varies; treat the ratio as an index of supply, not a precise seat-per-child guarantee.
• A program's ZIP is where it's licensed, not necessarily where its families live; children cross neighborhood lines for care.
• No cost data. Neither source publishes price in a machine-readable field, so this map covers availability only.
• Neighborhood = Modified ZCTA, a ZIP-code approximation of neighborhoods, not official Neighborhood Tabulation Areas.
• Denominator noise. A handful of low-population ZCTAs (business districts, e.g. 10282, 10007) can show extreme ratios in either direction because the child population is tiny; read those with caution.
• How robust are the rankings? The Bronx's first-place finish holds even on the current home-based data alone (50 slots per 100 under-5, versus 11–14 for every other borough), so it does not depend on the stale center layer. Manhattan's second-place rank, by contrast, rests almost entirely on the April-2023 center snapshot. And Queens (36.5) versus Staten Island (35.6) for last place is effectively a tie — don't read a firm ordering into that gap.
• Population overcount. Summed across ZCTAs, total population comes to ~8.58M, modestly above the city's ~8.3M, because some ZCTAs spill across the city line. This mainly affects edge neighborhoods; borough-level under-5 counts are close to standard estimates.

Reproducibility

Raw pulls and the build script (build_data.py) live in the project repo. The script joins the four sources, computes the metrics and writes childcare.geojson and summary.json, which this page loads statically. No live API calls at page load.