Industry Data: Dataset Structure and Field Coverage

Because capitation is a state-set contract price, not a federal statistic. KFF's 10 Things to Know About Medicaid Managed Care states the mechanism plainly in its seventh finding: each year, states develop MCO capitation rates that must be actuarially sound. The same finding records what happens to those prices in practice — a market-wide average medical loss ratio that rose from 88% in 2023 to 91% in 2024, with 33 states always requiring remittance payments when plans fall below minimum MLRs.

The consequence for anyone hunting a master file is structural. Rates vary by state, program, plan and delivery model, and the payment side of the bargain sits with each state agency. What the federal layer does publish is everything you need to weight and audit those prices: enrollment at state-program-plan grain on Data.Medicaid.gov, catalog discovery of the state rate files themselves through Data.gov, and the analytical frame from KFF. Payments to comprehensive risk-based MCOs ran to roughly half of total FY 2024 Medicaid spending, so the dollars riding on these files are enormous even where the rate schedules stay local.

One cataloged source carries the rate schedules directly, and three more carry the numbers you need to interpret them.

Data.gov — Managed Care Full-Text Search (quality score 9) is the direct hit. Its federated index of 552,271 catalog datasets surfaces the California Department of Health Care Services' Medi-Cal Managed Care Capitation Rates, published as separate files per delivery model — Geographic Managed Care (GMC), Two-Plan, Regional/Rural Expansion and COHS — alongside the Medi-Cal Managed Care Enrollment Report and the Newly Eligible Individuals enrolled in Medi-Cal Managed Care Health Plans series. Every verified record in Datadory's sample carried has_download: true, with direct CSV or JSON distributions and harvest timestamps current into July 2026.

KFF — State Health Policy: Medicaid & CHIP Indicators (quality 9) supplies the oversight variables: roughly 209 state-level indicators including an 11-indicator Managed Care Market Tracker whose contract-requirements section covers States Reporting Medicaid MCO Minimum Medical Loss Ratio (MLR) Remittance Requirements and total Medicaid MCO spending by state. Data.Medicaid.gov — CMS Medicaid Data Portal (quality 9) contributes the membership denominators — 549 public-domain datasets including the MMCC MCPAR PUF of MCO financial reports — and KFF's issue brief supplies the rate-setting rules and directed-payment caps that bound every schedule.

Six steps take you from zero to a defensible multi-state rate workbook, using only keyless free access routes:

1. Discover the state rate files. Query the JSON API returns a {after, results[], sort} envelope, and feeding the after token back paginates. Filter results to publisher = California Department of Health Care Services to isolate the four Medi-Cal capitation model files.
2. Pull the rate CSVs. Each result's dcat.distribution[].downloadURL points straight at files hosted on the publishing agency's site, in JSON, CSV, XLSX, XML or HTML formats. Respect robots.txt Crawl-Delay: 10 if you script repeated pulls.
3. Fetch the CMS catalog feed. GET returns all 549 DCAT-US records in one call of roughly 1.2 MB; filter titles containing “Managed Care Enrollment” for the denominators.
4. Query rows at plan grain. /api/1/datastore/query/{distributionIdentifier} takes the distribution UUID (not the dataset identifier — the datastore returns ‘No resource found’ otherwise) and responds with count, results and a schema block. Numeric columns arrive typed as text, so strip commas before aggregating.
5. Add the policy frame. From KFF's Medicaid & CHIP category, pull the MLR remittance and spending indicators through /wp-json/kff/v1/google-sheets-full?id=<gdocs_key>, which returns year-period tables without authentication.
6. Normalize before comparing. Rate files are per-member-per-month prices specific to one delivery model and eligibility group, so join them to enrollment counts rather than averaging raw rates across models — and label each figure with its program year.

Enrollment and financial tables are how you test whether a published rate is adequate. Data.Medicaid.gov's annual Managed Care Enrollment tables run from 2016 forward with identical column names — nine editions through 2024 — and the plan-level Managed Care Enrollment by Program and Plan file breaks membership out at state-program-plan-region level; the 2016 edition alone holds 7,804 rows. The national TOTALS row for 2016 reads 80,263,839 total Medicaid enrollees, 65,034,032 in some type of managed care and 54,547,048 in comprehensive MCOs.

Membership times rate is the revenue line, so these tables let you back into implied premiums. A verified sample row shows the grain available: Molina Healthcare of New Mexico's Centennial Care program (comprehensive MCO plus MLTSS) carried 225,105 Medicaid-only members plus 7,597 duals statewide in 2016. On the financial side, the portal hosts MCO reporting PUFs such as the MMCC MCPAR PUF, and T-MSIS analytic files add monthly claims-level detail. Pair those against KFF's finding that payments to comprehensive MCOs equal roughly half of total Medicaid spending (FY 2024) and that five publicly traded firms — Centene, UnitedHealth Group, Elevance, Molina and Aetna/CVS — held 47% of enrollment across the 291 MCOs states contracted with as of July 2024, with 15 parent firms over 62%.

The same hub's archive frames where rates go next: the 2025 reconciliation law caps state directed payments at 100% of Medicare rates in expansion states (110% non-expansion), with CBO-estimated savings of $149 billion from SDP changes and $226 billion from provider tax restrictions — both of which squeeze the supplemental layer that sits on top of base capitation. For machine-speed watchlisting, point a diff job at the Data.gov searches instead: daily harvests plus last_harvested_date timestamps make new rate-file publications detectable within a day. Policy researchers working this loop routinely pair it with the industry's wider landscape on the journalists and academics page.