States M1 Enrollment tracker: medicaid_mco_sec_tieout_v2

What it is

1. 1Medicaid product M1 (Enrollment tracker) of the States lane: the Medicaid MCO enrollment tracker: members by plan and parent by state by month, every cell graded (observed, estimated, untested, not published), with parent rollups and the quarterly tie-out to what public parents report to the SEC.
2. 2Start from medicaid_mco_enrollment_tracker_v2 (state x plan x parent x month) or its views vw_medicaid_mco_enrollment_tracker_v2_parent_state_month / _parent_national_month.
3. 3READ basis ON EVERY ROW: observed is the state's printed figure; estimated is ours, and method names how (M-a carry-forward, M-b plan share x state total, M-e the tracker's nowcast model, PUB-PRELIM the publisher's own preliminary estimate; docs/research/states/M1-TRACKER-V2-DESIGN.md section 4), with interval_low / interval_high, inputs_as_of and assumption_ids. modeled is an M-e value where the method missed the accuracy bar: INDICATIVE ONLY, its interval is the measured error, and it is excluded from members_total; never sum it into an observed or estimated total (the views carry members_observed, members_estimated, members_modeled and members_with_modeled apart). not_published and not_applicable rows have members null: they are not zeros.
4. 4Filter measure_class (members, member_months, weighted_members, average_monthly_members) and never add across it: average_monthly_members (with averaging_window;
5. 5RI from 2023-12 prints average daily enrollment per month) never adds into members. measure is the publisher's own label.
6. 6California is observed 2021-01..2026-09;
7. 7Kaiser's publisher-suppressed county cells carry a labelled lower bound (assumption A-SUPPRESSED-LB).
8. 8TWO PARENT QUESTIONS, two view pairs: the dated views vw_medicaid_mco_enrollment_tracker_v2_parent_dated_state_month / _parent_dated_national_month answer 'who owned it then' (parent_as_of_month;
9. 9WellCare is its own parent before 2020, so Centene July 2016 is 5.92M there); the current views _parent_state_month / _parent_national_month answer 'today's companies' history' (Centene July 2016 is 9.93M there), with parent the CURRENT binding today (parent_status, parent_basis); never mix the two in one series.
10. 10Assumptions: medicaid_mco_enrollment_assumption_v1 is the per-state register (populations counted, count basis, auto-assignment, carve-ins); medicaid_mco_enrollment_assumptions_v1 is the build's.
11. 11SEC: read capture ratios only from vw_medicaid_mco_enrollment_tracker_v2_sec_tieout (like for like, duals adjusted per docs/research/states/M1-SEC-COUNT-BASIS.csv); comparability_class says when the filed number is broader than the tracker.
12. 12SHOW THE GRADED adj_* COLUMNS, NEVER ONLY THE RATIO: each adjustment is its own labelled column (e.g. adj_subcontract_molina_la / _out measured from DMHC, adj_risk_share_healthy_blue_nc, adj_subcontract_lacare_anthem ESTIMATED with its low bound in the detail, adj_dsnp_from_ma_gap a JUDGMENT that a dual is counted in both lines, the 'once' alternative in the detail); a capture above 1.00 (UNH Q2 2026 1.007) means a judgment column over-explains.
13. 13The two NY columns keep their _held names for schema stability but are LIVE: adj_ny_chp_held (NY Child Health Plus members, CNC 188,720, ELV 40,405) and sec_ny_essential_plan_excluded_held (NY Essential Plan is not Medicaid and comes off the SEC side, CNC -388,602), source NY DOH healthweb-back.health.ny.gov, recorded by the States lane as allowed by Ben on 2026-11-01; null where no NY figure exists or it is not measured (UNH), never zero. sec_asc_excluded is per quarter; exits are listed apart.
14. 14ESTIMATE INTERVALS widen with the TOTAL horizon where the state total is carried past its last published month (months since the last observed plan value + months the total is carried): inputs_as_of is the last published total month and value_status says 'state total carried from <month>'.
15. 15TEXAS: observed monthly by MCO x program 2007-09..2026-02 from HHSC MCO Financial Statistical Reports (tx_medicaid_mco_fsr_member_months_v1: filter is_latest_filing = true, since restated filings stay as rows, and grain); after 2026-02 plan values scale by HHSC's own monthly total.
16. 16234 TX observed members rows are FRACTIONAL by rule (part-month adjustments in the filed member months;
17. 17HHSC defines the line as members enrolled in the month and it ties to HHSC final SFY averages within 0.04-0.38%, assumption A-TX-FSR-RETRO): they stay measure_class members, never rounded, so a fractional-members check exempts source_family state_fsr. measure_class members_hhsc_prelim_plan_total (HHSC 2026-05 preliminary per-plan totals) never adds to members. tx_medicaid_mco_enrollment_other_sources_v1 holds TDI HMO quarter-ends, 1115 waiver tables, CHIP 2003-13 and HHSC preliminary months, one series per row (filter series and period_kind), never mixed with the FSR months.
18. 18INPUTS behind the tracker, read only with their labels: medicaid_m1_program_observation_v1 is the programs the spine lacked (MN, PA CHC, NY MLTC/MAP/FIDA-IDD, IN PathWays, CO Denver Health, OH MyCare, WI Family Care, WV MHP, CA SCAN, DC Dual Choice), one row per published cell (read value_unit and source_time_basis: some are member months). medicaid_mco_plan_enrollment_hunt_v1 and _hunt_cogank_v1 are hunted series (ND VT CO GA monthly, MA AR quarterly, NE EQR points, KS year-to-date averages): filter tracker_use, because only observed_monthly and observed_point rows are observations; quarterly_anchor and average_anchor are anchors, check_only and not_read are never a figure, and grain says month, quarter, rolling_quarter, calendar_ytd or point_in_time. medicaid_mcpar_plan_avg_enrollment_v1 is MCPAR annual average monthly members by plan (PY2023-25, 48 states), an anchor only, never a monthly count; suspect = true rows (77) are flagged. medicaid_mco_state_month_reconciliation_mat_v1 is the earlier plan-sum vs state and CMS totals reconciliation (cms_monthly_denominator_available = false is a missing denominator, never zero).

Grain

Columns (84)

Query it

describe_data_source(source_fqn="medicare_silver.medicaid_mco_sec_tieout_v2")     # the card, live
query_substrate(sql="SELECT * FROM medicare_silver.medicaid_mco_sec_tieout_v2 WHERE <filter> LIMIT 100")