Methodology: Medicaid Coverage at Risk
How county-level Medicaid exposure ranges are computed under P.L. 119-21
Exposure is not disenrollment
This page identifies Michigan Medicaid enrollees who fall into categories that Urban Institute, CBO, and CBPP project may be affected by provisions of P.L. 119-21. Being in an affected category is not the same as losing coverage. Individual outcomes depend on state implementation decisions, work requirement compliance pathways, income and employment changes, and administrative factors that vary by county and over time. The ranges on this page describe populations at elevated exposure — they do not predict who will or will not retain Medicaid enrollment.
Sources
Urban Institute: Projected Reductions in Medicaid Expansion Enrollment Under OBBBA's Work Requirements
Urban Institute, March 2026. urban.org — Medicaid work requirements projection — Post-enactment Michigan-specific projection: 171,000–355,000 Michigan adults in categories projected to be affected by P.L. 119-21 work requirement provisions by 2028. This is the primary state-level input to the county allocation model. This is the most current post-enactment Michigan-specific figure available as of April 2026. Accessed April 2026.
KFF: Allocating CBO's Estimated Federal Medicaid Spending Reductions to States Under the Enacted One Big Beautiful Bill
Kaiser Family Foundation, December 2025. kff.org — Federal Medicaid spending reductions by state — Michigan 10-year federal Medicaid spending reduction: $31.6 billion under P.L. 119-21. This is a spending figure, not an enrollment count. Used as statewide spending-context statistic; not used as the county allocation input. Accessed April 2026.
CBO: Estimated Budgetary Effects of P.L. 119-21 — Medicaid Provisions
Congressional Budget Office, July 2025 (pub. 61570). cbo.gov — pub. 61570 publication landing page — National baseline: 7.5 million coverage loss by 2034; $326 billion federal savings from work requirements specifically; $63 billion / 700,000 coverage loss from redetermination changes. Urban Institute's Michigan projection is grounded in this CBO national score. Accessed April 2026.
GAO-20-149: Medicaid — Actions Needed to Address Weaknesses in Oversight of Costs to Administer Work Requirements
U.S. Government Accountability Office, 2020. gao.gov/products/gao-20-149 — GAO-20-149 documents the administrative costs Arkansas and other states incurred to implement Medicaid work requirements. Cited here for context on implementation burden, not as a source for coverage loss estimates. Coverage loss figures (approximately 18,000 Arkansas disenrollees) come from Sommers et al. NEJM 2019 (source 5 below). Accessed April 2026.
Sommers et al., NEJM 2019: Medicaid Work Requirements — Results from the First Year in Arkansas
Benjamin D. Sommers, Alister D. Martin, Robert J. Blendon, E. John Orav, Arnold M. Epstein. New England Journal of Medicine 2019;381:1073–1082. nejm.org — Arkansas work requirements study — Peer-reviewed Arkansas study. Approximately 18,000 Arkansas adults lost Medicaid coverage during the first year of work requirement implementation (June 2018 through April 2019), when a federal court halted the policy. Urban Institute's Michigan projection derives its loss rate partly from this Arkansas evidence. Accessed April 2026.
ACS B27010 / C27007 (5-year 2023): Michigan County Medicaid Enrollment
U.S. Census Bureau, American Community Survey 5-year 2023. data.census.gov — ACS C27007 — County-level Medicaid/means-tested public coverage enrollment. ACS B27010 is the design-document denominator; ACS C27007 (Medicaid/Means-Tested Public Coverage by Sex by Age) was used as the implementable county-level equivalent. County proportional shares — not absolute ACS values — are used for county allocation. The ACS county total (6,206,095 statewide) differs from the CMS MBES administrative enrollment (~2.4M point-in-time); the county proportion method preserves Urban Institute's statewide range regardless of this difference. ACS surveys are known to undercount low-income, non-citizen, and highly mobile populations; county shares used here may slightly underrepresent counties with higher concentrations of these groups. Accessed April 2026.
Perspectives supporting P.L. 119-21 Medicaid work requirement provisions
The Trump administration and P.L. 119-21 proponents frame work requirements as promoting self-sufficiency and labor force participation. CMS has issued guidance supporting Section 1115 demonstration waiver programs that include work requirements as a condition of Medicaid eligibility (see CMS Section 1115 demonstration waivers). Heritage Foundation and AEI have published analyses arguing that work requirements increase employment among Medicaid enrollees and reduce long-term dependency on public assistance. As of April 2026, no published analyses from these organizations provide Michigan-specific county-level enrollment projections comparable to the Urban Institute / CBO figures used here. The CBO score (source 3 above) is the enacted-law authoritative figure and is used here as a neutral baseline. If Michigan-specific analyses from any perspective are published, this methodology will cite them.
Projection methodology
Plain language
We take Urban Institute's Michigan-specific projection of 171,000–355,000 adults in affected categories under P.L. 119-21 work requirements by 2028 (March 2026) and apply it proportionally to each Michigan county based on each county's share of the state's total ACS-reported Medicaid enrollment. We use Urban's range directly — without adding a second uncertainty band — because Urban's low and high endpoints already encode their own implementation scenario modeling. We present the result as a range, never a point estimate.
Technical steps
- Baseline county enrollment — County Medicaid enrollment is sourced from ACS 2023 5-year, table C27007 (Medicaid/ Means-Tested Public Coverage by Sex by Age), variables C27007_003E (male) and C27007_012E (female) summed per county. All 83 Michigan counties returned clean values. Source: ACS B27010 / C27007 5-year 2023 (see Sources above).
- Statewide calibration — The ACS county survey total (6,206,095 statewide — sum of C27007 estimates for all 83 Michigan counties) differs from the CMS MBES administrative point-in-time enrollment (~2.4M). The ACS "means-tested public coverage" definition is broader, and the 5-year window averages PHE-era enrollment peaks. We do not apply a calibration multiplier to county values — doing so would require assumptions not supported by public data. The ACS county totals are used only for their relative distribution (shares), not as absolute enrollment counts. This is disclosed here and on every county row's tooltip.
- State-level projection input — Urban Institute (March 2026) projects 171,000–355,000 Michigan adults are in categories at elevated exposure under work requirement provisions by 2028. This range is used directly as the state-level input. Urban's range already encodes implementation scenario uncertainty; no additional band is applied by accessmi.org.
- County allocation — Each county's projected loss range is computed as:
county_loss = county_ACS_enrollment / sum(all_county_ACS_enrollments) × urban_state_figurecomputed at both the low (171,000) and high (355,000) endpoints. This uses straight ACS enrollment share as a proxy for work-requirement-eligible adult share at the county level. A more precise allocation would require county-level PUMS microdata on Medicaid participation by age, household type, and employment status — not available in published tables. - Floor —
Math.max(1, ...)applied to both endpoints. Prevents zero-displays for very small counties (Keweenaw, Luce, Schoolcraft) where the allocation would otherwise round to zero. - Display — Every county shows a low–high range. No point estimate is displayed. The "Projected loss low" and "Projected loss high" column headers each carry a "modeled range — not a point estimate" qualifier directly beneath the sort button. The page subtitle and the amber callout both repeat "Exposure is not disenrollment" and link to this page.
Limitations
- ACS enrollment values reflect means-tested public coverage (broader than administrative Medicaid enrollment) and include PHE-era enrollment peaks in the 5-year window. County absolute values should not be cited as Michigan Medicaid enrollment counts; they are used only for proportional allocation.
- The work-requirements-only scope does not model separate coverage loss from redetermination changes ($63B / 700,000 nationally per CBO pub. 61570) or provider tax restrictions. Those provisions would increase projected coverage loss beyond this range. MDHHS (August 2025) estimates >500,000 Michiganders at risk across all P.L. 119-21 Medicaid provisions combined.
- For small counties (Keweenaw, Luce, Ontonagon, Schoolcraft), ACS C27007 5-year 2023 coefficients of variation for Medicaid enrollment estimates are high (typically 20–50% for populations under 5,000). Keweenaw County (1,173 ACS-estimated enrollees) has a particularly high coefficient of variation; its county-level projection should be treated as illustrative only. Exact margins of error are available at data.census.gov table C27007. The
Math.max(1, ...)floor prevents zero-display for these counties. - Urban Institute's March 2026 projection reflects enacted-law work requirement implementation. It incorporates Urban's assumptions about compliance rates, state administrative capacity, and partial exemptions. Michigan's actual outcomes will depend on MDHHS implementation decisions, legal challenges, and federal rule-making not yet finalized.
Why we publish this
No public source currently provides county-level estimates of P.L. 119-21's Medicaid impact for Michigan. Urban Institute publishes a Michigan state range. CBO publishes national figures. Individual counties — which are responsible for implementing the law through local MDHHS offices, Federally Qualified Health Centers, and managed care contractors — have no public baseline for planning. accessmi.org publishes this model not as an advocacy document but as a planning input: a rough, sourced, range-bounded estimate that a county health department, hospital, or journalist can use to frame a question, not answer it. The methodology is public, the uncertainty is explicit, and the model will be updated when better data becomes available. Making the uncertainty visible is more useful than leaving this question unanswered in public planning records.
Change log
Every change to this methodology is recorded here with date and reason.