Calibrate ACA marketplace PTC usage by income/FPL band

[daphnehanse11]

Context

The ACA local impacts work uses CMS Marketplace Open Enrollment public use files. The compact county/district data we use on the page has aggregate enrollment, premium, APTC/CSR, income-band, and metal-level counts, but it does not include a county-level cross-tab of metal selection by income.

The fuller CMS 2026 State, Metal Level, and Enrollment Status PUF does include metal-level enrollment split by household income as percent of FPL. That table shows plan choice varies sharply by income/FPL, especially because cost-sharing reductions are tied to silver plans.

For PolicyEngine calibration, though, the target is not necessarily to predict every metal level. The practical modeling goal is to estimate how much premium tax credit (PTC/APTC) is actually used. We do not need to model gold separately if the current model treats silver and gold similarly for PTC use: both generally have premiums high enough to use the full available PTC. The more important margin is bronze or other low-premium selections, where the selected premium may be below the available PTC and some credit can go unused.

Sources:

• CMS 2026 Marketplace Open Enrollment Period Public Use Files: https://www.cms.gov/data-research/statistics-trends-reports/marketplace-products/2026-marketplace-open-enrollment-period-public-use-files
• 2026 OEP State, Metal Level, and Enrollment Status Public Use File
• 2026 Public Use Files Definitions

Evidence from the 2026 CMS PUF

Using HC.gov-platform states in the 2026 State/Metal/Enrollment PUF, a useful recode for the PTC-used question is bronze versus silver/gold:

Household income (% FPL)	Plan selections	Bronze / lower-premium proxy	Silver + gold / full-PTC proxy	Other
<100%	132,237	33.5%	64.3%	2.2%
100-150%	8,793,788	28.2%	71.8%	0.0%
150-200%	2,574,719	46.7%	53.2%	0.1%
200-250%	1,480,945	67.5%	32.4%	0.1%
250-300%	867,002	74.1%	25.7%	0.2%
300-400%	1,021,383	75.4%	24.3%	0.3%
400-500%	203,976	68.9%	27.0%	4.2%
>500%	357,059	69.3%	26.3%	4.5%
Other/unknown	340,288	64.0%	34.5%	1.5%

The largest practical contrast is still between low-income CSR-eligible enrollees and higher-income enrollees:

• 100-150% FPL: 28.2% bronze versus 71.8% silver/gold.
• 300-400% FPL: 75.4% bronze versus 24.3% silver/gold.

That is roughly a 47 percentage point shift away from silver/gold and toward bronze as income rises from 100-150% FPL to 300-400% FPL.

Suggested data/modeling update

Consider updating the ACA marketplace enrollment calibration to reflect PTC usage by household income/FPL band. This probably does not need to match every CMS metal and income band exactly.

A coarser calibration would likely capture most of the behavioral signal, for example:

• <=150% FPL
• 150-200% FPL
• 200-300% FPL
• 300-400% FPL

• 400% FPL / unknown

Potential target categories could be even simpler than metal level:

• bronze / lower-premium plan proxy, where some available PTC may be unused
• silver or gold / full-PTC-use proxy
• other / unknown

This avoids adding a gold-specific modeled choice that we do not need for the local impacts use case, while still preserving the large income gradient in PTC usage.

Notes / caveats

• These figures are plan selections, not necessarily effectuated enrollment.
• Silver/gold as a full-PTC-use group is a modeling simplification for the PTC-used question, not a claim that metal level alone mechanically determines APTC use in every case.
• The State/Metal/Enrollment PUF results above are for HC.gov-platform states. State-based marketplace reporting may need separate treatment or a fallback if comparable cross-tabs are unavailable.
• The county-level PUF has income-band counts and metal-level counts separately, but not a metal-by-income cross-tab at the county level, so this is probably best treated as a national/platform calibration target rather than a county-specific one at first.