Methodology and source transparency

The four categories of evidence

Every published number on this site falls into one of four buckets:

1. CMS marketplace rate filings. Every premium on this site is computed from the Centers for Medicare & Medicaid Services Health Insurance Exchange public use files: the rate, plan attributes and service area files for the federal exchange, plus the equivalent state-based exchange files, all in the public domain. The method is described in full below.
2. Employer coverage surveys. Employer-sponsored premium splits (what the employer pays, what the employee pays, COBRA continuation costs) come from the KFF Employer Health Benefits Survey, the annual survey of employer plans. These are labelled on the pages that use them and are the only premium figures on the site that are not computed from the CMS filings.
3. HHS Federal Poverty Level (FPL) tables. The annual FPL publication from the Department of Health and Human Services anchors every subsidy-cliff and Medicaid-eligibility calculation on the site. The 100 percent, 138 percent, 250 percent, and 400 percent FPL thresholds used in subsidy maths come from this single source.
4. CMS, IRS, and HHS published limits. The ACA out-of-pocket maximums ($10,600 self-only / $21,200 family for 2026) are set by CMS/HHS in the annual Notice of Benefit and Payment Parameters, not by the IRS. HSA contribution limits come from IRS Rev. Proc. 2025-19, and the ACA affordability percentage (9.96 percent for 2026) from IRS Rev. Proc. 2025-25. These are read directly off the agency PDFs.

Where a figure on a calculator page does not trace to one of these four sources, it is either a multiplier we derived ourselves (in which case the calculation is shown inline) or an industry convention we’ve labelled as such.

How the premium figures are computed

Nothing on this site is a hand-typed premium. The whole table is rebuilt from the plan year 2026 CMS filings by one script, and here is exactly what it does.

1. Eligible plans. Individual-market, non-dental, on-exchange standard plans, excluding child-only and catastrophic plans and excluding cost-sharing-reduction variants.
2. Benchmark premium. In each county, silver plans are ranked on the essential health benefits portion of premium (the rate multiplied by the plan’s EHB percentage, the basis premium tax credits use under 26 CFR 1.36B-3) and the second lowest is taken. Where the two lowest tie, the tied value is the benchmark, following the CMS rule that has applied since plan year 2018.
3. Which plans count in a county. Only plans whose service area covers that county, priced at the rate filed for that county’s geographic rating area. Rating areas come from the CCIIO published definitions for each state.
4. State and national averages. County figures are weighted by county plan selections from the CMS open enrollment file. That file covers HealthCare.gov states only, so state-based marketplace states are weighted by county population from the Census Bureau instead. The basis is shown against each state in the by-state table.
5. Verification. The same script was run against the plan year 2025 filings and compared with the state benchmark and lowest-cost-plan figures CMS published for that year. Across 31 state benchmarks and 93 lowest-cost-plan figures, the mean absolute difference was 0.16 percent and no single figure differed by more than 1 percent.

One known departure from the CMS method: where a silver plan does not cover pediatric dental, CMS adds a stand-alone dental premium before ranking. This site does not, which is part of the sub-1 percent residual above. Snapshot taken 2026-09-06.

Re-verification cadence

• CMS Exchange public use files: re-derived annually, when CMS publishes the next plan year’s files (typically late summer, ahead of open enrollment on 1 November).
• Employer survey data: refreshed when KFF publishes the next annual Employer Health Benefits Survey.
• HHS FPL tables: pulled annually in January when the new FPL is published.
• IRS Rev. Procs: pulled within 7 days of each new publication.

The footer on every page shows the “Last verified” stamp for the dataset the page depends on, so you can see at a glance whether you’re looking at the latest publication or a snapshot from the previous cycle.

What this site explicitly does not do

• We do not collect personal health information, income figures, or eligibility data. Calculators run entirely client-side; inputs never leave your browser.
• We do not quote you a plan or refer you to a broker. The CMS HealthCare.gov shop tool is the right place to get a binding quote; we link to it from every calculator output.
• We do not reproduce insurer claims data, and we do not project. Every premium figure is computed from published rate filings for the plan year named on the page.
• We do not name carriers or recommend plans. The goal is to translate the published dollar figures into a decision framework you can use; the binding shop is the marketplace itself.
• We do not invent provenance. Where a figure is a hand-typed estimate from an industry survey rather than a primary publication, the page labels it as such.

Corrections

If you spot a figure that disagrees with the source it cites, please write to [email protected] with the page URL and the contradicting source. Corrections are typically shipped within 48 hours and the page footer is updated with the correction date.

Reader questions go to questions@howmuchishealthinsurance.com. They are read by a person, and where a question exposes something this site does not explain, the answer goes onto the page it came from.