On a chartbook pulled 2026-08-08, unemployment insurance reached 28.07 percent of unemployed workers in the United States in 2025 — Florida 8.33 percent against Minnesota 49.86 percent, and no federal floor on coverage is in force

What is happening?

In 2025 about 28 percent of unemployed workers in the United States received a regular unemployment insurance payment. The value is 28.07 percent as read from the US Department of Labor unemployment insurance chartbook on 2026-08-08, and 26.5 percent as read from the same chart by the Federal Reserve Bank of Cleveland in a report published 2026-07-01. Both readings describe the same completed calendar year. Roughly three of every four unemployed workers received nothing from the program built to cover them.

The national average hides the spread that is the actual problem. For 2025, pulled 2026-08-08, the lowest jurisdiction was Florida at 8.33 percent and the highest was Minnesota at 49.86 percent, with a median of 23.86 percent across the 52 jurisdictions the chart covers. The highest is about six times the lowest. For 2024 the pair was Kentucky at 8.88 percent and Minnesota at 56.29 percent, a ratio of 6.34 to 1. Florida replaced Kentucky at the bottom of the table between those two years.

The decline is long and it is national. On chart a12 of the same chartbook the rate was 45.7 percent in 1950, 43.9 percent in 1980, 30.3 percent in 2010 and 27.3 percent in 2024. Coverage roughly halved across seventy years. The pandemic emergency programs lifted it to 74.8 percent in 2020, which is the proof that the ceiling is not structural, and it fell to 25.1 percent in 2022 — below the 2019 level of 28.1 percent.

Whose problem is this?

The body that writes the rules and the body that publishes the numbers are not the same body, and neither one is answerable for the result.

Where does this problem end?

The boundary matters here because the program already exists in every state and only the reach of it differs.

What is the state now, and what should it be?

Now

Needs a new measurementthe target state: no source opened here names a target recipiency level, for the nation or for any jurisdiction. The full text of the leading federal reform bill was searched for the terms recipiency, share of unemployed, take-up and takeup, and returned zero occurrences. That bill sets floors on inputs — weeks of benefit, wage replacement, maximum benefit — and names no target for the outcome. A measure with no target cannot be missed, so nobody is ever in violation of it.

How big is it?

About 5.3 million people. In an average month of 2025 the United States counted 7.315 million unemployed workers aged 16 and over on the not seasonally adjusted series LNU03000000, retrieved from the public statistics interface on 2026-08-08. Applying the recipiency rate for the same year leaves 5.26 million uncovered at the higher reading of 28.07 percent and 5.38 million at the lower reading of 26.5 percent. For 2024 the same chain gives 6.761 million unemployed against a recipiency of 26.75 percent, or 4.95 million uncovered.

The band is not a confidence interval. Both ends use the same denominator and the same indicator, and the entire width is produced by which day the chartbook was queried. The uncertainty here is instability in the source rather than sampling error.

The count has holes on both sides and they run in opposite directions. The 2025 denominator is an eleven month mean, because October 2025 is absent from the returned series and no cause for the absence was established. It is a stock rather than a flow — it counts people unemployed in an average month, while the number who pass through an uncovered spell across a full year is larger and no source opened here publishes it. It leaves out people who stopped searching, since leaving the labor force removes a person from the statistic instead of moving them to the uncovered side. It leaves out self-employed and contract workers, who are usually outside the program and therefore never appear as insured unemployed at all. And it treats everyone inside the covered share as covered regardless of amount, although Ohio replaces 43 percent of prior wages and Pennsylvania 50 percent.

One further caution about the arithmetic itself. The numerator is an administrative count of insured unemployed and the denominator is a household survey, so the two halves come from different systems and are not guaranteed to be coherent. Minnesota in 2020 returns 105.5 percent on this chart, a share above one hundred, and the chart displays it without comment.

Under what conditions does it arise?

1. The federal government funds administration and sets conformity conditions, and none of those conditions is a floor on coverage. States write their own earnings thresholds, disqualification rules and benefit duration. A jurisdiction that narrows any of the three faces no federal consequence, so the cheapest lawful design is a very narrow one.

2. Narrowing benefits and lowering employer taxes are the same arithmetic. Benefit outlays feed the employer reserve ratio, which feeds contribution rates. Kentucky House Bill 4 of 2022 carries both moves in one instrument — Section 7 replaces the 26 week entitlement with a 13 step scale bottoming at 12 weeks, and Section 3 shortens the employer reserve ratio look back from twelve calendar quarters to four.

3. The saving is local and immediate while the cost is diffuse and deferred. Lower employer contributions are visible now to an organized in-state constituency. Unemployed residents with no income are not organized as a constituency, and the weakened automatic stabilizer does not arrive until the next downturn.

4. Most of the gap is not written in any statute. The Council of Economic Advisers found that differences in estimated potential eligibility explain 11 percent of the variation across states, rising to 29 percent when actual maximum durations are used. Pennsylvania and Ohio both cap at 26 weeks and stood about twenty percentage points apart in 2025. The remainder lives in earnings thresholds, administrative friction and non-application.

Repealing every duration cut in the country would therefore leave most of the spread standing.

What has been tried?

Every instrument that raised coverage was either temporary or voluntary with money attached. The one instrument that would be permanent and binding has not moved in more than a year, while the instruments that lower coverage are permanent, binding and still being enacted.

What was found?

Why is it still unsolved?

Supply disparity — the same event, losing a job through no fault of your own, produces protection or nothing depending only on which state line it happened inside.

The first movement is that the federal lever exists and is unused. Washington funds administration and attaches conditions to the program through the tax code, and none of those conditions is a floor on how many weeks a jurisdiction must pay, how much prior work it may demand, or what share of its unemployed it must reach. The pending bill would exercise exactly this lever, which is the cleanest available evidence that the lever is there. Authority that exists and is not used is a choice inside an institution rather than a gap between institutions.

The second movement is that the incentive runs the wrong way at the level where the decision is made. Benefit outlays and employer contribution rates are connected by the reserve ratio formula, so paying fewer claimants for fewer weeks lowers a visible in-state tax. Kentucky House Bill 4 carries the duration cut and the reserve ratio recomputation in the same enacted instrument, which is as concrete as this coupling gets in primary text. Kentucky recipiency stood at 20.5 percent in 2019 and 8.88 percent in 2024. No source opened here joins the statute to the series, and the join is left to the reader rather than asserted.

The third movement is the one that keeps the pattern stable, and it is the reason repeal alone would not fix it. Most of the disparity is not in the statutes at all. Formal eligibility explains eleven percent of the variation across states, and about three quarters of unemployed workers never file a claim, of whom more than half believe they are ineligible. The source that establishes that belief declines to say how many of them are correct, which is precisely the number that decides whether this is an information problem or a statutory one. Underneath all of it sits a fourth condition that is quieter than the other three. Recipiency is published every year and owned by no one. No agency is accountable for it, no statute names a level, and the leading reform bill does not use the word.

What observation would mean it is solved?

Candidates — (a) the gap between the highest and lowest jurisdiction narrows sharply and stays narrow across several years (b) the national recipiency rate rises and holds above some named level (c) a federal floor on duration and eligibility is enacted and takes effect.

(a) alone is weak because the gap can close downward. If broad states narrow toward the narrow ones, the ratio improves while coverage falls everywhere. A convergence measure has to be read together with the level, and neither one alone tells you which direction the convergence came from.

(b) alone is weak because no level has ever been named. There is no target in any source opened here, so any threshold chosen after the fact would be chosen to be met. The national rate also moves with the composition of unemployment rather than with policy — a downturn that hits covered industries raises it without any rule changing.

(c) alone counts paper. A floor on weeks and earnings thresholds reaches the eleven to twenty-nine percent of the variation that formal eligibility explains. It does not reach the seventy-four percent of unemployed workers who never file, and a rule that nobody hears about does not raise take-up. The three have to be read together, and (c) has to be read against whether filings actually rose.

What is it connected to?

Fills with researchstate unemployment trust fund solvency and employer contribution rates, worker classification and the treatment of platform work, the design of automatic stabilizers in a recession, and administrative burden as a general feature of means-tested and insurance programs. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• No source names a target. Not the Department of Labor, not any state statute, not the Council of Economic Advisers, not the advocacy brief. The leading federal reform bill was searched in full text for recipiency, share of unemployed, take-up and takeup and returned zero occurrences. Nobody is failing to hit a number, because no number exists.
• The chartbook carries no data vintage. Values for the completed year 2025 differ by up to 5.8 percentage points between a pull made before 2026-07-01 and one made on 2026-08-08 — Massachusetts moves from 38.6 percent to 44.39 percent, Minnesota from 52.7 percent down to 49.86 percent. There is no revision notice, no preliminary or final flag and no as-of stamp on the chart. A citation of this indicator without a pull date is a citation of an unstable number.
• Two charts in the same chartbook disagree. Chart a12 gives 27.34 percent for 2024 and chart a13 gives 26.75 percent for the same year. Neither page reconciles them and neither acknowledges the other.
• Nobody splits denied from never applied. The recipiency rate merges a system that rejects people with a system that people do not approach. The same agency publishes denial and appeals data in the same reporting system, and no source opened here cross-tabulates the two.
• The take-up finding stops exactly where it matters. Fifty-five percent of non-applicants did not apply because they believed themselves ineligible, correctly or incorrectly, and no source resolves the split. If most were wrong the problem is information and is cheap to fix; if most were right it is statutory and expensive.
• No source attributes any state rate change to a specific statutory change. The Kentucky fall from 20.5 percent in 2019 to 8.88 percent in 2024 is visible in the published series and unattributed in every text opened here. The link has to be assembled by the reader from two separate documents.
• The 2026 value is presented as a year. The chart returns 26.80 percent for 2026 with no indication that only about seven months of it exist, and claiming is seasonal.
• The scorecard for the 2009 incentive payments is missing. The department page lists which states received Full, One-Third or Two-Third payments and publishes no dollar amounts per state, no total disbursed against the USD 7 billion authorized, and no evaluation of whether recipiency in adopting states subsequently rose.

See the evidence

Read directly and end to end — the enacted Kentucky statute, the full text of H.R. 4439, the Cleveland Fed report, the archived Council of Economic Advisers paper, and the three chartbook pages together with the statistics interface that supplied the unemployment level. Those are primary documents, and the evidence_tier: secondary declaration reflects the document as a whole rather than each row: the take-up figures, the maximum-weeks column and the oversight framing are all second hand because the hosts holding them refused automated retrieval. Where the sources disagree the disagreement is left visible rather than resolved — the advocacy brief gives nearly 59 percent for Minnesota in 2024 against 56.29 percent on the chart it cites, while its national and Kentucky figures match closely; the same chartbook returns 27.34 percent and 26.75 percent for 2024 on two different charts; and the 2025 state values shift by as much as 5.8 percentage points depending only on the date of the pull, which is why every figure above carries one. Two claims are arithmetic performed here rather than printed anywhere: the affected-population product, and the statement that the highest jurisdiction is about six times the lowest. This is a Path A output (research-based definition), so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range 5,260,000–5,380,000 As of 2025, average month

Sensitivity The width of the band is not statistical uncertainty. Both endpoints use the same denominator and the same indicator, and the entire spread of about 120 thousand people is produced by which day the chartbook was queried, because values for the completed year 2025 changed silently between a pull made before 2026-07-01 and one made on 2026-08-08. What this number fails to count: people who stopped searching for work, since leaving the labor force removes a person from the household survey rather than adding them to the uncovered side; self-employed, contract and platform workers, who are generally outside the program by design and so never appear as insured unemployed at all; and the difference between being denied a claim and never filing one, which this single figure merges. The limit in the opposite direction is that the number overstates nothing about hardship but understates the annual total badly, because it is a monthly stock rather than an annual flow and no source opened here publishes the flow. A further caution on the arithmetic itself: the numerator of the recipiency rate is an administrative count and the denominator is a household survey, so the two halves come from different systems and need not be coherent, and the same chart returns 105.5 percent for Minnesota in 2020 without comment.

Regional breakdown Recipiency rates are published for all 52 jurisdictions, but the matching state level unemployment counts were not retrieved in this round, so no state figure can be computed by the same chain. Splitting the national figure by population share is not permitted and would in any case be wrong here, since the whole point of the problem is that the rate varies from 8.33 percent in Florida to 49.86 percent in Minnesota.