Maryland's seasonally adjusted unemployment rate fell to 4.1% in August 2026, matching the national rate exactly.1 Through the first eight months of the year the state added 20,100 nonfarm jobs, growth of 0.7%, ahead of the national pace of 0.4%. After eighteen months of federal contraction, Maryland is adding jobs faster than the country.
By every headline measure, the state has recovered.
It has not. Maryland's unemployment rate improved in part because its labor force shrank. Whether displaced workers found work is not something the unemployment rate can tell you.
Set the three levels side by side. Between August 2025 and August 2026, Maryland's civilian labor force fell by 32,093. Employment fell by 23,857. The number of unemployed workers fell by 8,236.2 The second and third figures sum exactly to the first, because that is what the accounting requires.
The unemployment rate improved because unemployment fell while the labor force contracted more sharply. Fewer Marylanders were counted as unemployed, and fewer Marylanders were counted at all.
The participation rate tells the same story without the noise. It stood at 63.7% in August, down from 64.9% in January 2025, and it has not risen in a single month across that entire span. Nineteen consecutive months, eighteen consecutive declines or holds, no reversal. Over the 151 months BLS has published the series, August 2026 is the lowest rate.2
These figures do not establish why 32,093 people left the workforce. Retirement, out-of-state migration, discouraged search, demographic aging, and ordinary survey variation are all consistent with the same arithmetic, and the household survey cannot separate them. Nor, as it turns out, can any other tool Maryland uses.
The rate improved through a combination of lower unemployment and a smaller labor force, but public data cannot identify why people left the labor force, in the middle of the largest federal workforce contraction in the state's history.
One state, two labor markets
The recovery is real. It is simply happening to different people.
If you set the payroll series against each other, the shape is unmistakable. Maryland's total nonfarm employment has added 20,100 jobs since December 2025, yet still sits 33,700 below where it stood in January 2025. Federal employment over the same window fell from 162,000 to 132,700: a loss of 29,300 jobs, or 18.1% of the state's federal workforce, a contraction running at 1.6 times the national rate.3
Federal employment accounts for 87% of the payroll shortfall Maryland has still not recovered. The state is adding jobs in one column while the federal government column keeps subtracting them, and the two are not connected by any mechanism that moves a worker from the second to the first.
Exhibit 1 · Maryland's recovery has not reached the federal column
Change in payroll employment since January 2025, thousands, seasonally adjusted. Federal government in ruby, total nonfarm in ink.
Source: BLS Current Employment Statistics, series SMS24000000000000001 and SMS24000009091000001. Data through Aug 2026.
The contractor channel has already turned. Professional, scientific and technical services, the sector housing many federal contractors, fell from 280,400 in January 2025 to 270,800 a year later, but has added 2,400 jobs since December and now stands at 271,900.3 Contractors filed dislocation notices covering 1,234 Maryland positions through 2026, Amentum 418, Leidos 227, CBRE's facilities contract on the NIH campus 174, General Dynamics Information Technology 161, and eight smaller firms, of which 396 sit in Montgomery County.4 Those firms are hiring again, but federal agencies are not.
With that backdrop, Maryland's August job gains ran to leisure, hospitality, and to state and local government.
A senior NIH program officer does not become a hospitality worker. The jobs Maryland is creating and the jobs Maryland is destroying sit in different occupations, at different wages, in different counties. Maryland's aggregate recovery and Maryland's displacement event are running simultaneously, in the same state, to different populations, and the statewide statistics net them against each other.
The county data show the seam. In the third quarter of 2025, employment was down 3.4% year over year in Montgomery County, down 1.9% in Prince George's County, and up 0.2% in Baltimore City.5 Employment fell in seven of Maryland's eight largest counties. Statewide, it fell 1.3%.
What displacement actually costs
The canonical finding in the displacement literature is Jacobson, LaLonde and Sullivan's study of high-tenure Pennsylvania workers displaced in the early 1980s: earnings losses of roughly 25% of expected earnings, still present five years after separation.6 Three decades of subsequent administrative-data work has narrowed the range to 15–30% without overturning the result. The losses were largest under three conditions: long tenure with the former employer, a contracting former industry, and a local labor market with declining employment.
The exposed federal cohort may face all three conditions, but the available public data do not establish their prevalence or the resulting earnings loss. These are long-tenure workers in an industry that shed 287,000 positions nationally over calendar 2025, a decline exceeded only by the demobilizations of 1945 and 1946 across the entire eighty-seven-year history of the series3, living in the counties where employment fell hardest. This is the profile the literature identifies as having the worst and most durable earnings outcomes, and a falling unemployment rate tells you nothing about whether that has been avoided.
There is a natural objection to all of this. An aging state sheds participants, and a federal workforce with this much accumulated tenure has an unusually large population of annuity-eligible workers for whom retirement beats searching for a job. Maryland has roughly 153,000 households already drawing federal retirement income.7
The age data cuts against it, though not decisively. Among Marylanders aged 25 to 54, people who are not retiring, participation fell from 87.6% in 2024 to 85.9% in 2025. Over the same period, national prime-age participation was flat. That is the first annual decline in Maryland's prime-age series since at least 2016, and the largest single-year recorded.8
2024 was the high-water mark, annual moves in this state-level series routinely run a full point or more in either direction, and 85.9% still sits above Maryland's readings for 2016 through 2018 and 2021 through 2022. They also sit above the national rate. Some of the fall is reversion from the peak. State survey estimates for age subgroups are noisy enough that one year should not be used to explain too much.
What survives is direction, not magnitude: working-age participation moved down in Maryland while it held nationally, and the state's prime-age labor force shrank by 26,000 against a total decline on the order of 32,000. This is evidence against retirement as the whole story, not evidence for any particular alternative, and it does not identify where those workers went.
Openings are not absorption
Maryland employers advertised 127,000 job openings in December 2025, an openings rate of 4.4% against a national 4.0%. Across 2025, the state's openings never fell below 119,000, peaking at 144,000 in October.9 Advertised demand did not collapse during the federal contraction, but it did not hold either.
Set on one basis, advertised demand and actual hiring fell together. Openings averaged 152,200 a month in 2024 and 128,500 in 2025, a decline of 15.6%; hires fell 14.2% over the same comparison, from 1,167,000 to 1,001,000. The December-to-December reading on openings is the flattering one, because December 2024 already sat well below its own year's average. There is no divergence here to explain. What a vacancy count cannot tell you is whether the jobs being advertised are ones a displaced federal worker can take.
The scale of the churn underneath makes the point again. Those hires occurred in a year when Maryland's employment fell. Separations ran to roughly 1,020,000 against 1,001,000 hires, some two million hiring and separation events producing a net loss of nineteen thousand. These are transactions, not people; a worker hired and separated twice inside a year appears four times. But the ratio is the finding. A labor market can process two million events and still shrink. Volume is not absorption.
One series did not move at all. Layoffs and discharges stayed between 22,000 and 34,000 in every month of 2025, with no break in trend and no spike in any quarter of the federal contraction. The available layoff series shows no corresponding statewide surge. That absence does not identify the form the federal separations took, the series is statewide, model-based, and carries no sector detail, and several distinct mechanisms would leave the same flat trace. What it does establish is that the standard statistical signature of mass dislocation never appeared.
By December 2025, the evidence stops. BLS discontinued the monthly state JOLTS news release with that report; state estimates now publish once a year, covering the prior calendar year. At present, the Maryland series carries no 2026 observation at all, while the national series runs through July.9 The state's vacancy data for 2026 will not be public until well into 2027.
There is a second problem. Maryland's openings are the wrong denominator because the labor shed is regional. Federal employment fell 28,000 in the District and 26,200 in Virginia over the same window, against Maryland's 29,300, some 83,500 federal jobs gone across the three jurisdictions.3 A displaced program officer in Bethesda competes against displaced colleagues living across the DMV.
Maryland is monitoring the wrong point
It would be wrong to say the state is not watching. Maryland publishes weekly initial claims under Unemployment Compensation for Federal Employees, the program built specifically for separated federal workers, alongside regular state claims, initial claims by county, and weekly counts of open, closed and adjudicated claims.10
In the week ending 5 September 2026, Maryland recorded 25 UCFE initial claims. Continued UCFE claims the week before came to 317, against 23,497 insured unemployed statewide, former federal workers were 1.3% of Maryland's insured unemployment, in the state that had just shed 29,300 federal jobs.11 At 25 a week, the entire run of UCFE claims since January 2025 amounts to about 2,000, under a tenth of the jobs that actually disappeared.
The unemployment insurance system, the one instrument reporting weekly on federal workers, has barely registered the steepest federal contraction since 1946.
The national picture is the same shape: 398 UCFE initial claims in that week against a federal workforce down 336,000, and 5,478 continued claims, 1.6% of the national job loss. Across the DMV, 906 continued UCFE claims stand against 83,500 federal positions gone.
Exhibit 2 · The claims system is counting hundreds, not tens of thousands
Federal jobs lost January 2025 to August 2026 against continued UCFE claims, week ending 29 August 2026. Claims in ruby, jobs lost in ink.
Source: BLS Current Employment Statistics; DOL Employment and Training Administration, Unemployment Insurance Weekly Claims, 17 September 2026. Data through Aug 2026.
There are good reasons for this, and none of them is reassuring. Severance and deferred-resignation pay delay or preclude eligibility. Annuitants draw pensions, not benefits. Some people found work quickly, but some moved and filed elsewhere. Low claims are not evidence that displaced workers landed well. They are evidence that unemployment insurance is not the channel this shock ran through, and therefore that claims data cannot follow it.
Three systems should have caught a contraction this size, and each was built for a different shape of event. WARN does not cover federal agencies, so the 2026 dislocation log sees 1,234 contractor positions and none of the 29,300 federal ones. The shock did not appear cleanly in the unemployment rate, which fell as unemployment declined and the labor force contracted. The claims system logged twenty-five people last week. A shock does not have to be hidden to be missed. It only has to land in channels not properly monitored.
What the state does count, it counts carefully. Open claims peaked at 26,107 in the week ending 8 August and fell to 21,655 by 12 September, 4,452 claims closed in five weeks.10 Maryland knows precisely how many claims closed, but it does not publish why a single one of them did. A closed claim is consistent with a new job at comparable pay, a new job at half the wage, a move to Virginia or elsewhere, exhaustion of benefits, or giving up. Only one of those is a successful transition.
A year after the first wave, both Brookings and Capital News Service found the same gap: no systematic tracking of outcomes for Maryland's displaced federal workforce.712 What exists is case studies. One worker took contract roles across several states, earning more per hour with far less stability. One, at 57, retrained as a casino poker dealer. One pieced together part-time teaching and contract grant review, sustained by a spouse's income. Those are three different failures, out-of-state reemployment, occupational downgrade, underemployment, and not one appears in the unemployment rate. All three are counted as employed.
This is not an oversight. Maryland already runs Rapid Response, reporting WIOA performance measures, employment in the second and fourth quarters after exit, median earnings, credential attainment, matched against unemployment insurance wage records. The machinery exists. What it cannot do is start from a UCFE claims list that contains a few hundred names out of tens of thousands of separations. They must be identified some other way, from federal separation records, or from the wage records directly, before any of the existing performance machinery can see it at all.
The case for a transition register
The fix is a register that connects four decisions to evidence that can be tested for a defined population rather than for the state.
| Decision | Evidence required | Test of success |
|---|---|---|
| Who is in the cohort | Federal separation records by agency, occupation and duty station, matched into state UI wage records | Share of known separations located in the wage records, by quarter |
| Which occupations to target | Separation profile by occupation and tenure; demand sustained across reporting periods; wage and licensing comparability to the prior role | Share of the cohort employed in an adjacent occupation at 80%+ of prior earnings, twelve months after separation |
| Which employers to survey | Verified recurring demand, hiring capacity, willingness to recognize federal experience, location within the commuting shed | Hires from the cohort, retention after six and twelve months |
| Which programs to fund | Documented skill gap, time to completion against the worker's income runway, employer validation of the credential | Earnings in the fourth quarter after exit, against a matched comparison group |
| Which barriers are not training | Commuting distance, clearance transferability, licensing reciprocity, scheduling, childcare, income bridge | Transition completion rates by barrier type and county |
Four features distinguish this from what currently exists.
- The cohort is built from separation records rather than from claims, because the claims list holds a few hundred of tens of thousands.
- Every test is an outcome, not an output, no enrollment counts, no signed partnership agreements, no completion rates standing in for placement.
- Every test carries a comparison since a placement rate without a counterfactual measures the labor market, not the program.
- The unit of analysis is the separated worker, tracked through UI wage records across state lines, not the jurisdiction they happened to live in when they lost their job.
What states should want to know
Maryland's problem is not a weak labor market. In the aggregate, Maryland's economy is visibly strengthening, and faster than the nation's.
The problem is that the aggregate improved through a channel the state does not measure and report. Maryland's labor force shrank by 32,093 people over twelve months, its unemployment rate fell alongside that contraction, and no public dataset can currently say who those people were or where they went. This type of invisible displacement is how a region converts a temporary shock into a permanent reduction in earnings.
The regions that handle structural change well will be the ones that know, twelve months after a separation, where each worker went and what they now earn.
Maryland does not currently have those answers despite the state publishing weekly claims, county-level filings, and a program built specifically for separated federal workers. The problem is that this shock did not arrive as dislocation notices, because federal agencies file none; it did not appear cleanly in the unemployment rate, which fell as unemployment declined and the labor force contracted; and it did not arrive as claims, because twenty-five people filed last week. And the one data series that would show where displaced workers could transition has gone from monthly to annual and carries no 2026 reading at all.
A state absorbing the sharpest federal contraction in its history is navigating it with worse employment tooling than it had before the contraction started. None of what remains is pointed at the right people or tracking the right data. That, and not the headline jobs number, is the thing to fix.