The state jobs report
September state jobs.
Beyond the average.
September’s clearest monthly payroll changes were in just three states. The national unemployment rate cannot tell the whole local story.
New Jersey and South Carolina added payroll jobs in September 2026, while Wisconsin lost them. Those were the only statistically significant monthly state changes in the latest BLS report. Understanding what that means starts with separating people, jobs and uncertainty.
The important word is “significant”
September’s state jobs report does not describe an economy in which nothing happened outside three states. It describes an economy in which the clearest monthly payroll signals were unusually concentrated. In the Bureau of Labor Statistics release published October 20, New Jersey gained 14,900 nonfarm payroll jobs, South Carolina gained 12,500 and Wisconsin lost 13,700. Changes in the other 47 states and the District of Columbia did not meet the report’s test for statistical significance. The September state employment release makes that distinction explicit.
That is a more useful finding than a simple list of winners and losers. A state estimate can rise or fall numerically without providing sufficiently strong evidence of a change under the agency’s statistical test. “Essentially unchanged” is therefore not a claim that employers added exactly as many jobs as they removed. It is a statement about what the available evidence can reliably distinguish from sampling and estimation uncertainty.
BLS uses a 90 percent confidence level when identifying significant changes. The test concerns a specific estimate and its uncertainty, not whether every business or worker experienced stability.
The national rate conceals a wide local range
The national unemployment rate was 4.2 percent in September. South Dakota’s rate was 2.0 percent, followed by North Dakota at 2.2 percent. The District of Columbia was at 5.6 percent. California and Michigan, at 5.1 percent and 5.0 percent respectively, were the two states whose rates BLS identified as significantly above the national rate. Twenty states had rates significantly below it, while 28 were not appreciably different from it.
Those comparisons are not interchangeable with a raw numerical ranking. BLS assesses whether a state’s difference from the national rate is statistically significant. A reader who sorts every displayed rate and draws a hard line at 4.2 percent would be answering a different question from the one in the release. Rounded rates and state-specific uncertainty both matter. The published table is a guide to measured differences, not a scoreboard that establishes a universal best place to work.
A low rate is not a count of vacancies. A small state can have low unemployment and few employers in a specialized occupation; a large state can have a higher rate and more relevant opportunities. Skills, pay and actual openings still require separate evidence.
Five declines, with different starting points
Five states had statistically significant month-to-month declines in their unemployment rates. Maryland moved from 4.1 percent in August to 3.9 percent in September, and New Hampshire from 2.8 percent to 2.6 percent. Florida fell from 4.5 percent to 4.4 percent, Nevada from 4.8 percent to 4.7 percent, and Pennsylvania from 3.7 percent to 3.6 percent. These are changes in percentage points, not percent changes in the number of unemployed people.
The starting point changes the interpretation. New Hampshire’s decline occurred from a rate already well below the national figure. Nevada’s left its rate numerically above the national figure, although BLS did not classify Nevada among the places significantly above it. The common direction of a monthly movement does not mean the states have converged to the same labor-market conditions.
Some numerical changes at least as large as these declines did not pass the significance test. That is possible because uncertainty differs between estimates. The rounded rate alone does not convey that difference.
New Jersey, South Carolina and Wisconsin are not the whole country
The three significant monthly payroll changes point in different directions. New Jersey’s gain was 0.3 percent of its previous-month payroll level. South Carolina’s was 0.5 percent. Wisconsin’s decline was 0.4 percent. The percentages reveal why comparing only job counts can mislead: South Carolina added fewer jobs than New Jersey but recorded a larger proportional increase.
All three monthly payroll comparisons are seasonally adjusted. The adjustment addresses recurring calendar patterns, including those associated with schools and tourism, but does not turn the estimate into an administrative count of every job.
Nor should these three changes be added together and presented as the national result. The BLS technical note explicitly cautions against aggregating state payroll estimates into a substitute national employment series. State and national estimates are developed independently, and state estimates are not forced to sum to the national total. Combining them accumulates state-level errors. Selecting only statistically significant changes introduces an additional problem: it discards the remaining estimates entirely. The three-state list identifies the strongest signals in this release, not a complete national jobs ledger.
A strong September does not erase New Jersey’s earlier weakness
New Jersey provides a useful test of how much a headline can leave out. Its labor department’s October 15 release independently reports the same 14,900 monthly payroll increase and a September employment level of 4,377,400. But the state’s total nonfarm payroll gain over the year was just 3,600. A large latest-month improvement can coexist with a much smaller twelve-month advance because earlier gains and losses remain inside the longer comparison. New Jersey’s detailed September report supplies the industry breakdown behind that contrast.
Private education and health services added 7,200 jobs in September, professional and business services added 7,100, and leisure and hospitality added 5,500. Manufacturing and other services each added 1,000, while construction added 500. Those gains were partly offset by losses of 5,100 in trade, transportation and utilities, 1,100 in information, 100 in financial activities and 1,100 in the public sector. Together, those reported movements reconcile to the state’s 14,900 total.
The annual pattern is more revealing still. Private education and health services added 30,000 jobs over the year, while private employment overall rose only 6,900. Other industries therefore offset much of that strength. The report lists annual losses of 11,200 in trade, transportation and utilities and 4,500 in information, among others. For an applicant, the relevant lesson is not “New Jersey is hiring” in the abstract. It is that a strong aggregate month may be driven by industries very different from the one in which that applicant works.
The annual jobs leaders look different
Looking back twelve months changes the map. Seven states had statistically significant increases in payroll employment, while the District of Columbia had a significant decline. The remaining 43 states were essentially unchanged under the BLS test. Texas recorded the largest significant numerical gain, 154,300 jobs, followed by California with 129,300 and North Carolina with 67,000.
The full set of significant annual gains also included South Carolina, up 55,200; Minnesota, up 36,800; Louisiana, up 36,000; and Utah, up 26,300. These are changes between September 2025 and September 2026, not jobs created during September alone. The time label is not a small detail: a reader comparing Texas’s annual 154,300 gain with New Jersey’s monthly 14,900 gain would be comparing unequal observation windows.
California illustrates why two positive-sounding labels need not travel together: it had a significant annual payroll gain and an unemployment rate significantly above the national rate. Jobs at establishments and unemployed residents are different measures; neither establishes that finding suitable work is easy.
Percentage growth gives smaller states their proper scale
South Carolina led the significant annual percentage gains at 2.3 percent, ahead of Louisiana at 1.8 percent and Utah at 1.5 percent. Texas’s larger numerical gain represented 1.1 percent, while California’s represented 0.7 percent. North Carolina grew 1.3 percent and Minnesota 1.2 percent. These figures tell a different, complementary story from the absolute job counts.
The arithmetic is straightforward. A gain is divided by the starting payroll level. South Carolina rose from 2,386,100 jobs in September 2025 to 2,441,300 in September 2026. The increase of 55,200 was smaller than Texas’s, but larger relative to its initial employment base. Texas moved from 14,307,000 to 14,461,300. The state’s much bigger starting economy means an increase can be large in people’s everyday understanding of the word while more modest in percentage terms.
Absolute counts describe the scale of the payroll gain; percentage growth compares momentum relative to the starting base. Neither ranking, on its own, is a recommendation to relocate or expand a business.
Why D.C.’s payroll decline can coexist with a lower jobless rate
The District of Columbia lost 25,800 payroll jobs over the year, a 3.5 percent decline, the only statistically significant annual payroll loss identified in the release. Its unemployment rate nevertheless fell from 6.6 percent to 5.6 percent over the same period. The pairing is striking, but it does not require assuming that either measure is wrong.
The unemployment figures follow people where they live. Payroll figures follow the location of the establishments employing them. Consider a Maryland resident who works at an establishment in the District. That person belongs to Maryland’s resident labor-market count, while the payroll job belongs to the District’s establishment count. A commuter crossing a state line makes the distinction concrete. Changes in resident employment, labor-force participation and the geography of jobs need not line up one-for-one.
The counting units differ, too. A person with two payroll jobs can appear on two establishment payrolls but counts as one employed person in the resident framework. The release alone does not establish a single cause for the District’s annual decline.
A falling unemployment rate needs a second question
The unemployment rate is the number of unemployed people divided by the civilian labor force. The labor force includes employed and unemployed people. Under the definitions used in the report, unemployed people generally must have actively looked for work within the previous four weeks and be available to work; people on temporary layoff expecting recall need not be searching to be counted as unemployed.
That definition explains why a lower rate is not, by itself, proof that more residents found work. In a deliberately simplified example, suppose a labor force contains 100 people, 95 employed and five unemployed. The unemployment rate is 5 percent. If one unemployed person leaves the labor force while employment stays at 95, four unemployed people remain in a labor force of 99. The rate becomes about 4.0 percent, even though no additional person became employed. This is an arithmetic illustration, not a claim about what happened in any state in September.
Read labor-force and unemployed-person counts alongside the rate. A decline accompanied by rising employment is different from one caused by a smaller labor force. The arithmetic identifies the next question, but reasons such as retirement, school or caregiving require separate evidence.
The data are preliminary, and the measurement systems differ
The state unemployment estimates are not simply the answers of every resident collected in a monthly census. The Local Area Unemployment Statistics program uses models drawing on sources that include the Current Population Survey, establishment employment data and state unemployment insurance information. The technical note describes a process of controlling estimates to broader geographic totals, as well as smoothing and seasonal adjustment. Those methods are designed to produce coherent, usable local estimates from information that is incomplete and noisy.
Payroll estimates come from the Current Employment Statistics program. They refer to people on establishment payrolls who receive pay for any part of the pay period that includes the twelfth of the month. The estimates are subsequently revised as more information becomes available and are benchmarked annually, principally using employer tax reports covered by unemployment insurance laws. The first release is therefore an estimate with a defined method and revision process, not a final payroll audit.
When comparing releases, preserve the data vintage, reference month, geography and seasonal adjustment. An earlier published estimate may later be revised without having been inaccurately reported at the time. Residents and payroll jobs should never be silently substituted for one another.
How to use the report without asking it to do too much
For a job seeker, the most useful sequence is state, industry, occupation and actual vacancy. The state report can identify broad conditions and industry movements. It cannot establish that a particular employer is recruiting today or that a listed position pays enough to offset a longer commute. A strong statewide health-services gain, for example, does not automatically improve the prospects of an applicant looking only for manufacturing work in a distant county.
For a small business, the resident labor force and payroll series answer different operational questions. Resident measures describe the pool of people participating in the local labor market. Payroll measures describe the employment activity of establishments. Neither replaces information about qualifications, schedules, wages or the practical distance workers can travel. A statewide average can conceal a metropolitan labor market that crosses borders, as the District example makes especially clear.
For readers following the wider economy, employment should also be kept separate from production. Our analysis of September industrial output and its uneven manufacturing components examines what factories, mines and utilities produced. This report examines labor-market measures. More output is not mechanically the same as more payroll jobs: hours, productivity and the mix of activity can differ. Read together, the reports provide complementary evidence, not two interchangeable scores for the economy.
What to watch next
The next layer of detail is geographic. BLS has scheduled the September metropolitan-area employment and unemployment release for October 28 at 10 a.m. Eastern time. That release should help readers move beyond statewide averages toward the labor markets in which people actually look for jobs and commute. The October state employment and unemployment report is scheduled for November 20 at 10 a.m. Eastern time. Those are release dates, not predictions of what the data will show.
Three checks will make the next comparison more informative. First, see whether September’s estimates have been revised before treating the old and new numbers as a continuous story. Second, compare both the latest monthly movement and the twelve-month change, preserving each time window. Third, look for corroboration in the industries and geographic areas relevant to the question being asked, rather than assuming that a statewide headline describes every workplace.
September offers neither proof of a nationwide hiring boom nor a universal freeze. The clearest monthly payroll signals were concentrated in three states, while unemployment levels and annual growth varied widely. The headline identifies where to begin; definitions and industry detail keep the local reading honest.
Comments
Post a Comment