When BLS and Private Jobs Data Disagree: A Policy Analyst's Playbook

From January 2023 through December 2025, BLS and ADP both put U.S. employment growth at 2.4 percent; Revelio Labs put it at 2.9 percent.

Yet when the Bureau of Labor Statistics goes dark during a federal shutdown, analysts reach for those private series and find state and regional numbers that do not match public benchmarks. Boston Fed researchers found the alignment largely holds nationally but weakens sharply for New England.

The practical question for policy analysts is not which source is correct. It is when a private substitute is safe to use, and when a mismatch signals a methodology trap rather than a real labor market shift.

Payroll Surveys, Payroll Records, Profiles and Postings: What Each Source Actually Measures

Why does ADP differ from the BLS jobs report? The two products count different populations with different methods, a common pitfall in public sector vs private sector administrative employment. BLS Current Employment Statistics (CES) is a sample survey of roughly 121,000 nonfarm establishments¹ that asks employers how many workers were on payroll during the pay period containing the 12th of the month. ADP's National Employment Report tallies actual payroll records from more than 500,000 client companies covering over 26 million employees, roughly one in six U.S. workers.³ Revelio Public Labor Statistics models employment from professional profile records rather than surveying employers or processing payroll.nn## BLS Current Employment Statistics (CES)nCES samples approximately 121,000 businesses and government agencies, selected to represent nonfarm employers by state, industry, and size. It counts civilian workers on payroll in the reference period containing the 12th, and only employers reporting for two consecutive months enter the month-to-month change calculation.nn## BLS Job Openings and Labor Turnover Survey (JOLTS)nJOLTS draws on about 21,000 establishments² and measures job openings, hires, and separations across all 50 states and the District of Columbia. It is benchmarked to CES employment data but is a demand and turnover measure, not an employment count.nn## ADP National Employment ReportnADP uses anonymized payroll records from client companies, with pay periods interpolated to approximate the week containing the 12th. Its sample includes more than half a million companies and over 26 million employees, and it is benchmarked annually to QCEW tax records.⁴nn## Revelio Public Labor Statistics and Indeed Job Postings IndexnRevelio builds estimates from professional profile data, while Indeed's daily Job Postings Index tracks job postings, not filled positions. JOLTS and Indeed both measure openings and labor demand, so neither should be treated as interchangeable with CES or ADP payroll counts.nnBecause these sources differ in population, method, and timing, disagreement is expected rather than a sign that one source is wrong, a key caution for public sector labor statistics.

What the Boston Fed Found: National Alignment, Regional Drift

How closely do public and private labor market data align in New England, and where can an analyst inspect the underlying numbers instead of a headline?

Start with the primary brief

The core source is Boston Fed New England Public Policy Center Regional Brief 2026-3, "How Closely Do Public and Private Labor Market Data Align for New England Compared with the Nation?" by Nick Hall and Osborne Jackson. Search that exact title and author pair on the Boston Fed website and the NEPPC publications page¹ to retrieve the complete New England Revelio growth figure, state-level results, and regression alignment coefficients for ADP, Revelio, and Indeed versus BLS. This Federal Reserve labor market research indexes each series to January 2023 using data from January 2023 through December 2025. Its central finding is that private data show greater regional drift from BLS measures in New England than nationally.

Compare BLS benchmarks and private series directly

Go to BLS.gov for the official benchmarks: LAUS, QCEW, and OEWS state and national series. BLS defines payroll employment as nonfarm payroll employment and reports it separately for New England and the nation. When comparing those measures with ADP, Revelio Labs, and Indeed Hiring Lab data, check each source's methodology, release notes, and definitions of employment and wage growth.

Verify method and numbers before relying on them

Once you locate the brief, read its tables, appendix, and references. Look for replication files or author contact information, and cross-check any cited figures against original ADP Research Institute, Revelio Labs, Indeed Hiring Lab, and Boston Fed publications. Ask how each private provider allocates workers across state lines and handles seasonal adjustment; those are among the reasons the brief gives for the regional gap. For regional context, pair the brief with state labor market information offices, local workforce boards, and relevant professional or industry associations that report area-specific labor market data.

Methodological Traps: Seasonal Adjustment, State Attribution and Coverage

The regional misalignment is less about one source being wrong and more about which geography a vendor thinks it is measuring. For an analyst working at the state or regional level, that difference can change policy decision-making.

Why the series drift at the state level

The Boston Fed brief¹ points to seasonal adjustment methods and the way multi-state employers allocate jobs to individual states as likely drivers, but the authors describe the evidence as limited. Treat these as hypotheses to test with a private vendor, not established defects. A hospital system headquartered in Boston with facilities in Rhode Island and Connecticut could report all workers under Massachusetts, inflating that state's count while undercounting the others.

What each series covers

Coverage gaps compound the problem. The Bureau of Labor Statistics payroll survey includes federal, state, and local government employees. ADP covers only the payrolls of its client firms. Revelio builds estimates from public profiles, and certain worker classifications may be excluded depending on how the vendor handles self-reported or incomplete data. Area-specific measurement error may also play a role, but its exact contribution is still unclear.

One question to ask before using a private series

Ask the vendor whether a job is assigned to the headquarters location or the work site. The answer often explains a large share of state-level divergence and tells you whether the series is safe for subnational data-driven policymaking.

Source Comparison Matrix: Coverage, Timeliness, Revisions, Cost and Access

This matrix compares the five labor market series on coverage, release lag, revision behavior, geography, and access. Cells marked Not published reflect details not confirmed in the published sources used for this comparison. Replicability is often limited for private series because the underlying records are proprietary; when a public methodology does not make the series reproducible, the relevant cell is marked Not published.

Private labor market data track the nation well, but their alignment with public data weakens sharply across New England and its states.

Revisions and Benchmarking: Why the First Print Is Not the Final Word

Analysts often trade timeliness for accuracy, and labor market data make that tradeoff explicit. The first estimate arrives fast, but it is not the final word.

Two Stages of Revision in Public Data

BLS's establishment survey revises monthly as more survey responses arrive. The first preliminary estimate is based on reports received by the last Friday of the reference month. A second preliminary follows the next month, and a final sample-based estimate lands two months after the initial release.¹ Once a year, that sample-based history is benchmarked to unemployment insurance tax records, which can revise several years of data.¹

The scale of those revisions matters. The March 2025 annual benchmark, reported in the CES Benchmark Revision 2025, lowered total nonfarm employment by 861,000 jobs, or 0.5 percent. Over the prior ten years, annual benchmark revisions averaged 0.2 percent in absolute terms and ranged from 0.05 percent to 0.4 percent, according to the CES Frequently Asked Questions. Monthly first-to-third estimate shifts are smaller on average: since 2003, the average absolute revision has been about 51,000 jobs, with a signed average of 9,000.²

Private Series Revision and Rebasing

ADP and Revelio also revise and rebase their series, but their revision schedules and historical lookback windows are not laid out on the same public calendar as BLS. A rebased private series can be hard to compare historically unless the analyst keeps vintage files and re-runs comparisons after each benchmark.

Practical rule: label every number as preliminary or revised, and repeat any trend or regression analysis after the annual benchmark.

The Data Infrastructure Problem: Response Rates, Funding and Shutdowns

Can private jobs data stand in when the Bureau of Labor Statistics goes dark during a shutdown?

When the gold standard goes quiet

The 43-day federal shutdown that began October 1, 2025 showed why the "gold standard" label still matters. BLS cancelled the October Employment Situation because the underlying Current Population Survey was never collected; September's release was delayed more than six weeks, according to the BLS Statement on the Federal Government Shutdown. November CPS response rate fell to 64.0 percent, below the prior 12-month average of 68.4 percent and just under the previous low of 64.9 percent in June 2020.¹ Establishment survey collection rates held at 73.9 percent and 73.8 percent for October and November, but only because of extended collection periods.

Why the strain is structural

The shutdown was not an isolated event. BLS inflation-adjusted budget fell from $810 million in 2010 to $636 million in 2025, a 20 percent drop, as documented in the Economic Policy Institute report Consequences of Austerity: How Reductions in BLS Funding Threaten the Credibility of Our Statistics. Staffing is projected to fall to 1,851 positions in FY26, down 10 percent from 2,058 in FY24.² The agency has already cancelled products, samples and services in 2025.²

What private data can and cannot cover

Private data can partly replace public jobs numbers during a shutdown at the national level; ADP, Revelio and Indeed continued through the 2025 gap. But for states and regions, the Boston Fed analysis shows private series align less well, so substituting there carries more risk. Private vendors also depend on public benchmarks for calibration and seasonal adjustment, so a weakened BLS weakens the private alternatives too.

Related Articles

A Decision Rule for Analysts: When Private Data Is Safe and When It Isn't

The central tradeoff is simple: private labor market data offer speed, frequency and continuity when official series are delayed or interrupted, but official series remain the benchmark for geographic precision and methodological transparency. For policy analysts and public policy consultants, the decision rule is equally clear: private data can supplement or temporarily stand in for national monitoring, but using it for subnational formulas, budgets or economic development decisions carries materially higher error risk. The choice is not whether to use private data, but when and at what geographic level.

Separate tasks into three tiers

• National trend monitoring: Use ADP, Revelio or Indeed alongside BLS to track direction and turning points. This is the lowest-risk setting, and it is where private data performed best in the Boston Fed comparison.
• Regional directional reads: Treat private state or regional figures as a signal, not a measurement. The Boston Fed found New England private data aligned less closely with BLS than the national private data did, so disclose the uncertainty and avoid quoting levels as facts. A regional private figure should be treated as a range, not a point estimate.
• Formula or allocation decisions: Stay with official BLS data. Funding formulas, revenue estimates and program eligibility rules need the consistency, benchmarking and legal defensibility that only official statistical series provide.

Keep BLS as the anchor

Use private series as a cross-check on direction, never on level. If ADP or Indeed moves up while BLS moves down, that is a reason to investigate coverage or seasonal adjustment, not a reason to reject the official number. Treat private series as early warning indicators, not as substitutes for the official level. When official data are unavailable, private data can fill the gap temporarily, but label it explicitly as a proxy.

Document the choice

Before the memo reaches a decision-maker, state which source you used, why you used it, and what bias or error it might introduce. That one-sentence disclosure turns a fragile data substitution into a defensible analytical judgment.

Triangulation Checklist and Vendor Due-Diligence Questions

• Compare indexed levels

  Look at cumulative indexed changes since a common start month, not just the latest monthly percentage move.
• Match the base period

  Confirm every series is set to the same reference point before comparing direction or magnitude.
• Require two independent sources

  Treat a single private series as provisional until a second source shows the same directional pattern.
• Ask about seasonal adjustment

  Request the vendor's method and whether it differs from the BLS approach used in public data.
• Ask how jobs are assigned to states

  Check whether the vendor allocates employment by worker residence, establishment location, or a modeled proxy.
• Ask about sample and profile coverage

  Clarify which employers, industries, workers, or postings are included and which are underrepresented.
• Ask how revisions are handled

  Find out when historical values are restated and whether the vendor publishes revision studies.
• Ask if the method can be replicated

  Preference goes to sources that document enough methodology for an analyst to reproduce key series.

What This Means for Workforce Boards, State Budgets and Economic Development

A local workforce board can choose between two signals for workforce development policy: a federal payroll benchmark built for formula allocations, or a fast private estimate built for market monitoring. They answer different questions, and at the state level they may not agree.

Where the numbers enter real decisions

In these federal-state partnership formulas, WIOA Adult and Youth allocations weight three one-third components: unemployment in Areas of Substantial Unemployment, excess unemployment, and disadvantaged residents.¹ Wagner-Peyser allocates 97% by formula, with two-thirds civilian labor force share and one-third unemployed share.¹ Allocations are protected within 90% to 130% of the prior-year share.¹ A small shift in LAUS unemployment can change which counties qualify as an ASU (at least 6.5% unemployment and 10,000 people) or a Labor Surplus Area (20% above the national average, bounded between 6% and 10%).²³

Hypothetical misreads

Illustrations, not real cases: A state revenue forecast built on an overstated regional payroll gain may inflate expected receipts. A workforce board that moves training funds after a postings spike may chase intent rather than actual hiring, especially when private regional data align less closely with public benchmarks.

Contingency planning

Pre-agree fallback sources such as QCEW, UI wage records, and LAUS.³ Set trigger points tied to missed revisions, shutdown gaps, or large public-private divergences. Disclose data limits plainly in board packets and budget memos. Use private data for early warning, not for final allocation changes.

Private data can substitute for national labor market monitoring, but for state and regional budget and formula decisions, substitution carries materially higher risk of error.

publicadministrationpolicy.org

Building Data Literacy for a Public Policy Career

Government data infrastructure is under visible strain, and analytical skill expectations now center on source evaluation, triangulation, and revision awareness rather than passive reliance on a single jobs report. In MPA or MPP degree policy analytics courses, these competencies are increasingly framed as everyday practice. A useful habit is keeping a source log: record the provider, access date, seasonal adjustment choice, and state attribution rule each time a dataset informs a decision. Another is tracking revisions, comparing an initial payroll print against later benchmark updates to see what changed and why. Treat vendor estimates from ADP, Revelio, or Indeed as corroborating evidence, not replacements, especially at the subnational level where alignment weakens.

These habits make policy analysts easier to evaluate. Public policy analytics roles value analysts who can explain methodology limits to budget staff, workforce boards, and elected officials. The policy analytics and public-sector career path pages on this site offer deeper guidance on building these habits during a career change.