BLS salary data: Bureau of Labor Statistics wages, OEWS wage data and how to turn percentiles into pay bands
BLS salary data comes from the Occupational Employment and Wage Statistics survey, which asks about 1.1 million US establishments what they actually pay. It publishes a mean wage and the 10th, 25th, 50th, 75th and 90th percentiles for roughly 830 occupations, in every state and about 530 metro and nonmetro areas, for free. The wages are straight-time gross pay, so they behave like base pay: annual bonuses and benefits are not in the number.
That makes it the strongest free foundation for a pay band in the United States, and the easiest data in the market to misread. Below: the full specification, what the wage figure does and does not contain, the six limitations that decide whether you can use it for a given role, the averaging mistake BLS explicitly warns against, and the percentile-to-band arithmetic worked through end to end. Price a role on the right while you read.
- P25
- $0
- P50 ยท Median
- $0
- P75
- $0
Suggested posted range
Built from public U.S. BLS OES wage data (May 2024 release), adjusted for market, seniority and company stage by the multipliers published in our methodology.
Get bands for your whole team
Get startedWhat Bureau of Labor Statistics wage data actually is
Almost every US salary figure you have ever seen traces back to this survey, including the wage numbers in the Occupational Outlook Handbook and in most careers websites. It is a cooperative effort: BLS funds it and sets the methodology, and the state workforce agencies collect most of the data. Here is the whole specification in one place.
| Attribute | May 2025 estimates |
|---|---|
| What it is | The Occupational Employment and Wage Statistics (OEWS) program, an employer survey of wage and salary workers in nonfarm establishments run by BLS with the state workforce agencies. |
| Occupations covered | About 830 occupational categories under the 2018 Standard Occupational Classification. |
| Geographies | The nation, every state, DC, Guam, Puerto Rico, the US Virgin Islands, and about 530 metropolitan and nonmetropolitan areas. |
| Industry detail | About 410 industry classifications at the national level, using the 2022 NAICS. |
| Sample | About 1.1 million establishments, drawn from state unemployment insurance files. |
| Collection | Two semiannual panels a year of roughly 186,000 to 189,000 establishments each, contacted in May and November. |
| What one release contains | Six semiannual panels collected over three years. The May 2025 estimates combine May 2025, November 2024, May 2024, November 2023, May 2023 and November 2022. |
| Sampled employment | 84.7 million unweighted across the six panels, roughly 55 percent of total national employment. |
| Response rate | 66.2 percent by establishment, 67.2 percent by weighted sampled employment. |
| Statistics published | A mean wage plus the 10th, 25th, 50th (median), 75th and 90th percentiles, both hourly and annual. |
| Annual wage basis | Annual figures assume a full-time, year-round schedule of 2,080 hours. |
| Current release | The May 2025 estimates, published May 15, 2026 (USDL-26-0725). |
| Price | Free. There is no license, no participation requirement and no login. |
Source: BLS news release USDL-26-0725, Occupational Employment and Wages, May 2025, technical note, and the OEWS frequently asked questions at bls.gov/oes.
What the wage figure includes, and what it leaves out
This is the single most consequential thing to understand about BLS salary data, and the part almost every secondary write-up skips. OEWS wages are straight-time, gross pay, exclusive of premium pay. That definition has a precise inclusion list and a precise exclusion list.
Included in the number
- Base rate
- Cost-of-living allowances
- Guaranteed pay
- Hazardous-duty pay
- Incentive pay, including commissions and production bonuses
- Tips
Excluded from the number
- Overtime pay
- Severance pay
- Shift differentials
- Nonproduction bonuses (the annual bonus most office employees get)
- Employer cost for supplementary benefits
- Tuition reimbursements
Read the two lists together and a rule falls out: an OEWS wage is base pay plus sales-style incentive pay, and nothing else. Commissions and production bonuses are in. The annual or discretionary bonus that most salaried office employees receive is a nonproduction bonus, and it is out. Every dollar of benefits cost is out.
The exclusion that trips up round-the-clock operations is shift differentials. A published median for a role has the night premium stripped out of it, so benchmark the base rate against the survey and treat shift differential pay as a separate layer on top. Compare a differential inclusive hourly rate to an OEWS median and the night crew will look overpaid while its base sits under market.
Benefits are the larger gap, and there is one slice of them federal tax law puts an exact price on for you. Group-term life coverage above $50,000, a company car used personally and health coverage for a non-dependent partner all become imputed income with a published valuation method. That figure is not a full benefits load, but it is a real, defensible dollar amount sitting entirely outside the OEWS wage you are benchmarking against.
So when you compare a BLS figure to one of your own employees, compare it to their base salary. Comparing it to their total cash package will make your pay look generous when it is merely competitive, and that error compounds quietly across a whole structure. If you want to hold total cash constant instead, split the band and the incentive target apart first, the way a salary structure separates grade from variable pay.
One more definitional detail with real consequences: annual wages are computed on the assumption of a full-time, year-round schedule of 2,080 hours. For occupations that are genuinely paid annually but do not work 2,080 hours, such as teachers, pilots and flight attendants, employers report an annual rate directly. For some entertainment occupations only an hourly wage is reported at all.
Who is not in the survey at all
OEWS covers wage and salary workers in nonfarm establishments. It excludes the self-employed, owners and partners in unincorporated firms, household workers and unpaid family workers. It excludes most of agriculture, keeping only logging and the support activities for crop and animal production, and it excludes private households entirely. Federal coverage is limited to the Postal Service and the executive branch, while state and local government are fully covered. If a large share of the people doing a job work for themselves, the OEWS picture of that job is partial by construction. Whether the people doing your work are employees in the first place is the 1099 vs W2 classification question.
How to turn OEWS percentiles into a pay band
A percentile wage is a boundary, not an average. The 25th percentile is the line below which a quarter of workers in that occupation are paid. The 50th, the median, is the line that splits the occupation in half. BLS publishes this worked example on its own percentiles page, and it is a clean one to build a band from.
| Percentile | 10th | 25th | 50th (median) | 75th | 90th |
|---|---|---|---|---|---|
| Hourly wage | $11.00 | $15.00 | $20.00 | $24.00 | $29.00 |
| Annual wage | $22,880 | $31,200 | $41,600 | $49,920 | $60,320 |
The most common construction is to take the 25th percentile as the band minimum and the 75th as the maximum, which here gives a band of $31,200 to $49,920. The spread, meaning maximum divided by minimum minus one, is 60 percent. That is a wide band, appropriate for a job family where people stay and grow for years, and too wide for a role with a narrow scope. How to choose that width is covered in how wide a salary range should be.
Now the detail that trips people up. The median is $41,600, but the arithmetic middle of $31,200 and $49,920 is $40,560. A band built from percentiles is not symmetric around the median, because wage distributions lean right. You have to decide which number is your midpoint and then be consistent, because the midpoint is the denominator of every compa ratio you will ever calculate. Anchor on the median at $41,600 and someone paid $45,000 has a compa ratio of 1.08. Anchor on $40,560 and the same person is at 1.11, and nothing about their pay has changed.
Whichever you pick, write it down in the structure documentation. A pay band is only defensible if you can explain in one sentence where the minimum, midpoint and maximum came from, and half of the arguments about compa ratio are really arguments about an undocumented midpoint.
Choose the geography before the percentile
OEWS publishes the same occupation for the nation, the state and the metro area, and the three numbers can differ by 30 percent or more. Benchmark against the market where the person actually does the work, not where you are incorporated. For a fully remote hire, pick one policy and apply it to everyone: national, or the employee's own metro, or a small number of defined tiers. The tradeoffs are in geographic pay differentials.
Six limitations that decide whether you can use it for a role
None of these make BLS salary data bad. They make it specific. Knowing which one bites on a given job is the difference between a benchmark you can defend in a compensation committee and a number you pulled off a website.
It has no concept of level
OEWS does not contain information about pay according to the level of work performed. BLS says so directly, and runs a separate survey, the National Compensation Survey, for that purpose. So OEWS can tell you what software developers earn in Denver, but not the difference between your Engineer II and your Engineer IV.
It has no employer size cut
BLS notes that workers in large establishments generally earn more than workers in small ones, and that OEWS does not publish a breakdown by establishment size. A 40-person company benchmarking against a figure that includes 40,000-person employers is reading a number shaped partly by companies nothing like it.
It is roughly a year behind, and blended over three
The May 2025 reference period was published on May 15, 2026. Each release also blends six panels stretching back three years, with the older panels adjusted forward to the current reference date by a modeling procedure. The number is a three-year blended estimate centered on a date about a year old, which matters most in fast-moving job families.
It is not a time series
BLS warns that OEWS estimates are not designed for comparing employment or wages over time. Sample rotation, classification changes and methodology revisions all move the numbers independently of the labor market. Charting last year against this year to prove a raise budget is a misuse of the data.
Its scope has real holes
The survey excludes the self-employed, owners and partners in unincorporated firms, household workers and unpaid family workers. It excludes most of agriculture, and private households entirely. Federal government coverage is limited to the US Postal Service and the executive branch. There is no county-level data either: the smallest geography is a metropolitan or nonmetropolitan area.
Some estimates are simply missing
Where an estimate does not meet OEWS publication standards it is not released. Small occupations in small areas are the usual casualties, which is exactly the combination a niche role in a mid-sized metro tends to fall into.
The level problem is the one that matters most in practice, and it has a workable answer. Because OEWS mixes every seniority level of an occupation into one distribution, the percentiles themselves carry the level information: the 25th percentile of a software developer distribution is largely junior people, the 90th is largely senior people. So instead of asking BLS for a level it does not have, map your levels onto percentile targets and hold that mapping steady across the whole job family. That is exactly the method behind market pricing a job.
Do not average percentile wages, and here is the proof
This is the most common analytical error in homemade benchmarking, and BLS warns about it explicitly: BLS does not recommend calculating averages of percentile wages. Averaging percentiles fails to account for the different sizes and different wage distributions of the occupations you are combining, so the result does not equal the real percentile for the combined group.
BLS demonstrates it with human resources workers, which is a convenient example for anyone reading this page. The broad group 13-1070 contains three detailed occupations. Take the 90th percentile annual wage of each, average them, and compare to the published 90th percentile for the group itself.
The simple average of the three lands at $122,280, nearly six thousand dollars below the published figure of $128,180, because it gives Farm Labor Contractors, an occupation with 410 people in it, the same weight as HR Specialists, an occupation with 917,460. Weighting by employment gets much closer at $128,292, but still does not match, because weighting corrects for size and not for the shape of each wage distribution.
The practical rule: when you need a number for a group of jobs, pull the published estimate for that group. Build averages out of raw wages, never out of percentiles.
| Occupation | Employment | 90th pct |
|---|---|---|
| HR Specialists (13-1071) | 917,460 | $126,540 |
| Farm Labor Contractors (13-1074) | 410 | $86,860 |
| Labor Relations Specialists (13-1075) | 64,590 | $153,440 |
| Simple average of the three | n/a | $122,280 |
| Employment-weighted average | n/a | $128,292 |
| Published, HR Workers (13-1070) | 982,460 | $128,180 |
Source: BLS, OEWS national cross-industry estimates, May 2024, reproduced from the BLS percentile wage page.
The May 2025 release fixed the biggest problem with benchmarking senior roles
For years, anyone trying to benchmark a well-paid role in OEWS hit the same wall. Instead of a number, high percentiles for high-paying occupations returned a footnote: this wage is equal to or greater than $115.00 per hour or $239,200 per year. The survey collected wages into intervals, and the top interval was open-ended, so OEWS could not produce a specific estimate for any percentile that fell inside it. For a director-level or specialist role, the 75th and 90th percentiles were often exactly the ones you needed and exactly the ones that came back censored.
Two changes fixed most of it. OEWS moved to its current model-based estimation approach, and from the May 2022 panels onward it began using reported wage rates where employers supplied them rather than relying only on interval assignments. Together those let the program determine the distribution above the top interval far more accurately.
Beginning with the May 2025 estimates, OEWS started publishing specific wage estimates for many percentiles that would previously have carried only that footnote. If you tried OEWS for senior roles a few years ago, found it top-coded and wrote it off, that judgment is now out of date. Go and look again.
A note on the 2025 shutdown
The lapse in federal appropriations from October 1 to November 12, 2025 meant the May 2025 OEWS panel needed extra collection and processing time once funding resumed. BLS reported that the response rate for that panel still came in within the normal range, and that no additional modifications to OEWS methodology or procedures were needed as a result. The release arrived on schedule in May 2026. It is worth knowing because it is the kind of thing a skeptical executive will ask about, and the answer is reassuring.
BLS salary data against a paid salary survey
We build on OEWS, so treat this table as interested rather than neutral. It is still honest, including about the places a purchased survey genuinely wins. Most companies under 200 people should start with public data and buy a survey only when a specific job family forces it.
| Dimension | BLS OEWS | Commercial salary survey |
|---|---|---|
| Cost | Free to anyone, with no participation requirement. | Priced per job, per module or per year, and most vendors quote rather than publish. |
| How the sample is built | A stratified probability sample drawn from state unemployment insurance files, so participation is not self-selected. | Employers who choose to join and submit their own data, which is a self-selected group by construction. |
| Level of work | None. One occupation code covers every seniority level in it. | Usually the main reason to buy one. Levels, scopes and job families are the product. |
| What the wage covers | Straight-time gross pay, so effectively base plus commissions and production bonuses. | Typically base, target bonus, actual bonus and often equity, reported separately. |
| Freshness | Annual release, reference period about a year old, blended across three years. | Often semiannual or quarterly, with a shorter lag. |
| Job matching | You match your role to an SOC code yourself, and the codes are broad. | Vendor job descriptions to match against, which is slower but far more precise. |
| Antitrust posture | A published federal statistic. Reading it involves no exchange of competitively sensitive information between employers. | An exchange of employer pay data through a third party, which is lawful but is exactly the activity the 2025 DOJ and FTC guidelines address. |
The last row is the one that has changed most recently. The Department of Justice withdrew its 1996 safe harbor for information exchanges on February 3, 2023, and the DOJ and FTC guidelines issued on January 16, 2025 replaced the earlier HR guidance without restoring it, while addressing data shared through an algorithm or a third party. Salary surveys remain lawful, and plenty of companies should still buy one. What is gone is the assurance that meeting four structural conditions puts the exchange beyond challenge. Reading a published federal statistic raises none of that, because there is no exchange to join. The detail sits on salary survey providers and in how to conduct a salary survey.
Where to get the data, in the order most people should try
01
Occupational profiles
One page per occupation with the definition, national employment and wages, then the same by state, area and industry. Fastest route for a single role.
02
The OEWS tables
National, state, metropolitan and industry tables in HTML and Excel at bls.gov/oes/tables.htm. Use these when you are pricing a whole structure at once.
03
The full data files
Every published estimate in one download. Large, and worth it if you want to filter on the group field to avoid double counting broad and detailed occupations.
04
State workforce agencies
Most states republish OEWS for their own areas, sometimes with extra local cuts. Useful when you need a nonmetropolitan area explained.
A warning about the download route, straight from the BLS documentation: the files contain both broad occupation groups and the detailed occupations inside them, so summing across everything double counts employment. Filter on the group field and keep one level or the other. And if you are combining an occupation across several areas, remember section 05: do not average the percentiles.
What our software does with all of this is the boring, repetitive part: pull the right occupation for the right geography, apply your chosen percentile targets consistently, and produce a salary band with the source and date recorded next to it, so that when someone asks where the number came from a year later there is an answer. The how it works page walks through the sequence.
What to do once you have the market number
A benchmark is an input, not an outcome. It becomes useful when it turns into a band, a structure and, in a growing number of states, a range you are legally required to publish in the job ad.
There is also a use for this data that has nothing to do with hiring. When an S corporation owner has to justify the salary they pay themselves, the IRS factor in play is what comparable businesses pay for similar services, and published wage statistics are how that gets priced. The government expert in the leading case built his figure from an industry survey the same way, which S corp reasonable salary walks through step by step.
Salary bands
Turn a market percentile into a minimum, midpoint and maximum you can defend.
Salary structure
Grades, spread and overlap across every job at once, not one role at a time.
Job posting salary ranges
What has to appear in the advertisement once the band exists.
Pay transparency by state
Which states require a posted range, the headcount thresholds and the penalties.
BLS salary data questions people actually ask
What is BLS wage data?
BLS wage data usually means the Occupational Employment and Wage Statistics program, or OEWS. It is a survey of about 1.1 million establishments that reports what employers actually pay across roughly 830 occupations, for the nation, every state and about 530 metropolitan and nonmetropolitan areas. It publishes a mean wage and five percentiles for each one.
Where does the BLS get its salary data?
From employers. BLS and the state workforce agencies draw a sample from state unemployment insurance files and ask those establishments to report employment and wage rates by occupation. It is payroll information supplied by the business, not survey responses from workers, which is why it is not affected by what employees think they are worth.
Is BLS salary data accurate?
It is accurate for what it measures, and the sample is far larger than any private source: 84.7 million sampled jobs, about 55 percent of national employment, with a 66.2 percent response rate. The risk is not accuracy but fit. It has no level detail, no employer size cut, and a reference period about a year old.
How often is BLS wage data updated?
Once a year. Data are collected in two panels each May and November, and a single set of estimates is published each spring. The May 2025 estimates were released on May 15, 2026. Each release folds in six panels collected over the previous three years, adjusted forward to the current reference period.
Does BLS salary data include bonuses?
Only some. OEWS wages are straight-time gross pay and include incentive pay such as commissions and production bonuses. They exclude nonproduction bonuses, which is the category most annual and discretionary office bonuses fall into. Treat an OEWS figure as base pay plus sales-style incentives, not as total cash compensation.
Does BLS wage data include benefits?
No. The employer cost for supplementary benefits is explicitly excluded, along with overtime pay, severance, shift differentials and tuition reimbursement. If you need the cost of benefits, that is a different BLS product: the Employer Costs for Employee Compensation series from the National Compensation Survey.
What is the difference between the mean wage and the median wage?
The mean is the average: total wages divided by the number of employees. The median is the 50th percentile, the boundary where half the workers earn more and half earn less. The mean is pulled upward by a few very high earners, so in most occupations it sits above the median. For pay bands, use the median.
Can I use BLS data to set salary ranges?
Yes, and it is a defensible starting point because it is public, free and drawn from a probability sample. The usual approach is to anchor the midpoint on the median or the 75th percentile for the right geography, then set a minimum and maximum around it. Adjust for level and company size, because OEWS captures neither.
What percentile should I pay at?
There is no universal answer, only a policy you should write down. Paying at the median means half the market pays more, which is fine for roles you can refill easily. Paying at the 75th percentile is normal for hard-to-hire and revenue-critical roles. What matters is choosing a target percentile per job family and applying it consistently.
Does BLS have salary data by state and metro area?
Yes. Estimates are published for every state, the District of Columbia, Guam, Puerto Rico, the US Virgin Islands, and about 530 metropolitan and nonmetropolitan areas. There is no county-level data, so for a role outside a metro area you use the nonmetropolitan area estimate for that part of the state.
Is BLS salary data free?
Yes, completely. The tables, the occupational profiles and the full downloadable data files are all published without charge or registration at bls.gov/oes. It is the only comprehensive source of regularly produced occupational wage information for the US economy that costs nothing to use.
Why is there no BLS estimate for my occupation in my area?
Because it did not meet OEWS publication standards, usually because too few establishments in that area employ the occupation for the estimate to be reliable or confidential. When that happens, step up one geography at a time: metro area, then state, then national, and note which level you used.
Can I average BLS percentile wages across occupations?
BLS specifically recommends against it. Averaging percentiles does not account for the different sizes and wage distributions of the occupations being combined, so the answer will not match the published percentile for the group. If you need a figure for a combined group, use the published estimate for that group instead.
Does BLS salary data include self-employed people?
No. OEWS covers wage and salary workers only. The self-employed, owners and partners in unincorporated firms, household workers and unpaid family workers are all outside the survey. That matters if you are benchmarking a job family where a lot of the market works on contract.
Figures on this page describe the May 2025 OEWS estimates published on May 15, 2026, and the percentile example reproduces BLS's own published illustration. BLS updates the estimates annually, so check bls.gov/oes for the current release before quoting a number in a compensation decision.
Wagelist
Skip the spreadsheet, keep the source
Enter a job title and a location and get a band built from public federal wage data, with the occupation, geography, percentile and release date recorded alongside it. Pricing starts at $99 a month with no annual contract. If you are still choosing between public data and a purchased survey, the compensation management software roundup compares the ten tools most teams shortlist.