Insights · Data guide
Groundwork research · Published October 19, 2026 · Data as of September 29, 2026 · 12 min read
The Occupational Employment and Wage Statistics survey (OEWS) is the Bureau of Labor Statistics' (BLS) count of what employers pay in about 830 occupations. It gives the wage at five points of the pay range for each occupation in every metropolitan area, nonmetropolitan area and state it can estimate, and the Department of Labor builds the four H-1B wage levels from the same figures. The May 2025 metro workbook holds 141,015 occupation rows for 393 metropolitan areas. This guide explains what a row holds, what the symbols and percentiles mean, how far to trust the numbers and how the wage levels come from them.
OEWS is a survey of employers. BLS sends it twice a year, in May and November, to a sample of nonfarm establishments drawn from the state unemployment insurance lists, and asks for the employees in each occupation and the wage range they are paid in. One release pools six of these semiannual panels, collected over three years: the May 2025 estimates draw on the panels from November 2022 to May 2025. Together the six panels sampled about 1.1 million establishments and 84.7 million jobs, 55% of U.S. employment, and 66.2% of the establishments responded. BLS updates the wages of the earlier panels to the reference date and released the May 2025 estimates on May 15, 2026.
The survey covers wage and salary workers. It leaves out the self-employed, owners and partners in unincorporated firms, household workers, unpaid family workers and farm workers.
One row is the estimate for one occupation in one area, across all industries and all kinds of ownership. The metro workbook (MSA_M2025_dl.xlsx) holds 150,023 rows of three
kinds, named by its O_GROUP column: 393 total rows with the figures for all occupations in an area, 8,615 major rows for the
groups of related occupations, and 141,015 detailed rows, the occupations themselves. Every row carries
NAICS code 000000, which stands for all industries, and ownership code 1235, which stands for the private sector and federal, state and local government together.
Detailed rows follow the 2018 Standard Occupational Classification (SOC). The metro workbook has 824 detailed codes, and BLS publishes a row only where it can estimate: New York-Newark-Jersey City, NY-NJ has 769 detailed rows and Guayama, PR has 71. Seven of the codes stand for several SOC occupations at once. They end in 0, and each holds the wage of its whole group:
For example 13-1020 stands where 13-1021, 13-1022 and 13-1023 would be, so a search for a farm products buyer finds the wage of all buyers and purchasing agents together. Here is one row of the table the site keeps: Software Developers (15-1252) in New York-Newark-Jersey City, NY-NJ has 121,000 jobs, a 10th percentile wage of $101,690, a 25th of $130,650, a median of $166,830, a 75th of $205,970 and a 90th of $224,590.
The workbook has 32 columns. These are the ones a reader needs, with how many of the 141,015 detailed rows in the metro workbook hold a number in each.
Where BLS holds back a number it writes a symbol in the cell. The workbook's own notes define four, and the metro workbook shows three of them:
The # rows are mostly doctors and surgeons: every wage is withheld on 207 rows and only the top wages on 5, and Anesthesiologists, Emergency Medicine Physicians and Cardiologists head the list. A number above the cap is a published number, because the symbol marks an estimate BLS withheld: 4,382 of the 90th percentile wages in the metro workbook reach $239,200 or more, the highest at $929,520. Read every symbol as a missing value before you sort, average or chart a column.
OEWS wages are straight-time gross pay. Base rates, cost-of-living allowances, guaranteed pay, hazardous-duty pay, commissions, production bonuses and tips are in. Overtime, severance, shift differentials, nonproduction bonuses and the employer's cost of benefits are out, so an OEWS wage sits below a total compensation figure that adds them.
An annual wage is the hourly wage times 2,080 hours, BLS's typical work year. The workbook bears it out: on all 131,342 rows with both medians, the annual median is within 0.077% of 2,080 times the hourly one. Occupations paid by the year that usually work fewer hours (teachers and school administrators) carry only annual wages, 7,847 rows in the metro workbook. Occupations paid by the hour with fewer hours (musicians and actors) carry only hourly wages, 262 rows, so the annual columns hold a * for them.
A percentile wage is a boundary. The median, the 50th percentile, separates the highest-paid half of an occupation's workers from the lowest-paid half, and the mean is their average. The five annual percentiles rise from the 10th to the 90th on all 139,104 rows that carry all five, and none is out of order.
On 81% of the rows, the relative standard error of the mean wage is under 5%, and on 0.5% (671 rows) it is 20% or more. Employment is thinner: 32% of the detailed rows that give employment stand for fewer than 100 jobs. Read a wage over a few dozen jobs as a rough guide, and use the employment column as the weight of evidence behind it.
Each release pools three years of panels, so a change in one year shows up slowly, and BLS advises against using the estimates to follow wages over time, because its classifications and methods also change. A wage covers every level of experience in the occupation together, which is why an entry-level offer and a senior one both sit away from the median. The survey records wages and jobs, with no information on the worker's age, sex or education.
BLS publishes the estimates in three area workbooks with the same columns: 393 metropolitan areas, 137 nonmetropolitan areas (the parts of each state outside every metropolitan area) and 54 states and territories. The May 2025 estimates use the metropolitan area definitions in OMB Bulletin 23-01. The nonmetropolitan workbook has 45,542 detailed rows for 751 occupation codes, and the state workbook 36,168 for 829.
A job title maps to a SOC code, and the code decides the wage. Two cautions come with that. An occupation named in a combination (the list above) shares one wage with its neighbors, and the occupations of hourly-only and annual-only pay appear in only one set of wage columns.
The site's wage tables load the annual wages of the detailed, cross-industry rows that have an annual median: 139,108 rows for 393 metropolitan areas and 818 occupation codes in the metro table, and 35,452 rows for 54 states and territories in the state table. Each count equals the rows in its workbook with a numeric annual median. A row with a * or # in the annual median is absent, and so are the 262 rows of hourly-only occupations. A * or # in one of the other wage columns of a kept row becomes a missing value.
The site holds no rows from the nonmetropolitan workbook (45,076 rows with an annual median), none for a single industry and none for the national estimates.
An employer filing a labor condition application for an H-1B worker attests to paying the higher of the wage it pays similar workers and the prevailing wage for the occupation in the area of work. When the Department of Labor uses a government survey for the prevailing wage, the statute (INA section 212(p)(4), 8 U.S.C. 1182(p)(4)) requires at least four wage levels commensurate with experience, education and the level of supervision. OEWS is that survey, and it publishes no levels. DOL's Office of Foreign Labor Certification builds them for each occupation and area: Level I (entry), II (qualified), III (experienced) and IV (fully competent).
DOL describes the method in its 2026 proposed rule (91 FR 15454). Since 1997 it has taken two figures from OEWS: the mean wage of the lowest paid third of an occupation's workers, about the 17th percentile of the distribution, and the mean wage of the upper two thirds, about the 67th. The statute's formula places the two intermediate levels a third and two thirds of the way between them, which puts Levels I to IV at about the 17th, 34th, 50th and 67th percentiles. DOL calls these approximations, because a mean over part of a distribution seldom lands on a percentile.
The filings we hold let us check the arithmetic. DOL's wage year from July 1, 2025 to June 30, 2026 is built on the OEWS estimates of May 2024, a year older than the workbook above, so the comparison is approximate. 599 occupation and metropolitan area pairs have a filing at every level, each level at one prevailing wage. In the median pair, Level II sits 33% of the way from Level I to Level IV and Level III 67%, the thirds the statute gives, and 513 and 507 of the pairs are within half a percentage point of them.
Against the OEWS percentiles, the levels climb as described. Among 6,364 pairs, Level I lies under the 25th percentile in 94%. Among 6,599, Level II lies between the 25th percentile and the median in 69%. Level III sits near the median: its middle pair is 2% above the OEWS median, and 59% of its 4,989 pairs lie above it. Level IV lies above the median in 93% of 4,937 pairs and above the 75th percentile in 25%.
The median prevailing wage at each level as a share of the OEWS median for the same occupation and area
DOL's guidance starts every job at Level I and compares the job's requirements for experience, education, training and special skills with those generally required for the occupation in O*NET. Points for requirements above the norm, added on a worksheet, raise the level (the Prevailing Wage Determination Policy Guidance).
Since the fiscal 2027 cap season, USCIS enters an H-1B registration in the selection one time at Level I, two at Level II, three at Level III and four at Level IV, by the highest OEWS wage level the offered wage reaches for the occupation and area (90 FR 60864, in effect from February 27, 2026). What each level pays in the filings is in our article on the wage levels.
On March 27, 2026 DOL proposed to set Levels I to IV at the 34th, 52nd, 70th and 88th percentiles of the OEWS distribution (91 FR 15454, docket ETA-2026-0001). Comments closed on May 26, 2026, and DOL estimated that the change would raise the average certified wage by about $14,000 a year. It was a proposal when this guide was last checked, on October 19, 2026.
The wage values themselves are in OFLC's wage files, which hold the four levels for each occupation and area for a wage year that runs from July to June (OFLC wage data downloads). Each wage year starts in July and is built on the estimates BLS has just published: July 2025 to June 2026 uses May 2024 and July 2026 to June 2027 uses May 2025. OFLC applies $115.00 an hour, $239,200 a year, where OEWS withholds a wage as too high to estimate.
OEWS supplies the numbers. OFLC turns them into four levels for each occupation and area, builds each July to June wage year on the estimates BLS has just published, and applies the wage of the level a job requires.
OEWS counts straight-time pay for every worker in the occupation at every level of experience, with nonproduction bonuses and benefits left out. A posting advertises one job at one employer, and OEWS describes every job in the occupation.
BLS publishes a row only where it can estimate, so a small area holds far fewer occupations than a large one (71 against 769). An occupation can also sit inside a combination, or carry only hourly or only annual wages.
BLS advises against reading the estimates as a time series, because each release pools three years of panels and the classifications and methods change.
The estimate sits at or above $239,200 a year and BLS withholds it. Treat the cell as missing, and read the occupation's other wages in the same row if they hold numbers.
OEWS wage data: the May 2025 estimates BLS published on May 15, 2026, read from the three area workbooks for the counts on this page and from the site's wage tables built from them. The
wage levels were compared on the filings with a wage level and a prevailing wage decided from July 1, 2025, matched by ZIP code to a metropolitan area and to the OEWS row for the same
occupation. BLS's technical notes and questions and answers, and the Federal Register and OFLC pages, were read on October 19, 2026. Built by scripts/build_insights.py.
Sources: BLS OEWS; technical notes for the May 2025 estimates; OEWS questions and answers; DOL proposed rule, 91 FR 15454; DHS final rule, 90 FR 60864; OFLC technical release notes, wage year 2025-26; OFLC technical release notes, wage year 2026-27; 8 U.S.C. 1182.
Informational only; not legal, immigration, or financial advice.
See what a job pays in your city, with your own target salary range, beside what employers offer H-1B workers for it.
Check a job or an employer → · What each H-1B wage level pays → · Where our other numbers come from →Get new articles and the biggest layoff notices
One email a week, sent only when there is something new.
Every message has an unsubscribe link. How we use your address
Amazon: 9,746 workers in 30 layoff notices
What state layoff notices and Labor Department H-1B filings show for Amazon: the notices over the past year, the sites, and the sponsorship record.
How to read BEA regional price parities: 387 metro areas
What a regional price parity of 95 or 110 means, how BEA builds one, how 387 metro areas spread around 100, why rent drives it and how the site uses it for pay after cost of living.
How to read DOL wage and hour enforcement data: 267,635 cases
The fields of a Wage and Hour case, the laws behind the numbers, what back wages agreed to pay and penalties assessed mean, and the cases a copy of the dataset leaves out.
How to read the H-1B LCA disclosure data: what 661,885 filings hold
The fields of an LCA, the case statuses, wage units and wage levels, and what the quarterly files leave out, with what 661,885 certified filings show.
How to read a WARN notice: what 4,099 state filings contain
What the notices can say, why some are missing, the federal rules and the state lists, the fields of a WARN notice and how we count them, with what a year of notices shows.
Kroger: 19 layoff notices naming 3,066 workers
The state layoff notices filed by Kroger, its warehouse company and its grocery chains: 19 notices and at least 3,066 workers in a year, with the wage and H-1B records alongside.
Wage data: U.S. Department of Labor OFLC LCA Disclosure Data and U.S. Bureau of Labor Statistics OEWS. LCA filings reflect sponsored-worker wage offers, which skew toward mid-large employers and specific occupations — treat any estimate built from a small sample cautiously. Informational only; not legal, immigration, or financial advice.