Methodology
Methodology and sources
The files, the years, and the rules.
Read this note
Every figure on a career, major, state, or ranking page is either copied from the files below or calculated from those files with the formula on this page. The build stops if a few well-known published figures (the all-occupations median, the registered-nurse median, the anesthesiologist median, computer-science and nursing field medians, and the all-occupations growth rate) do not match the files. That check is there to catch a parser bug, not to freeze the numbers forever. A later BLS release means updating the pins and the citation year together.
Occupational Employment and Wage Statistics, May 2025
Source: U.S. Bureau of Labor Statistics. National file oesm25nat.zip and state file oesm25st.zip, reference period May 2025, news release May 15, 2026. We keep the cross-industry, all-ownership rows. Detailed occupations are the pages. The all-occupations row is the national benchmark.
OEWS covers wage-and-salary workers in nonfarm establishments. It leaves out most of the self-employed, most farm jobs, and private households. Pay is straight-time gross wages. Overtime, stock grants, and nonproduction bonuses are not included. Where BLS publishes both an hourly and an annual wage, the annual figure is generally the hourly wage times 2,080 hours. That is a publishing convention, not a timesheet. Some occupations, including teachers and airline pilots, are published as annual wages only. We do not invent an hourly wage for them.
An asterisk means the estimate was suppressed. A double asterisk means it did not meet BLS reliability criteria. A hash mark means the wage is at or above the open-ended interval. The BLS OEWS FAQ describes that footnote as “equal to or greater than $115.00 per hour or $239,200 per year,” and notes that beginning with the May 2025 estimates many percentiles that used to carry only that footnote now have a specific dollar figure. We print the specific figure when the file has one, and “$239,200 or more” when the cell is still a hash mark.
The 10th percentile is not a starting salary. It is the wage below which 10 percent of employees fall, including experienced workers in lower-paid settings.
The median is the headline wage: half of employees earn more, and half earn less. The mean, when a page shows it, is labeled “Average (mean)” and sits apart from that headline. The mean is higher than the median when a smaller number of very high earners pull the average up. It is not a second estimate of the typical wage.
A career page is published only when the national median, the 10th, 25th, and 75th percentiles are numeric, the 90th is numeric or top-coded, and wage-and-salary employment is at least 5,000. A state page is published only for six high-demand occupations in California, Texas, Florida, and New York, and only when that state's employment is at least 5,000 and the same percentile rule holds. State rankings on a career page require at least 1,000 jobs. We do not adjust for the cost of living. We do not publish city pages. A New York page says it is the state, not New York City.
Employment Projections, 2025–35
Source: U.S. Bureau of Labor Statistics, Table 1.2 in the occupational projections workbook, base year 2025 and projected year 2035. Employment in that table is in thousands, and our pages multiply by 1,000. The projection includes the self-employed, so it will not match OEWS employment. We label both. Growth is the published percent change, compared with the all-occupations percent from the same table (3.5 percent in this file). Openings are the projected annual average and include people leaving the occupation, not just new jobs. Typical education, related experience, and on-the-job training are the BLS entry assignments. They are not a census of workers already in the job.
The trades hub is the hands-on skilled trades with a published page: construction, mechanical, and industrial occupations, plus truck driving and line work. It is not every job that can be entered without a bachelor's degree. The training line on each trade page is the BLS assignment from the projections table.
College Scorecard field of study
Source: U.S. Department of Education, College Scorecard, file Most-Recent-Cohorts-Field-of-Study_10152026.zip (October 15, 2026 release). We use the national variables, not an average of schools. On some rows the national value is privacy-suppressed along with the school's own value. We keep a national figure only when the non-suppressed rows agree on a single number. If they disagree, we do not publish the field.
Earnings are for graduates who were working and not enrolled. They include W-2 wages and positive self-employment earnings, inflated to 2024 dollars. The cohort map in the College Scorecard data dictionary dates them as follows.
- One year out: award years 2020–21 and 2021–22, measured in calendar 2022 and 2023.
- Five years out: award years 2016–17 and 2017–18, measured in calendar 2022 and 2023.
- Seven years out: award years 2014–15 and 2015–16, measured in calendar 2022 and 2023.
Those are different graduating classes. A higher five-year median is not a raise for the one-year class. A major page requires a numeric one-year median, a numeric five-year median, and at least 2,000 people in the five-year cohort. Fields are four-digit CIP codes. A four-digit field can mix several programs, and the page says so when that mix would mislead. Pell medians are the national Pell figures from the same file, shown only when a single number is published.
The “midpoint of bachelor's fields” is the median of those field medians, among bachelor's CIP codes that clear the same 2,000-person rule. It is the middle field, not the median graduate.
CIP to SOC crosswalk
Source: National Center for Education Statistics and the Bureau of Labor Statistics, CIP 2020 to SOC 2018 crosswalk. NCES states that the match is based on program and occupation descriptions, not on where graduates actually go. We use it only as a list of related occupations, with May 2025 wages beside the titles.
What we calculate
- Percent gap: (state median − national median) ÷ national median.
- Employment share: state employment ÷ national OEWS employment.
- Rank: states and the District of Columbia with a published median and at least 1,000 jobs, ordered by median. Ties keep file order. Puerto Rico, Guam, and the Virgin Islands are not ranked.
- Offer check: which published percentile band a number falls in. No interpolated percentile.
- Map colors: five groups with an equal count of states, so a single high-wage state does not flatten the rest. The legend says so.
Files in this repository
The zip and spreadsheet downloads are in data/raw/, with SHA-256 hashes written to dist/data/manifest.json at build time. python build.py rebuilds the site from those files. Refresh the downloads, update the pins if the known figures moved, and rebuild. Do not hand-edit numbers in the HTML.