Pay by Role editorial · Updated August 25, 2026 · 12 min read
How Pay by Role sources salary data
Salary pages are only useful when you know what the numbers represent. A city median pulled from labour-market microdata is not the same thing as a national average scaled by a pay index — and treating them as identical is how people over- or under-value an offer.
This guide explains how Pay by Role sources salary data, how we label observed versus modeled figures, and what we refuse to publish. If you are deciding whether to trust a number on a city page, start here, then read when to trust salary data.
Why transparency matters
Pay by Role is a research tool, not a job board. We publish country and city pay ranges from public APIs at build time, label how each figure was produced, and refresh the dataset on a schedule.
When you open a salary page, you should be able to tell whether you are looking at a metro-specific percentile set or a national figure adjusted for local pay levels. Those two paths answer different questions.
What public sources feed the dataset?
National and city percentiles primarily come from OfficialSalary aggregations of public statistical programmes — including BLS OEWS-style occupation wages, OECD series, and ILOSTAT where available — plus SalaryByRole for occupations OfficialSalary does not yet cover.
FX rates come from a public exchange-rate feed so local-currency figures can be compared in a display currency of your choice. Country context such as consumer-price inflation, GDP per capita, and unemployment comes from World Bank Open Data. Public holiday counts come from Nager.Date.
Every country–job pair on the site requires a real median from one of those salary sources. If a source does not cover the pair, we do not publish that combination. Coverage gaps usually mean the underlying programmes do not publish a usable median — not that the role pays nothing.
See live source counts and the last refresh timestamp on About & methodology.
What is observed city data vs a modeled estimate?
Observed city data is a full percentile set returned for a specific metro and occupation. Those pages are our strongest local estimates for relocating or judging a local offer.
A modeled city estimate starts from a real national median, multiplies by that city’s pay index, then applies a spread model for entry-to-senior bands. Modeled pages are labelled and kept for navigation and comparison — they are not metro microdata.
How to read the two city-page types
| Type | What it is | Best use |
|---|---|---|
| Observed | Metro-specific percentiles from the source | Relocating or judging a local offer |
| Modeled | National median × city pay index, plus a spread model | Directional comparison when the metro is not broken out |
Cost-of-living adjustments, rent models, simplified tax take-home, and illustrative salary trends are derived helpers. They add context; they are not government statistics. For the index math, see understanding pay and COL indexes.
How often does the data refresh?
The dataset is regenerated by an automated pipeline, typically monthly. Wage surveys lag the market. A fast-moving specialty — new AI titles, niche contractor rates — can shift between builds.
If your offer sits far outside a published band, check whether the job title maps cleanly to the occupation we publish, and whether the employer’s geo tier matches the city you selected.
What this data is not
- Not an offer, contract, or tax filing.
- Not a promise that you will land at the median or the 90th percentile.
- Not a substitute for employer bands, recruiter feedback, or official local series when the decision is large.
Employer pay bands, equity, overtime, and local taxes can move individual outcomes well outside a published range. Use how to read a salary range before you treat any single number as “the market.”
Key takeaways
- Country–job pages require a real source median; we do not invent uncovered pairs.
- Observed city percentiles are stronger local estimates than modeled pay-index pages.
- COL, rent, tax, and trend modules are helpers — not official statistics.
- The pipeline typically refreshes monthly; surveys still lag fast-moving titles.
- Prefer the city you will actually be paid in, then verify with the employer’s band.
Frequently asked questions
- Where does Pay by Role get salary numbers?
- Country and city percentiles come from public labour-market aggregations (OfficialSalary, and SalaryByRole where needed), built from programmes such as BLS OEWS-style wages, OECD, and ILOSTAT. FX, World Bank economics, and holiday counts add context. We only publish a country–job pair when a salary source returns a real median.
- What does “observed” salary data mean?
- Observed means the source returned a percentile set for that metro and occupation, not a national figure scaled to the city. Use observed pages when you are relocating or judging a local offer. They are the figures we treat as the strongest local references on this site.
- What does “modeled” salary data mean?
- Modeled city pay starts from a real national median, then applies the city’s pay index and a spread model for entry-to-senior bands. It is useful when sources do not break out that metro. It is not metro microdata — widen your confidence interval before you negotiate hard on it.
- Why is a job missing in some countries?
- We do not publish a country–job pair without a source median. A gap usually means the statistical programme does not cover that occupation there. Use a closely related job family as a peer benchmark, not as a substitute for employer data.
- How current is the salary dataset?
- An automated pipeline typically regenerates figures monthly. Underlying wage surveys still lag. Check the About page for the last build timestamp. If your specialty moves faster than survey cycles, treat published bands as an anchor and confirm with recent offers.
- Can I cite these figures as official statistics?
- Cite the underlying public programmes when you need an official series. Pay by Role is a research layer: we aggregate, label observed versus modeled paths, and add derived helpers. Do not present COL-adjusted pay, rent sketches, or tax estimates as government statistics.
Next step
Knowing the source path is the difference between using this site well and treating every median as gospel. Open a city page, check observed versus modeled, then run the same gross figure through take-home and cost-of-living tools.
See methodology, source counts, and last refresh →
Then pick a country or job and open an observed city page for the role you care about.