SeoulMarkets · data for licence
The National Pension wage panel
553,978 active workplaces · 11,561,661 subscribers · month of 2026-06
Every month, the national pension bills each Korean workplace 9% of its workers' pay. Nobody games that number to look good — it is what the pension actually collects. Divide it by headcount and roll it up, and you get a pay ladder for almost the entire workforce, one that no survey and no tax return can match for coverage. Here is that panel, with the one thing a buyer needs and competitors leave out: how far each figure is a floor.
- What it is — the pension bill (9% of pay) per subscriber, rolled up into an industry, region and firm-size pay ladder.
- Why trust it — it is money the pension really collects, not a survey or a self-report. There is nothing to inflate.
- How big — 553,978 active workplaces, ~11.56 million subscribers. Not a sample — close to a census.
- The limit — pay above the income ceiling is billed as if it were at the ceiling, so those cells read low. Every figure is therefore a floor.
- The form — one CSV per axis (industry 1,232, region 16, size 7), each row carrying a cap-pressure column. Refreshed monthly. Redistribution-free source.
① The sample — how many
| Active workplaces (status code 1) | 554,570 | per-head figures use the 553,978 that also file a monthly bill |
|---|---|---|
| Subscribers (people) | 11,561,661 | one workplace row can carry many people |
| Withdrawn workplaces | 38,557 | dropped, never counted as zero |
| Month of record | 2026-06 | monthly release, 1–2 months in arrears |
Close to a census. This counts almost every registered workplace, not a thousand-person poll.
② Where the ceiling presses — and how much
When a worker earns above the monthly income ceiling (a bill of ₩573,300), the pension is figured as if they sat exactly at the ceiling. Their cell reads low. Of all 11.56 million subscribers, 634,927 (5.5%) are at the ceiling. We count that pressure on every axis — and hand it over with the data, because it is the part a buyer has to see.
| Axis | Most-pressed cells | At the ceiling |
|---|---|---|
| Industry (1,232) | Video & imaging-equipment manufacturing | 95.0% |
| Computer manufacturing | 51.4% | |
| Truck & special-vehicle manufacturing | 42.7% | |
| Region (16) | Gyeonggi | 9.4% |
| Seoul | 6.1% | |
| Daegu | 0.3% | |
| Firm size (7) | 1,000+ employees | 21.0% |
| 300–999 employees | 4.1% | |
| 10–29 employees | 0.3% |
Read the region row carefully: Gyeonggi is pressed harder than Seoul, so a raw ranking that ignores the ceiling would put them in the wrong order. That is exactly the mistake the cap-pressure column stops a buyer from making.
③ Every figure is a floor — the error runs one way
This data's error has a fixed direction. That is what makes it safe to quote.
- The ceiling only cuts downward. A pressed cell reads below the true pay, never above it.
- The count itself under-counts. We mark a cell pressed only where its workplace average reaches the ceiling, so workplaces where only some workers are capped are missed — the real number of pressed people is above 634,927.
- Therefore every number is a floor. A buyer can cite it as "at least this much" and stay safe.
A dataset that knows its own limit is worth more than one that hides it. Rival tables rank without removing cap pressure; we attach it, cell by cell, so a buyer can decide for themselves whether a figure belongs in their report.
④ What you get
| Files | Three CSVs — industry.csv (1,232 rows),region.csv (16), size.csv (7). Columns: group · subscribers · average monthly bill (pay proxy) · pressed subscribers · share at the ceiling. | |
|---|---|---|
| The floor column | Every row carries the share-at-ceiling figure, so a buyer sees at a glance how far that cell is a floor. | |
| Definitions | A method.md that quotes tenure, bill and ceiling from the source verbatim — not paraphrased into our own words. | |
| Refresh | Monthly, on the pension's release cycle (1–2 months in arrears). Every changed figure is logged in corrections.csv, not silently edited. | |
| Source | National Pension registered-workplace file (data.go.kr dataset 15083277). Redistribution unrestricted. The CP949 (Korean-Windows) encoding is decoded by us. | |
Ask for the panel
Want a sample cut or the full three-axis panel for your desk? Leave an email and we send you the sample and the current terms. It reaches us directly — no account, and this is not the newsletter.
While you are here, the company-level workforce table this joins to is free to download — the whole listed-company table (CSV), or browse it on the rankings page.
Data as of 2026-06 (National Pension), retrieved via data.go.kr · Figures are floors, as described above · Not investment advice. Browse all SeoulMarkets datasets. See our terms of use and refund policy.