Definitions before comparisons.
The Atlas measures public employment and compensation, without equating public ownership with exclusive tax financing. It does not automatically include employees of private contractors delivering public services.
Three different scopes
- Public administration — a narrower functional or institutional boundary. NACE O is not the whole public sector.
- General government (S.13) — central, state and local governments and social security funds, including public non-market units such as some schools and hospitals. Public corporations are excluded. This is the main OECD comparison scope.
- Survey-defined public sector — may include state-owned enterprises. WWBI harmonises household surveys, but exact coverage varies across surveys.
What we compare
- Employment share: S.13 employment / total employment, as reported by OECD. Total employment is different from paid employment.
- Per 1,000 residents: OECD S.13 persons / World Bank SP.POP.TOTL population in the same year × 1000. Years are never mixed. Domestic employment and resident population differ for cross-border workers.
- Compensation: Eurostat D.1, wages and salaries plus employers’ social contributions, as a percentage of GDP. COFOG functions show costs, not headcounts.
- WWBI: a separate descriptive view. BI.EMP.TOTL.PB measures the public share of total employment. The API returns a proportion; we multiply it by 100. It is not merged with the OECD series.
WWBI limitations
WWBI Codebook 3.1 describes the use of NACE O/P/Q as a public-sector proxy in some EU-SILC data for the EEA in 2004–2018. This can include private education and health providers. The API does not identify the underlying survey for each observation. Geography and year alone do not establish exact scope.
WWBI therefore has no international ranking or median in the Atlas. Its map and charts are descriptive. International comparisons and trends require source review; coverage changes may cause methodological breaks. WWBI Codebook 3.1 (pp. 8–9).
Integrity rules
- The map, table and ranking use one selected year. No last-known value is substituted for a missing observation.
- Missing means null or an absent observation, never zero. No interpolation; chart lines break at missing years.
- The median is unweighted and includes only countries with data in the selected year. It is not a global average. OECD and EU aggregates are excluded from rankings.
- Totals are never added to their components; overlapping demographic or sectoral breakdowns are not summed. Eurostat uses only COFOG level 1.
- Persons, FTEs, sample counts, gross wages, adjusted premiums and ratios of means are distinct measures. They are not substituted for one another.
- Source flags are preserved. OECD A indicates a normal observation; Eurostat p means provisional, e estimated, and b a break in series. Consult source code lists.
What the MVP does not yet claim
We do not publish a guessed public-to-private mean wage ratio. The WWBI adjusted wage premium is a model result, not a ratio of means. No fictional breakdown of central, local, tax, education or health staffing is presented. The next stage requires verification of ILOSTAT and national sources, institutional coverage and persons/FTE units.
Reproducibility and sources
Every value has a country identifier, year, unit, scope, definition, method, source URL, retrieval timestamp and raw-response SHA-256. Derived values identify their inputs. JSON and CSV contain this metadata.
Download data and raw responses →
OECD: Employment in general government, methodology and country exceptions