- Generated by
1.9.8
|
Occupation Generator 3.0.1
Employment status and occupation by country, age, sex and education for C++23
|
The data in resources/ and how occupation-generator uses it (rule 6 of the dasmig standard). For using the library, see the usage guide.
There are seven tables. Five are measured, from the ILO's labour force statistics (ILOSTAT): the employment status of the population by country, sex and age; its education; the ISCO-08 group of the employed by country and sex; and how much the age and education of workers shift those groups. Two are curated: the catalogue of jobs, and how the inactive split into students, retired and others.
What is measured and what is curated. The rates and shares are measured, and each row says whether its country was surveyed or filled in from its region. The jobs are curated: no open dataset lists job titles with the skills and gear a game needs. A job is mapped to an ISCO-08 group, and the group is what the measured data speaks about. How the jobs of a group share the group's weight is a judgement.
| Tier | Countries | Jobs | Notes |
|---|---|---|---|
lite | 213 | 95 (1 to 5 per ISCO group) | The measured tables are the same in both tiers |
full | 213 | 320 (2 to 18 per ISCO group) | The full catalogue |
The 213 countries are every independent country in country-generator's full list plus the territories ILOSTAT has a survey for (Hong Kong, Macau, Puerto Rico, Taiwan, Palestine, New Caledonia and 12 others). Antarctica and the other territories without a population or a survey are left out. Because the tiers differ only in the catalogue, a group in lite often has a single job that takes the group's whole share: Software developer is every ICT professional there.
Source: ILO, ILOSTAT, indicators downloaded through its bulk API by scripts/fetch_ilostat.py. ILOSTAT data is published under CC BY 4.0. The indicators:
| Indicator | Used for |
|---|---|
EMP_TEMP_SEX_OC2_NB_A: employment by sex and ISCO-08 level 2 | shares.tsv |
EMP_TEMP_SEX_AGE_OCU_NB_A: employment by age class and ISCO-08 level 1 | factors.tsv, age_band rows |
EMP_TEMP_SEX_OCU_EDU_NB_A: employment by ISCO-08 level 1 and education | factors.tsv, education rows |
POP_XWAP_SEX_AGE_EDU_NB_A: working-age population by sex, age and education | education.tsv |
POP_XWAP_SEX_AGE_NB_A, EMP_TEMP_SEX_AGE_NB_A, UNE_TUNE_SEX_AGE_NB_A: working-age population, employment and unemployment by sex and five-year age band | status.tsv |
Every measured row has a basis and a year. basis is ilostat when the country has a survey, and region when it does not. For each country and table the script takes the latest survey since 2010 that has the complete table (the same source and year for every cell), so one table's year can differ from another's.
Countries with no survey. A country gets the unweighted mean of the surveyed countries in its UN subregion (else its region, else the world), with basis region and year empty. The number of countries with a survey of their own:
| Table | Surveyed | Filled in |
|---|---|---|
status | 139 | 74 |
education | 185 | 28 |
shares | 149 | 64 |
factors, age_band | 166 | 47 |
factors, education | 148 | 65 |
Most surveys are from 2022 to 2025 (23 of the 139 status surveys are from 2024, 45 from 2025); the oldest are from 2010. A regional mean is a guess for a country that may be far from the average of its neighbours. Rows with basis region should not be read as a statement about that country.
| Column | Meaning |
|---|---|
country_code | ISO 3166-1 alpha-2 code (the key used by country-generator) |
sex | male or female |
band | Five-year age band: 15-19, 20-24, ... 60-64, and 65+ |
employed, unemployed, inactive | Percent of the population of that sex and band: employment divided by the working-age population, unemployment divided by it, and the rest. They add up to 100 |
basis, year | As above |
"Employed" and "unemployed" follow the ILO's definitions (a person is unemployed when they have no work, are available and are looking for it); the inactive are everyone else of working age. The 65+ row covers everyone from 65 to the end of life, so an 85-year-old has the employment rate of the average person over 65. Under 15 there is no data: a person is a student from 6 to 14 (school age, which is 4 to 7 depending on the country; 6 is the common one), and inactive under 6.
If a survey has the three indicators from different sources, the script uses the employment source's and takes the others from the same year, else any source with a complete set. Where employment and unemployment together exceed the population (inconsistent sources) they are scaled down to fit.
How the inactive of an age band split into student, retired and other. Curated: no table describes this directly. The shares follow what the labour-force surveys suggest about reasons for inactivity (education dominates under 25, retirement over 55, care and illness in between), but the numbers are judgements and the same for every country:
| Band | Student | Retired | Other |
|---|---|---|---|
15-19 | 0.85 | 0 | 0.15 |
20-24 | 0.55 | 0 | 0.45 |
25-29 | 0.12 | 0 | 0.88 |
30-34 | 0.04 | 0 | 0.96 |
35-54 | 0.01 | 0 | 0.99 |
55-59 | 0 | 0.20 | 0.80 |
60-64 | 0 | 0.50 | 0.50 |
65+ | 0 | 0.92 | 0.08 |
The country matters here: where women stay home the "other" share should be larger and where schooling is long the student share should be, but this table does not know. A person who is employed and studying at once counts as employed.
| Column | Meaning |
|---|---|
band | Ten-year age band: 15-24, 25-34, 35-44, 45-54, 55-64, 65+ |
less_than_basic, basic, intermediate, advanced | Percent of the working-age population of that sex and band at each of the ILO's four aggregate education levels, among those whose level is known |
The levels are ILOSTAT's EDU_AGGREGATE groups: less than basic (below primary or none), basic (primary and lower secondary, ISCED 1 to 2), intermediate (upper secondary and post-secondary non-tertiary, ISCED 3 to 4) and advanced (tertiary, ISCED 5 to 8). Countries classify their own systems into these, and not all use all four: Japan reports no "basic" level at all, so its shares have 0 there in every band.
Age limits. The age band of a young person is wide: the 15-24 row describes ages 15 to 24 together. The generator therefore applies minimum ages (curated): basic needs 14 years, intermediate 17 and advanced 21. A person too young for a level counts as the one below, so a 16-year-old who would be "intermediate" in the band is "basic".
The 43 groups of ISCO-08 at level 2: code (11 to 96), label (the ISCO-08 title), major (the first digit, which ILOSTAT uses for the age and education tables; 0 for the armed forces), and skill, the ISCO-08 skill level of the major group (1, 2 or 3-4; empty for other ranks of the armed forces, which span levels). Written from the ISCO-08 structure published by the ILO.
share is the percent of the employed of that sex who work in the ISCO-08 group code (omitted when 0). Each country and sex adds up to 100. Source: EMP_TEMP_SEX_OC2_NB_A. The cells ILOSTAT flags as unreliable (a small sample) are kept, because dropping them would bias the total. ILOSTAT's "not classified" group is left out and the rest rescaled.
How well a country's occupations map to ISCO-08 varies: some use ISCO-08 directly, others convert from a national scheme, and a few groups are not reported separately (for example group 63, subsistence farmers, is often folded into 61). Differences between countries in a group's share are partly real and partly classification, so compare the major groups rather than single level-2 groups.
Multipliers on a group's share, by the ISCO major group major (1 to 9, and 0 for the armed forces) of the group, for two dimensions:
| Dimension | Values | Meaning | | — | — | — | | age_band | 15-24, 25-54, 55-64, 65+ | P(age class | major group) / P(age class), among the employed | | education | the four levels | P(education | major group) / P(education), among the employed |
A factor above 1 means the major group employs more of that age class or level than the labour force as a whole does; below 1 means fewer. Rows within 3% of 1 are left out (a missing row is 1) and factors are limited to 0.05 to 8. They are the Bayes factors that turn the country's shares for a sex into shares for someone of a given age and education: with P(group) the share and the factors above, P(group | age, education) is proportional to share × age factor × education factor.
Sources: EMP_TEMP_SEX_AGE_OCU_NB_A (the age classes are the only ones ILOSTAT cross-tabulates with ISCO-08: 15 to 24, 25 to 54, 55 to 64 and 65 and over) and EMP_TEMP_SEX_OCU_EDU_NB_A. Both are for men and women together.
What this assumes. The factors are the same for men and women, and the age and education factors are multiplied as if age and education were independent given the group. Neither is true: young men and young women differ in the groups they enter, and the young are more educated within every group. Where the data has both, the product over-corrects a little (a young graduate is a little too sure to be a professional). The factors also describe the employed and are applied to the unemployed, retired and others as their usual or last job.
| Column | Meaning | Source |
|---|---|---|
key, label | Stable identifier and display name | Curated |
isco | The ISCO-08 level-2 group. The group's measured weight is spread over its jobs | Curated mapping, using the ISCO-08 unit-group definitions |
unit | The ISCO-08 unit group (four digits) the job belongs to, for reference | Curated |
weight | Relative weight within its group (0.1 to 8) | Curated: a judgement of how common the job is within the group, not a measurement |
min_age, max_age | Hard limits, such as 29 for a surgeon, 30 for a judge and 40 for a model; blank for none | Curated |
tags | ;-separated character tags (manual, caring, desk, risk, leadership, ...) for other generators, such as nickname-generator | Curated |
skills | ;-separated skill keywords | Curated, with O*NET and ESCO as a guide; no text was copied |
gear | ;-separated clothing and gear keywords for a game's art kit | Curated |
The catalogue is in resources/raw/occupations_curated.tsv with a tier column saying which tier a job belongs to. lite has between one and five jobs per group and full up to 18. Every ISCO group that has employment in any country has at least one job in each tier (the build script checks this), and a group in which no job is open to a person's age is skipped when the group is drawn.
The weights, tags, skills and gear are not from a dataset; they are one person's idea of the job. A layer can replace any of them (see Extending Data).
For a call with country c, sex s and age a:
"country"). The data has no population column; pass a country chosen by country-generator for a population-weighted world."sex").status.tsv row for c, s and a's band (stream "status"); if inactive, student, retired or other by inactivity.tsv (the same stream, next value).a is 15 or more: one of the four levels by the education.tsv row (stream "education"), after moving the weight of levels a is too young for down a level. Under 15: none.g has the weight share(c, s, g) × the age factor for g's major group and a's age class × the education factor for g's major group and the person's level; a group with no job open to a has weight 0 (stream "isco"). Then a job of the group by catalogue weight, among those open to a (stream "job"). The age class of a person under 15 is 15-24.Each field draws from a stream of its own (call.split(label)), so forcing one field in the request never changes another.
prepare_occupations.py reads the downloads and the curated inputs in resources/raw/ and writes both tiers. ILOSTAT revises its figures and adds years, so a rebuild today gives different numbers from the ones shipped; the committed resources/ are the reference. They were built on 2026-10-14.
1.9.8