If you are searching for how to translate ISCO codes, how to translate occupation codes, how to translate job titles for labour statistics, or how to keep an occupational classification correct when a workforce dataset moves into another language, the first rule is simple: translate the human-language title and explanation, but do not translate the classification code itself. ISCO-08 codes are reference data. They identify categories in the International Standard Classification of Occupations and should remain stable while the surrounding language changes.
This matters in employment surveys, population censuses, labour-force statistics, migration systems, skills databases, vacancy analytics, career guidance, HR reporting, payroll analysis, occupational health research, education-to-work studies and international development. A translation can be fluent and still be statistically wrong if a job title is mapped to the wrong ISCO-08 unit group, if a national occupation code is mistaken for an ISCO code, if a manager title is interpreted from prestige rather than actual tasks, or if a translator changes the code because the target-language occupation name looks different.
This guide explains how to translate ISCO-08 occupation codes and job titles accurately. It shows how the hierarchy works, why job title alone is not enough, how skill level and skill specialisation affect classification, how to distinguish occupations from industries and qualifications, how to handle national crosswalks, and how to protect occupational data in spreadsheets, surveys, APIs and multilingual reports. The aim is not to create another broad translation hub, but to solve one narrow problem inside the wider eduKateSG translation architecture: moving occupational meaning across languages without moving the worker into the wrong statistical category.
1. ISCO-08 classifies jobs by occupation, not by employer
ISCO-08 is designed to classify jobs according to the kind of work performed. That means the occupation is determined by tasks, duties, skill level and skill specialisation rather than by the name of the company, ministry, hospital, school, factory or industry employing the person. A software developer employed by a bank is still classified by the software-development work being done, not automatically under a banking occupation. A translator must therefore resist the temptation to infer the occupation from the employer’s sector.
Translation control: preserve the distinction between occupation and industry. QA: check whether the translated title still points to the worker’s actual tasks rather than the employer’s business activity.
2. The ISCO-08 hierarchy has four classification levels
The International Labour Organization describes ISCO-08 as a four-level hierarchy with 10 major groups, 43 sub-major groups, 130 minor groups and 436 unit groups. The code becomes more specific as digits are added. A one-digit code identifies a major group; two digits identify a sub-major group; three identify a minor group; and four identify a unit group. When translating a table, every digit therefore carries structural meaning. Dropping a leading zero, adding punctuation or shortening a four-digit code can change the classification level.
Translation control: keep code length and hierarchy intact. QA: verify that every translated row retains the same one-, two-, three- or four-digit level as the source.
3. Major groups are broad occupational families
At the highest level, ISCO-08 groups occupations into major families such as Managers, Professionals, Technicians and Associate Professionals, Clerical Support Workers, Service and Sales Workers, Skilled Agricultural, Forestry and Fishery Workers, Craft and Related Trades Workers, Plant and Machine Operators and Assemblers, Elementary Occupations, and Armed Forces Occupations. These are classification labels, not employment grades or social-status rankings. A target-language term that sounds more prestigious or less prestigious can distort the statistical meaning.
Translation control: translate major-group labels by occupational function, not social connotation. QA: compare each target label with the official group definition rather than relying only on a dictionary equivalent.
4. Unit groups are the most detailed ISCO-08 level
The 436 unit groups are the most detailed standard categories in ISCO-08. A unit group brings together occupations that share a high degree of similarity in tasks and skill requirements. When a source dataset already contains a verified four-digit code, the translation should preserve it. If the source contains only a job title and asks the translator to assign a code, that is no longer ordinary translation; it is occupational coding and requires enough information about the job’s duties.
Translation control: distinguish translating an assigned code from assigning a new code. QA: flag uncoded titles that lack task descriptions instead of guessing a unit group.
5. Job titles are clues, not proof
A title such as “manager,” “engineer,” “assistant,” “technician,” “consultant” or “coordinator” can mean very different work in different organisations and countries. Some titles are honorary, some are internal pay grades, and some are intentionally broad. ISCO classification should follow the actual job. Translation should therefore avoid treating the source title as if it uniquely determines the code. The safest method is to translate the title and preserve the verified code, or examine tasks before any new classification decision.
Translation control: do not promote or demote a worker statistically because of the wording of a title. QA: review ambiguous titles against task descriptions.
6. Tasks and duties are the strongest semantic evidence
When classification is genuinely required, what the person does matters more than what the business card says. A “project manager” who primarily performs software development may need a different occupational classification from someone who truly plans, directs and coordinates a project function. A “research assistant” may perform laboratory work, statistical analysis, field interviewing or clerical support. Translating task statements precisely is therefore essential because weak translation can corrupt the evidence used to assign the occupation.
Translation control: preserve verbs describing actual work: diagnose, install, teach, operate, supervise, design, reconcile, repair. QA: compare the translated task list with the source before coding.
7. Skill level is a classification concept, not a compliment
ISCO-08 uses skill level to help organise occupations. The ILO describes four broad skill levels based on the complexity and range of tasks and duties, together with typical education, training and experience. “Higher” skill level does not mean that the person is more valuable as a human being, nor does “lower” mean unimportant work. Translators should avoid target-language words that turn technical classification into status judgement.
Translation control: use neutral statistical terminology for skill level. QA: remove wording that implies social worth, intelligence or moral rank.
8. Skill specialisation distinguishes kinds of work
Occupations can require similar overall skill levels while differing sharply in subject matter, tools, materials, products, services or work context. ISCO calls this skill specialisation. Translators need enough technical vocabulary to keep those distinctions intact. An electrical engineering technician is not interchangeable with a medical technician; a metal machinist is not a generic machine operator; a primary teacher is not simply an education worker.
Translation control: preserve domain-specific nouns and task verbs. QA: test whether the target title could accidentally fit a neighbouring occupational group.
9. Occupation and qualification are different dimensions
A person’s degree, diploma, licence or certificate can provide evidence about likely skills, but ISCO classifies the job performed. Someone with an engineering degree may work as a teacher, manager, analyst or entrepreneur. Someone without a particular formal degree may perform a job through experience or vocational training. Translating “occupation” as “qualification” creates a serious conceptual error.
Translation control: keep occupation, educational attainment and professional credential fields separate. QA: audit datasets where degree names and job titles appear together.
10. Occupation and industry are different dimensions
Industry classifications such as ISIC describe the economic activity of establishments or enterprises. ISCO describes the work people perform. These can cross in every direction. Hospitals employ accountants, cleaners and IT specialists; software firms employ sales staff and office managers; schools employ administrators and maintenance workers. eduKateSG already has a separate `Translate |` owner for NAICS, NACE and ISIC industry codes, so this article intentionally stays on the occupation side of the boundary.
Translation control: never substitute an industry code for an occupation code. QA: ask whether the field answers “what does this person do?” or “what does this establishment produce or provide?”
11. ISCO-08 and national occupation systems are not automatically identical
Countries often maintain national classifications adapted to local labour markets. These systems may be based on or related to ISCO but use different codes, titles or levels of detail. A translator should never assume that a national four-digit code is an ISCO-08 code merely because the format looks similar. Crosswalks can be one-to-one, one-to-many, many-to-one or version-specific.
Translation control: label the code system explicitly. QA: verify the classification name and version before any mapping.
12. Crosswalks are mappings, not translations
Mapping a national occupational code to ISCO-08 is a statistical harmonisation task. Translation may help users understand the titles and notes, but the mapping relationship must come from an authorised correspondence table or documented methodology. A good linguistic match does not prove a valid statistical match. This distinction becomes critical in migration, credential recognition and international workforce comparison.
Translation control: separate language equivalence from classification correspondence. QA: record the official crosswalk source and version.
13. Version names belong in the data
ISCO-08 replaced earlier ISCO versions. A code may have a different meaning, scope or correspondence across revisions. Translating a historical dataset without preserving the version can make later analysis unreliable. The text “ISCO” alone is often insufficient when codes from several eras coexist.
Translation control: retain “ISCO-08” where that is the source standard. QA: check metadata, correspondence tables and publication date before interpreting unfamiliar codes.
14. Historical datasets should remain historical
A census conducted under an older occupational classification should not be silently recoded into ISCO-08 just because the translation is being produced today. Recoding can be useful for comparison, but it should be a documented analytical transformation with its own method. Translation should first preserve what the source actually recorded.
Translation control: translate the original classification faithfully. QA: label any harmonised or recoded series separately from the historical source.
15. Leading zero in Major Group 0 is meaningful
Armed Forces Occupations use Major Group 0. Systems that treat occupation codes as ordinary integers can remove leading zeros from detailed codes beginning with zero. That changes the stored representation and can break lookups, joins and exports. Spreadsheet software is a common source of this problem.
Translation control: store occupational codes as text, not numbers. QA: compare literal strings before and after spreadsheet import.
16. Do not localise the digits of the machine code unless the system explicitly permits it
A user interface may display local numeral glyphs in some languages, but the machine-facing classification value should remain the canonical code expected by the database and exchange format. The safest architecture separates the stored code from presentation formatting.
Translation control: preserve canonical code values in the data layer. QA: inspect API payloads, CSV exports and copy-paste behaviour across locales.
17. Occupation titles often contain culture-specific wording
Some job titles have no neat one-word equivalent across languages because administrative systems, trades, licences and organisational structures differ. Translating word by word can suggest a target-country occupation that is not actually equivalent. In those cases, a descriptive translation may be safer than replacing the source title with a familiar but inaccurate local title.
Translation control: preserve source-system specificity. QA: distinguish lexical similarity from institutional equivalence.
18. “Engineer” is a classic translation trap
The word translated as “engineer” can refer to a professional engineer, an engineering technician, a locomotive driver, a maintenance specialist, a machine operator or another role depending on language and industry. The source title is therefore insufficient evidence. Task description, regulatory context and equipment handled can determine the correct occupational meaning.
Translation control: do not map every engineering-like word to one target occupation. QA: inspect duties and professional context.
19. “Manager” is another dangerous title
Many organisations use “manager” in titles for senior individual contributors who do not actually plan, direct and coordinate organisational units. Conversely, some people perform management work without the word “manager” in their title. ISCO coding should follow the role rather than office prestige.
Translation control: translate the title naturally but do not infer Major Group 1 automatically. QA: look for responsibility for planning, directing and coordinating people or operations.
20. “Technician” varies across education systems
Technician titles can imply different qualification routes and scopes of practice across countries. Some are tightly regulated; others are general workplace labels. ISCO classification uses work content, not a universal assumption about the educational level attached to the word “technician.”
Translation control: avoid importing target-country credential assumptions. QA: read the occupation definition and task list.
21. “Assistant” tells you almost nothing by itself
An assistant can be a professional support role, health worker, laboratory worker, teaching assistant, administrative employee, personal aide or trainee. Translating “assistant” accurately at the word level does not solve occupational classification. The noun being assisted, tasks performed and degree of autonomy matter.
Translation control: retain the full compound title and task context. QA: reject generic mappings based only on the word “assistant.”
22. Self-employment does not create a separate occupation
Employment status and occupation are separate variables. A self-employed plumber and an employed plumber can perform the same occupation. A business owner who mainly manages the enterprise may belong elsewhere than an owner who primarily performs the trade. Translation should not treat “self-employed” as an occupation title.
Translation control: keep employment status outside the occupation field. QA: determine the actual job performed by working proprietors.
23. Multiple jobs require a declared coding rule
A respondent may hold more than one job. Labour-force surveys often designate a main job according to hours, income or survey rules. Translation should preserve that rule because the occupation code applies to the selected job, not necessarily every activity the person performs.
Translation control: translate “main job,” “second job” and related concepts consistently. QA: confirm which job the code refers to.
24. Apprentices and trainees need task-aware treatment
People in apprenticeships or structured training may already perform substantial occupational tasks while still learning. Other trainees may mainly observe or study. The correct statistical treatment depends on the survey and classification rules. A translator should not upgrade or downgrade the occupation merely from words such as “apprentice,” “intern” or “trainee.”
Translation control: preserve training status and occupation separately where the source does. QA: follow the coding manual used by the data owner.
25. Supervisors are not automatically managers
Supervision can be part of many occupations. A working supervisor may still spend most time performing the trade or operational job while directing a small team. Management classification generally depends on the broader nature of planning and coordination responsibilities, not simply whether someone oversees others.
Translation control: distinguish “supervisor,” “foreperson,” “team leader” and “manager” carefully. QA: examine the balance between operational and managerial duties.
26. Public-sector titles need institutional caution
Civil-service ranks and administrative titles can be highly country-specific. A title that literally resembles “secretary,” “officer,” “director” or “commissioner” may not correspond to the same occupational content elsewhere. Translation should preserve the institution’s role while classification follows duties.
Translation control: avoid substituting a target-country rank that implies false equivalence. QA: use descriptive notes when necessary.
27. Health occupations demand regulated vocabulary
Health professions often have protected scopes of practice. “Nurse,” “nursing associate,” “midwife,” “medical assistant,” “physician assistant,” “paramedic” and “technician” can refer to different roles across systems. A mistranslation can push a record into the wrong occupational group and may also imply a licence the person does not hold.
Translation control: preserve professional scope and regulatory status. QA: compare the source role with official occupation definitions rather than title resemblance alone.
28. Teaching titles should distinguish level and function
Early-childhood educators, primary teachers, secondary teachers, vocational instructors, university academics, special-needs teachers and teaching assistants can belong to different occupational groups. “Teacher” as a single translation may be too broad when the source carries level information.
Translation control: retain education level and instructional function. QA: ensure the target title preserves whether the person teaches, assists, lectures, trains or provides specialist support.
29. Digital occupations evolve faster than titles
Job titles such as data scientist, cloud engineer, product designer, cybersecurity analyst and machine-learning engineer may develop faster than international classification revisions. A modern title can still be classified by examining its tasks and matching the closest defined occupation. Translation should not invent a new ISCO code for a fashionable title.
Translation control: translate emerging titles descriptively while preserving assigned codes. QA: document coding decisions for new occupations.
30. Gig-platform labels can hide the real occupation
Platform workers may be called “partners,” “creators,” “taskers,” “riders” or “independent professionals.” Those commercial labels do not necessarily describe the occupation. A delivery rider, driver, cleaner, designer or tutor still needs classification by actual work.
Translation control: separate platform branding from occupational function. QA: map tasks, not app terminology.
31. Surveys need questions that survive translation
Occupational coding quality starts before the code itself. Survey questions such as “What is your main job title?” and “What are your main tasks and duties?” must be translated so respondents understand them similarly across languages. If the target question asks for employer, rank or qualification instead, the resulting code will be biased even if the coding team works perfectly.
Translation control: preserve the construct measured by each survey item. QA: conduct cognitive testing with target-language respondents where possible.
32. Open-ended occupation answers need controlled interpretation
Respondents often enter free-text answers with spelling variants, abbreviations, local slang and incomplete titles. Translating those answers into another language before coding can remove useful local clues. A strong workflow keeps the original text, produces a target-language rendering, and records the assigned code separately.
Translation control: never overwrite the raw answer. QA: retain source text, translated text, coder decision and confidence or query status.
33. Occupational indexes help, but they do not replace definitions
ISCO indexes list occupational titles and associated groups, helping coders find candidate classifications. An index hit is a starting point, not always the final answer. Local titles can be ambiguous, and some jobs require task information to distinguish neighbouring groups.
Translation control: use indexes for navigation and definitions for confirmation. QA: read inclusion and exclusion notes where available.
34. Machine translation should protect codes as variables
Most translation engines will leave a four-digit code untouched, but a production workflow should not rely on probability. Codes should be locked, masked or stored outside the translatable segment. The human-language title and definition can be translated while the code passes through deterministically.
Translation control: treat ISCO codes as protected tokens. QA: compare the complete source and target code sets.
35. Generative AI can explain, but should not guess missing occupational evidence
A language model can summarize an ISCO definition or suggest candidate groups from a detailed task description. It should not invent duties that were never supplied in order to make a title fit a code. A plausible explanation is not the same as a documented classification decision.
Translation control: use AI for assisted review, not fabricated evidence. QA: require the source task description for every ambiguous assignment.
36. Translation memories can carry stale job-title mappings
A translation memory may suggest that a source title previously translated one way. That does not prove the occupational code is the same in the current record. Similar titles can occur in different domains, and organisational naming can change. The code and task description remain the stronger anchors.
Translation control: treat title translations as linguistic suggestions, not coding rules. QA: compare each accepted title against its current code and description.
37. Spreadsheets should store ISCO codes as text
CSV and spreadsheet workflows are common in labour statistics. If the code column is numeric, leading zeros can disappear and hierarchical codes can be reformatted. Sorting only one column can also attach translated titles to the wrong code.
Translation control: define code fields as text and use stable row keys. QA: round-trip a sample through export and re-import before full release.
38. Database joins require both code and version
Using “2211” as a key without recording the classification version can become ambiguous when systems mix standards or national variants. Strong schemas store the classification namespace and version alongside the code.
Translation control: keep metadata such as classification=ISCO and version=08 outside translatable text. QA: validate joins against the correct reference table.
39. APIs should separate code, source title and localized title
A useful multilingual API model might carry the canonical occupation code, an official or source-language label, and one or more localized display labels. That architecture prevents the translated phrase from becoming the database identity.
Translation control: localize display fields while preserving code keys. QA: switch locale and confirm the same record identifier is returned.
40. Search systems should index synonyms without rewriting the code
Users search occupations using everyday language, official titles, abbreviations and local variants. Search quality improves when systems index synonyms and translated labels while retaining the same ISCO code as the classification anchor. Vocabulary richness belongs in the retrieval layer, not in code mutation.
Translation control: expand searchable language around the stable code. QA: test multiple synonyms and confirm they resolve to the intended occupation.
41. This is where vocabulary knowledge supports classification
Occupational language is a vocabulary system: near-synonyms, technical terms, institutional labels, verbs describing work and culturally specific names all interact. eduKateSG’s protected Vocabulary Learning Hub helps explain how word meaning, context and lexical choice work. This article does not replace that hub; it applies those principles to the narrow problem of occupational-classification translation.
Translation control: use vocabulary analysis to clarify meaning, not to redefine the classification. QA: document domain-specific synonyms and false friends.
42. Worked example: “software engineer” in a bank
The employer is a financial institution, but the worker designs and develops software systems. The occupation should be classified from the software-development tasks, not from banking. The target language may use a term closer to “software developer” than “engineer.” That linguistic choice can be correct while the verified ISCO code remains unchanged.
Lesson: translate the role; do not translate the industry into the occupation field.
43. Worked example: “operations manager” who does not manage
A company calls a senior scheduler an “operations manager,” but the person does not direct a department, set budgets or coordinate broad organisational resources. The work mainly involves planning daily schedules and handling records. The title alone could mislead a coder into a management group. The task description provides the evidence needed to avoid that error.
Lesson: title prestige is not a classification rule.
44. Worked example: multilingual health survey
A survey asks respondents for job title and main tasks. One language uses a general word that could mean nurse, nursing aide or caregiver. The coder reviews the translated duties: medication administration, clinical monitoring and licensed nursing tasks. The tasks support the correct occupational interpretation. Without that second question, title-only translation would remain uncertain.
Lesson: preserve the survey evidence needed for coding.
45. Worked example: national code mapped to ISCO-08
A national labour dataset uses a domestic six-digit occupation code. An international report requires ISCO-08. The translation team translates titles and notes, but the statistical team applies the official national-to-ISCO correspondence table. Where one national category maps to several ISCO groups, additional information or an allocation rule is required.
Lesson: crosswalking is a classification process, not a bilingual dictionary exercise.
46. Worked example: historical census recoding
A 1990 census used an older occupational classification. The archive translation preserves the original code and title. A separate research dataset later maps those records to ISCO-08 for time-series analysis. Both versions are retained so users can distinguish what the census originally recorded from what analysts later harmonised.
Lesson: translation should not erase provenance.
47. Worked example: new digital title
A vacancy advert uses the title “AI prompt engineer.” The actual tasks include writing evaluation datasets, testing model outputs, scripting automation and designing quality-assurance procedures. Rather than inventing an ISCO category from the fashionable title, the coder compares the dominant duties with existing information-technology and analytical occupation definitions.
Lesson: new titles still require old-fashioned evidence: what work is actually performed?
48. Search intent: “translate occupation code”
A user typing this phrase may want to translate the title attached to a code, understand the code in another language, or convert a national classification to ISCO. These are different tasks. Translation changes language; classification mapping changes the coding system. A strong answer identifies which transformation is needed before touching the data.
Best practice: state the source classification, version, code and desired target output explicitly.
49. Search intent: “translate job title to ISCO code”
This query sounds simple but often lacks enough information. Job title alone can be ambiguous. The practical workflow is to collect main tasks and duties, identify the occupational family, check skill specialisation and use the official ISCO index and definitions to confirm the unit group. If a national code already exists, consult an authorised correspondence table rather than recoding from title alone.
50. Search intent: “what ISCO code is this job?”
The answer should begin with evidence, not confidence. Ask what the person actually does, what they supervise, what tools or knowledge they use, and whether the source already carries a national or international occupation code. When evidence is incomplete, report candidate groups or unresolved status rather than pretending certainty.
51. Reference route: the ILO ISCO-08 structure
The International Labour Organization’s official ISCO-08 reference site describes the four-level hierarchy, the 10 major groups, 43 sub-major groups, 130 minor groups and 436 unit groups, and provides structure, definitions and downloadable materials. Production translation and coding should use the authoritative definitions rather than copied lists from unknown websites.
52. Reference route: ILOSTAT classification guidance
The ILOSTAT occupational-classification guidance explains the hierarchy and skill-level framework used in international labour statistics. It is useful when translating statistical notes, metadata and analytical reports because it shows how occupation concepts fit into the broader measurement system.
53. Connection to the master translation architecture
This specialist article sits beneath eduKateSG’s Master Art of Translation — Technical Translation System. The master owns broad questions of terminology, specifications, QA and change control. This page owns a narrower problem: occupational titles, ISCO-08 codes, hierarchy and cross-language statistical integrity. Keeping that boundary clear prevents duplication with the master and with the separate industry-classification owner.
54. Release checklist for translated ISCO-08 data
Before release, confirm the classification namespace and version; preserve every code exactly; retain leading zeros; distinguish occupation from industry, qualification and employment status; verify title meaning against tasks; keep national codes separate from ISCO codes; use authorised crosswalks; preserve historical classifications; lock code fields in translation tools; store codes as text in spreadsheets; compare source and target code sets; check row relationships; and document any unresolved or manually coded cases.
Then perform ordinary linguistic QA on occupational titles, task descriptions, skill terms, institutional vocabulary and explanatory notes. Code integrity and language quality are separate controls. A target can pass one and fail the other.
55. Final rule: translate the work language, preserve the occupational identity
Occupation translation succeeds when a reader in another language can understand what a person does while statisticians, researchers and systems still recognise the same occupational category. ISCO-08 provides the shared classification layer; vocabulary and context provide the human meaning. Translate job titles and task descriptions carefully, preserve the assigned code, separate mapping from translation, and keep enough evidence that another reviewer can reconstruct why the occupation was classified the way it was.