If you are searching for how to translate COICOP codes, how to translate household expenditure categories, how to translate consumer price index spending groups, or how to preserve the meaning of household-consumption statistics across languages, the most important rule is that COICOP is a classification by purpose. Translate the category title and explanatory text, but do not translate or redesign the code itself. A COICOP code points to a defined consumption purpose inside the United Nations statistical framework.
This matters in household expenditure surveys, consumer price statistics, cost-of-living research, inflation analysis, national accounts, poverty studies, market research, public policy, social protection, household budgets, international comparison and multilingual statistical dashboards. A translation can read beautifully and still change the data if restaurant meals are grouped with food bought for home preparation, if school fees are confused with private tutoring, if a separately priced delivery service is attached to the product rather than the transport-service category, or if a translated label causes analysts to map a purchase to the wrong purpose.
This guide explains how to translate COICOP 2018 accurately. It shows how division, group, class and subclass hierarchy works; why purpose matters more than the physical object alone; how goods and services, package pricing, second-hand purchases and household context affect classification; how COICOP differs from product and industry classifications; and how to protect codes in multilingual spreadsheets, survey systems, APIs and reports. It is a specialist owner under eduKateSG’s master translation architecture, not a competing general translation hub.
1. COICOP classifies individual consumption according to purpose
COICOP stands for Classification of Individual Consumption According to Purpose. The United Nations Statistics Division describes COICOP 2018 as an international reference framework for grouping household consumption expenditure on goods and services into homogeneous categories according to the purpose those goods and services fulfil. The same physical product can therefore need different treatment from a classification based on production or product type.
Translation control: preserve the idea of purpose, not merely object name. QA: ask what need or household consumption purpose the expenditure serves.
2. COICOP 2018 is the current operational UN reference version
The United Nations Statistical Commission endorsed the revised COICOP in 2018. UNSD lists COICOP 2018 as operational and makes the structure and explanatory material available through its classification system. Translation projects should record that version explicitly because earlier COICOP structures and national adaptations can differ.
Translation control: retain “COICOP 2018” in metadata. QA: never mix labels from a previous version into a 2018 code table without a documented correspondence process.
3. The hierarchy runs from division to subclass
COICOP 2018 uses a hierarchy of divisions, groups, classes and subclasses. Codes become more detailed as digits are added. For example, a division can cover a broad purpose such as transport, while groups and classes separate vehicle purchases, operating costs, passenger transport and other services. The subclass provides the most detailed standard category.
Translation control: preserve code punctuation and hierarchy exactly. QA: confirm that translated labels remain attached to the correct level.
4. Division 01 shows why context matters
Division 01 covers food and non-alcoholic beverages purchased mainly for household consumption or preparation, while food and drink provided by restaurants and similar services are classified elsewhere. A translator who sees the word “food” and groups everything into the same category misses the purpose-and-service distinction that COICOP is designed to preserve.
Translation control: translate the purchase context as carefully as the food name. QA: distinguish groceries, food-serving services and separately priced delivery services.
5. Goods and services can serve the same broad need but use different codes
A household can meet a need by buying a good, paying for a service or purchasing a package that combines both. COICOP’s explanatory notes often distinguish these cases. The translator should therefore preserve whether the source says purchase, hire, repair, maintenance, subscription, admission, tuition, accommodation or another service relationship.
Translation control: protect transaction verbs and service nouns. QA: confirm that the target still tells the coder whether money was spent on a good or service.
6. Purpose is not the same as product identity
A product classification such as the UN Central Product Classification asks what the product is. COICOP asks what household-consumption purpose the expenditure serves. Those dimensions can be linked, but they are not interchangeable. eduKateSG therefore keeps this COICOP owner separate from the CPC owner in the same `Translate |` family.
Translation control: keep purpose codes and product codes in separate fields. QA: do not substitute a CPC code simply because the product name is known.
7. Purpose is not the same as industry
Industry classifications such as ISIC classify the economic activity of producers. COICOP classifies household expenditure by consumption purpose. A restaurant belongs to an industry; a meal purchased from a restaurant is a household-consumption expenditure. These can be connected analytically but answer different questions.
Translation control: distinguish producer activity from consumer purpose. QA: identify whether the dataset describes the establishment or household spending.
8. Consumer price index baskets often use COICOP or national adaptations
Many national statistical offices organise consumer price index weights and expenditure categories using COICOP-based structures. National versions can include local detail, renamed categories or additional coding layers. Translators should preserve the national code namespace while explaining its relationship to COICOP rather than assuming every CPI category is the international code.
Translation control: label national adaptations explicitly. QA: use the statistical office’s official correspondence table when mapping to COICOP 2018.
9. Household expenditure surveys need questionnaire language that preserves purpose
Survey responses are only as good as the questions. If a source asks how much a household spent on restaurant meals, school fees or vehicle maintenance, the translation must retain the same boundaries. A broader or narrower target phrase can move reported expenditure into another code before analysts ever see it.
Translation control: preserve inclusion and exclusion language. QA: test translated questions against the intended COICOP category.
10. “Food” is not one universal expenditure category
Food purchased for preparation at home, takeaway items, restaurant meals, school cafeteria purchases and food included inside accommodation or institutional packages can be treated differently depending on pricing and purpose. Translating every food-related line as one generic “food” category destroys the classification detail.
Translation control: keep preparation, service setting and pricing context. QA: use explanatory notes when the boundary is unclear.
11. Separately priced services are often decisive
COICOP explanatory notes repeatedly distinguish services included in a package from services priced separately. A delivery charge, parking fee, breakfast, maintenance service or accommodation add-on may move to another category when separately invoiced. Translation should preserve words such as “included,” “separately priced,” “bundled” and “optional.”
Translation control: treat pricing relationship as classification evidence. QA: check whether the target still indicates what the household paid for separately.
12. Package purchases need the source methodology
Package holidays, bundled communications services, accommodation packages and combined subscriptions can contain several components. Statistical treatment depends on the COICOP rules and the available price information. Translators should not split or merge the package merely because the target language describes components differently.
Translation control: preserve package boundaries from the source. QA: apply the survey or CPI methodology rather than linguistic intuition.
13. Durable, semi-durable, non-durable and service markers are analytical attributes
COICOP materials can mark categories according to durability or service status. These markers help economic analysis but are not ordinary adjectives to translate creatively. A durable good is a statistical concept based on expected use, not simply a product that feels “strong” or “long-lasting.”
Translation control: use standard national-accounts terminology. QA: preserve the marker independently from the category label.
14. Housing expenditure needs careful boundary language
Housing-related consumption can include actual rentals, maintenance, utilities and household services, while home purchase can involve capital rather than consumption treatment in national accounts. A translator who casually uses “housing cost” for every payment can obscure these distinctions.
Translation control: distinguish rent, purchase, maintenance, utility and service concepts. QA: keep the wording aligned with the specific COICOP class.
15. Utilities require service-specific terminology
Electricity, gas, water supply, refuse collection and related services may appear near one another in household budget data but serve distinct categories. “Utilities” is useful as a broad English umbrella but may be too vague as a translated statistical label.
Translation control: name the actual utility service at detailed levels. QA: ensure the category code and target noun remain aligned.
16. Health expenditure is not the same as health outcome
COICOP classifies household spending on health-related goods and services. It does not classify diseases, clinical procedures or health outcomes. Those belong to different systems. A medicine purchase, outpatient service and hospital service should not be translated into diagnostic language simply because all relate to health.
Translation control: preserve expenditure purpose and service setting. QA: do not mix COICOP with ICD, SNOMED CT or LOINC codes.
17. Transport illustrates purpose layers clearly
UNSD’s COICOP 2018 transport division distinguishes purchase of vehicles, operation of personal transport equipment, passenger transport services and transport services for goods. A household buying a car, paying for fuel, buying a bus ticket and paying a courier are all spending on transport-related purposes, but not the same purpose category.
Translation control: preserve whether expenditure concerns acquisition, operation, passenger movement or goods movement. QA: test category labels with real transaction examples.
18. Vehicle purchase and vehicle operation must remain separate
The purchase of a car is conceptually different from fuel, tyres, repairs, parking or tolls. A translated household-budget app may use one friendly “Car” heading, but its data layer still needs the detailed categories if the analysis depends on COICOP.
Translation control: localise user-facing headings without collapsing detailed codes. QA: compare child categories after interface translation.
19. Passenger transport and parcel delivery are different purposes
A taxi ride, train ticket, air fare, postal service, courier fee and separately priced delivery charge can all involve movement, but the household purpose differs. Translators should not render every transport-service category with a generic word meaning “delivery” or “transport.”
Translation control: keep people movement and goods movement distinct. QA: review verbs such as travel, carry, deliver and ship.
20. Information and communication categories have changed with digital life
COICOP 2018 includes modern information and communication categories such as software and communication services. Digital subscriptions can look like entertainment, information, software or communication depending on the service. Product branding alone may not reveal the statistical purpose.
Translation control: describe what the household buys and uses, not just the platform brand. QA: consult the current explanatory note for borderline digital services.
21. Software is not automatically recreation
COICOP 2018 has a category for software excluding games, while game software belongs elsewhere. Translating “app,” “software,” “game,” “subscription” or “digital service” too broadly can move spending across divisions.
Translation control: preserve functional purpose. QA: distinguish game content from general software even when both are downloaded through the same store.
22. Education expenditure needs level and service context
Households can spend on formal education, private lessons, tutoring, extracurricular activities, books, accommodation and transport. These expenditures do not all belong to one education category. The statistical treatment depends on what the service is and whether it is part of formal education or another household purpose.
Translation control: distinguish formal education from recreation or cultural lessons where the classification does. QA: do not map from the word “class” alone.
23. Recreation, sport and culture include diverse services
COICOP groups many leisure and cultural expenditures, but explanatory notes distinguish equipment, subscriptions, admissions, lessons and services. A dancing class taken as recreation may differ from formal education in a recognised programme. Translation should preserve the purpose described by the source transaction.
Translation control: keep “formal education” and “leisure lesson” conceptually distinct. QA: review ambiguous training and class purchases.
24. Restaurants and accommodation are service categories, not simply food and housing
Restaurant expenditure pays for food-serving services, while accommodation expenditure pays for lodging services. Meals included in a hotel price can be treated differently from separately priced meals. A translator needs to preserve whether the source amount is included, separate or part of a package.
Translation control: preserve service and pricing context. QA: compare invoice structure with the intended category.
25. Personal care categories need precise everyday vocabulary
Hairdressing, grooming products, personal effects and social-protection services can sit near one another at broad levels but differ in detailed purpose. Everyday vocabulary varies greatly across cultures, so the translator should work from definitions and examples rather than a single familiar label.
Translation control: use explanatory notes to settle ambiguous consumer terms. QA: test translated labels with representative purchases.
26. Insurance expenditure is not the insured object itself
A household may pay insurance related to housing, health, vehicles or travel. COICOP treats insurance service charges according to defined purposes and categories. Translating an insurance premium as though it were a purchase of the insured good can distort the expenditure structure.
Translation control: keep insurance service terminology explicit. QA: distinguish premium/service charge from repair, medical care or asset purchase.
27. Financial services need transaction-specific wording
Bank charges, investment services, credit services and other financial-service expenditures can be difficult to translate because everyday language may use one word such as “fee.” The statistical meaning depends on what service the household receives.
Translation control: translate the service function, not merely the word “fee.” QA: preserve whether the charge concerns payment, credit, investment or another service.
28. Second-hand goods require explicit status where the category distinguishes it
COICOP 2018 includes distinctions in some areas between new and second-hand purchases, such as motor cars. A translator who drops “used,” “pre-owned” or “second-hand” because it sounds commercially unattractive can change the statistical category.
Translation control: preserve condition descriptors when classification depends on them. QA: review retail euphemisms such as “certified pre-owned.”
29. Rental and purchase are different economic relationships
Hiring equipment, renting a vehicle, leasing a dwelling and purchasing an asset are not interchangeable. Some languages use broad verbs that can cover temporary use and ownership transfer. Statistical translation must make the transaction clear.
Translation control: distinguish buy, rent, hire, lease and subscription. QA: confirm whether ownership changes.
30. Repair and replacement are different purposes
Repairing an appliance, buying a replacement part and buying a new appliance can belong to different detailed categories. Translation that compresses all three into “appliance expense” makes the data easier to read but less useful statistically.
Translation control: preserve maintenance, repair, parts and new-purchase language. QA: check invoice-level examples.
31. Tips and gratuities can belong with the underlying service
COICOP explanatory notes can explicitly mention tips and gratuities within particular service categories. Translation should not automatically create a separate generic “tips” bucket if the source methodology assigns them to the service purchased.
Translation control: follow the classification rule rather than local budgeting habits. QA: preserve the relationship between gratuity and service.
32. Taxes, fees and compulsory payments need scope checks
Not every payment made by a household is individual consumption expenditure. Taxes, fines, social contributions and transfers can fall outside COICOP consumption categories or require specific treatment. Translators should not assume that every line in a household cash-flow survey is COICOP expenditure.
Translation control: preserve legal payment type. QA: confirm statistical scope before assigning a purpose code.
33. Reimbursements can change the analytic amount
Households may pay for health care, travel or work-related items and later receive reimbursement from an employer, insurer or government. The treatment depends on the survey and national-accounts methodology. Translation should retain whether an amount is gross payment, reimbursement or net household expenditure.
Translation control: preserve payer and reimbursement relationships. QA: check variable definitions before translating “spent” or “paid.”
34. Gifts and transfers need purpose and ownership context
A household can buy goods for another household, receive gifts or make cash transfers. The economic recording can depend on who acquires, pays for and consumes the item. Language that hides the giver, recipient or purchaser can make classification and accounting difficult.
Translation control: preserve actor roles. QA: keep transaction parties visible in survey instructions.
35. Household production for own consumption has special boundaries
Households may produce food or other goods for their own use and purchase services to process primary goods. COICOP 2018 includes categories relevant to some such services. Translators need to distinguish buying a finished product from paying someone to process a household-provided input.
Translation control: preserve who supplies the material and what service is purchased. QA: compare wording with inclusion and exclusion notes.
36. Informal-market purchases still need purpose classification
Households buy goods and services from shops, online platforms, informal sellers, neighbours and community providers. The seller’s formality does not by itself determine the COICOP purpose. A translator should preserve the item or service description even when the vendor category is vague.
Translation control: classify the consumption purpose separately from seller type. QA: avoid equating “informal” with “miscellaneous.”
37. E-commerce creates new labels, not necessarily new purposes
Online purchasing changes how households buy, but many consumption purposes remain familiar. A grocery ordered through an app is still food expenditure; a streaming subscription may belong to a digital service category; a separately priced delivery fee may require its own treatment. Translation should not create an “internet purchase” category merely because the channel changed.
Translation control: keep sales channel separate from consumption purpose. QA: review platform fees and bundled services carefully.
38. Subscriptions require service-purpose analysis
Households subscribe to software, entertainment, news, telecommunications, fitness, transport and many other services. “Subscription” describes payment form, not consumption purpose. The translated label should identify what service is received.
Translation control: translate the service function, not only the billing model. QA: check recurring charges against product descriptions.
39. Membership fees can serve different purposes
A household may pay membership fees to sports clubs, professional associations, cultural organisations or other bodies. The statistical treatment depends on the service or purpose represented by the payment. A generic translation of all membership fees into one target term can remove that context.
Translation control: retain the type of organisation and benefit. QA: distinguish consumption service from transfer or professional expense where relevant.
40. National adaptations need versioned crosswalks
Countries can extend COICOP with additional digits or local subcategories. Those national codes should not be presented as if they were universal UN subclasses. A crosswalk can connect them to the international structure, but the national detail may not map one-to-one.
Translation control: preserve the national namespace. QA: document the correspondence table and reference year.
41. Historical COICOP series need controlled correspondence
Long inflation and household-expenditure series may have been collected under earlier COICOP structures. Reclassifying them into COICOP 2018 can improve comparability but may require split weights, assumptions or unavailable detail. Translation should not silently make that analytical change.
Translation control: preserve original version labels. QA: distinguish translated historical data from harmonised series.
42. Translation memories can attach old labels to new codes
Classification titles repeat across versions, but some scopes and boundaries change. A translation memory can suggest a familiar label that no longer matches the current explanatory note. Accepting the old phrase without review can preserve language while changing statistical meaning.
Translation control: bind terminology to classification version. QA: review changed or new 2018 categories against current notes.
43. Machine translation should lock COICOP codes
Codes such as 07.2.4 or 11.2.0.3 contain punctuation and hierarchy that should never be reformatted by a language engine. The code belongs in protected structured data; the title and explanatory note can be translated around it.
Translation control: mask or lock codes before MT. QA: compare literal source and target code strings.
44. Generative AI should not classify from merchant name alone
A merchant can sell products serving many purposes. A supermarket sells groceries, household goods and sometimes services; an online marketplace sells almost everything. A language model asked to infer COICOP from merchant name may produce plausible but unreliable classifications.
Translation control: use transaction description and methodology, not brand familiarity. QA: require evidence for automated classification.
45. OCR can detach amounts from categories
Scanned household-budget tables contain short codes, labels, weights and monetary values. OCR may shift a decimal, merge a code with an amount or place a label on the wrong row. Translation can then faithfully reproduce the wrong relationship.
Translation control: preserve table structure and row identifiers. QA: reconcile code-label-amount triples after extraction.
46. Spreadsheets should store COICOP codes as text
COICOP codes contain leading zeros and periods. Numeric spreadsheet types can remove zeros or reinterpret the value as a decimal number. A code such as 01.1 is not the number 1.1; it is a hierarchical classification string.
Translation control: import and export code columns as text. QA: compare code length, punctuation and parent relationships after every round trip.
47. Sorting only translated labels can corrupt the dataset
If a translator alphabetically sorts the target-language label column without sorting the entire table, perfectly preserved codes can become attached to the wrong descriptions and weights. The result may look clean but be analytically unusable.
Translation control: use stable row keys and whole-table operations. QA: compare code-label-weight relationships before and after localisation.
48. APIs need canonical code plus localized label
A multilingual statistics API should expose the canonical COICOP code as a stable field and provide language-specific labels separately. The consumer can switch languages without changing the category identity used for aggregation or time-series analysis.
Translation control: localize display strings only. QA: request the same code in several locales and confirm the same hierarchy and data values return.
49. Search systems can use synonyms without changing classification
Households use many words for the same purchase: mobile phone plan, cellular service, telco bill; takeaway, takeout, food delivery; rent, housing lease. Search and coding assistance can index these synonyms while keeping the COICOP category stable.
Translation control: enrich retrieval vocabulary around the code. QA: test regional synonyms and false friends.
50. Vocabulary knowledge improves coding questions
Household-consumption language sits at the intersection of everyday vocabulary and statistical definition. eduKateSG’s protected Vocabulary Learning Hub explains how meaning, context and word choice work. This COICOP owner uses those principles narrowly: it helps translators distinguish consumer terms whose everyday similarity hides different statistical purposes.
51. Worked example: groceries versus restaurant meal
A household buys uncooked rice, vegetables and fish from a supermarket, then later pays for dinner at a restaurant. Both expenditures involve food, but the purpose and service context place them in different COICOP areas. The translation preserves “groceries for home preparation” and “restaurant meal” rather than flattening both to “food.”
Lesson: the noun alone does not determine the category.
52. Worked example: separately priced delivery
An online retailer sells household goods and charges a separate delivery fee. The product expenditure and the separately priced transport service may need distinct classification treatment. If the target invoice simply says “total purchase,” the translator has removed evidence needed for statistical coding.
Lesson: preserve the pricing relationship.
53. Worked example: driving lessons
A household pays for driving lessons and a driving test. The word “lessons” might tempt a translator to classify the spending under education. COICOP 2018’s transport structure places driving lessons, tests and licences within services related to personal transport equipment. The purpose is transport-related rather than formal education.
Lesson: classification purpose can differ from the everyday form of the service.
54. Worked example: music class versus formal music education
A child attends a weekend guitar lesson for recreation, while another student is enrolled in a formal music programme at an educational institution. Both involve learning music, but COICOP explanatory notes distinguish recreational or cultural lessons from formal education according to the service context.
Lesson: “class” is not enough; purpose and institutional context matter.
55. Search intent: “translate COICOP code”
A user may want the target-language title for a code, help understanding a household expenditure category, or a mapping from a national CPI classification to COICOP 2018. Those are different tasks. Code translation preserves the code and translates the official title; mapping requires an authorised correspondence table; classification of a raw purchase requires transaction evidence and coding rules.
56. Search intent: “which COICOP category is this expense?”
A reliable answer needs more than the merchant or product name. Ask what was bought, whether it was a good or service, the household purpose, whether any components were separately priced, whether the item was rented or purchased, and which COICOP version the dataset uses. When the evidence is incomplete, preserve uncertainty instead of guessing a precise subclass.
57. Reference route: United Nations Statistics Division
The official UNSD COICOP page provides the COICOP 2018 structure and revision materials. The UN classification detail pages also provide hierarchy and explanatory notes for individual codes. These are the correct reference points for production classification work rather than copied consumer-category lists on commercial websites.
58. Connection to the master translation architecture
This article sits beneath eduKateSG’s Master Art of Translation — Technical Translation System. The master owns broad terminology and QA principles. This page owns one narrower search intent: translating household-consumption classifications while preserving purpose, code hierarchy and statistical comparability. It does not replace the existing industry, procurement or product-classification owners.
59. Release checklist for translated COICOP 2018 data
Before release, record the COICOP version; preserve every code as text; keep division, group, class and subclass hierarchy intact; distinguish purpose from product and industry; retain good-versus-service status; preserve package and separately priced relationships; distinguish purchase, rental, repair and subscription; retain second-hand status where relevant; use national crosswalks for adaptations; protect codes in MT and CAT tools; verify code-label-amount relationships; and check explanatory notes for every ambiguous boundary.
Then perform linguistic QA on household vocabulary, transaction verbs, service labels, consumer terms, pricing language and statistical notes. A code can be perfectly preserved while the translated label points users toward the wrong purpose, so both layers must pass.
60. Final rule: translate the purchase language, preserve the consumption purpose
Households buy millions of different things, but COICOP makes those purchases analytically useful by grouping them according to purpose. Translation should make the categories understandable without turning them into a different classification. Preserve the COICOP 2018 code, translate the label and explanatory note accurately, retain the transaction context that determines purpose, and keep mappings to national systems explicit. The result is multilingual household-spending data that still measures the same economic behaviour.