best practice guidelines international classification usd ONU 2022

Best Practice Guidelines for Developing International Statistical Classifications October 2022 1

Best Practice Guidelines for Developing International Statistical Classifications

Best Practice Guidelines for Developing International Statistical Classifications October 2022 2

Abbreviations

CPC Central Product Classification ICD International Classification of Diseases
ILO International Labour Organization
ISCO International Standard Classification of Occupations ISIC International Standard Industrial Classification of All Economic Activities
ISO International Organisation for Standardisation UNESCO United Nations Educational, Scientific and Cultural Organisation UNSD United Nations Statistics Division WHO World Health Organization

Best Practice Guidelines for Developing International Statistical Classifications October 2022 3

Contents

Introduction ................................................................................................................................ 4 Background ................................................................................................................................ 4 Defining a statistical classification .............................................................................................. 5 Principles to consider when developing an international statistical classification ........................ 7 Components of a Classification .................................................................................................13 Other Issues/Definitions ............................................................................................................18 Appendix One: Checklist for Developing a Statistical Classification ..........................................20

Best Practice Guidelines for Developing International Statistical Classifications October 2022 4

Introduction

The purpose of this document is to provide guidelines for best practice in the development, maintenance and implementation of international statistical classifications. The best practice guidelines contained in this document have been developed for use by international agencies, national statistical agencies, and other organisations that may develop, maintain and implement statistical classifications.

Best practice is defined in this document as the general principles and guidelines used in the creation of international statistical classifications. It includes:

• defining what is an international statistical classification; • components of an international statistical classification; • guidelines on principles to consider when developing an international statistical classification.

Background

Why have international statistical classifications?

It is a fundamental need for any statistical system to have standard concepts, definitions and classifications. This is enshrined in Principle 9 of the Principles of Official Statistics1 which states: “The use by statistical agencies in each country of international concepts, classifications and methods promotes the consistency and efficiency of statistical systems at all official levels.” International statistical classifications are developed and adopted by international agencies to ensure that there is a standardised and consistent approach to classifying statistics and data. The aim is to provide a basis for:

• statistics and data that are reasonably comparable between countries; • developing national classifications for the same variable/characteristics.

International statistical classifications are intended to group and organise information meaningfully and systematically, in exhaustive and structured sets of categories that are defined according to a set of criteria for similarity. A primary purpose of an international statistical classification is to provide a simplification of the real world and to provide a useful framework for collecting, organising and analysing statistical and administrative data , as well as providing a framework for international comparability of and reporting on statistics and data.

International statistical classifications are also developed to support policy making at the global level e.g., to support the Sustainable Developmentg Goals (SDGs)and to facilitate the collection and organisation of statistics.

•

To coordinate the international work on statistical classifications, the United Nations Statistical Commission mandates that a Committee of Experts on International Statistical Classifications (CEISC) be “the central coordination body for the current and future work on classifications that

1 https://unstats.un.org/unsd/dnss/gp/FP-Rev2013-E.pdf Best Practice Guidelines for Developing International Statistical Classifications October 2022 5

are the responsibility of the United Nations Statistics Division, and for the coordination and review of other classifications that are the responsibility of other international organizations and that have been proposed for adoption by the Statistical Commission.”2

It is important to note that international statistical classifications developed by international organisations other than the United Nations Statistics Division will, as well as being approved by the CEISC and United Nations Statistical Commission, also be officially approved and adopted through the relevant mechanisms of those organisations, for example the UNESCO General Conference of Member States, or an International Conference of Labour Statisticians authorised by the Governing Body of the International Labour Organization (ILO).

Defining a statistical classification

This part of the document defines a statistical classification and introduces the essential components and guidelines for its use. For the purposes of best practice, the definition below applies to all classifications used in the production of official statistics or data at the international, regional and national level.

What is a statistical classification?

The following definition should be used by national statistical agencies:

“A statistical classification is a set of categories which may be assigned to one or more variables registered in statistical surveys or administrative files and used in the production and dissemination of statistics. The categories are defined in terms of one or more characteristics of a particular population of units of observation. A statistical classification may have a flat, linear structure or may be hierarchically structured, such that all categories at lower levels are sub-categories of a category at the next level up. The categories at each level of a statistical classification structure must be mutually exclusive and jointly exhaustive of all objects in the population of interest.”

Guideline

A statistical classification is a set of discrete, exhaustive and mutually exclusive categories which can be assigned to one or more variables used in the collection and presentation of official statistics and data, and which describe the characteristics of a particular population.

Practices and principles of statistical classifications

The United Nations Statistical Commission endorsed a set of principles for statistical classifications and how they are developed or revised. (See 'Standard Statistical Classifications: Basic Principles' by Eivind Hoffmann, Bureau of Statistics, International Labour Office and Mary Chamie, United Nations Statistics Division, 1999). These principles should be read in conjunction with this document.

Essential components of a statistical classification

The essential components of a statistical classification are: • a consistent conceptual basis; • a flat or hierarchic structure;

2 https://unstats.un.org/UNSDWebsite/statcom/session_52/documents/BG-3k-TOR-UNCEISC-E.pdf Best Practice Guidelines for Developing International Statistical Classifications October 2022 6

• categories that are mutually exclusive and exhaustive; • definitions that are clear and unambiguous, and which define the content of each category; • that it is up-to-date and relevant; • that it is robust enough to last for a period of time; • that it meets user needs; • that it provides comparability over time and between collections; • that it provides guidelines for coding and output of data collected using it.

A statistical classification is one that follows prescribed rules and guidelines which are generally recommended and accepted. They ensure that the information is classified consistently and can be developed by any international agency, who will also act as custodians of that classification. The essential components listed above are a requirement for all international statistical classifications but may be waived, as appropriate, when the classification is deemed to be for analytical purposes only.

For example, the International Standard Classification of Occupations (ISCO) is developed by the International Labour Organization (ILO), which is also the custodian of ISCO; the International Classification of Diseases (ICD) is developed by the World Health Organization (WHO), which is also the custodian of ICD.

These rules and guidelines are also applicable for regional and national statistical agencies in developing statistical classifications.

Best Practice Guidelines for Developing International Statistical Classifications October 2022 7

Principles to consider when developing an international statistical classification

This section provides guidelines on what principles to consider when developing an international statistical classification. It is important to recognised that as classifications move from hardcopy publications to web enabled or digital format, many of the methods required for developing an international statistical classification can be dealt with in alternative ways but it is up to individual agencies to make that determination.

Custodians The custodianship of international statistical classifications generally resides with the United Nations Statistics Division. A number of major international classifications are, however, under the custodianship of key international agencies e.g. WHO, ILO, UNESCO, ISO. In consultation with international and multi-national agencies, including national statistical offices and/or thematic experts, the custodians of international statistical classifications are responsible for the development and maintenance of the classifications they look after. The custodians are also responsible for supporting the implementation of international statistical classifications in relevant countries, and in national statistical and administrative agencies. Custodians are required to present international statistical classifications to the UN Committee of Experts on International Statistical Classifications before their finalisation through the UN Statistical.

UN Committee of Experts on International Statistical Classifications

The role of the Committee is .

(a) Promoting the adoption of best practices in the development of international statistical classifications; (b) Coordinating work on those classifications that are members of the International Family of Statistical Classifications, including examining changes to classifications and clarifying their relationship to other international statistical classifications; © Promoting the adoption of the classifications of the International Family of Statistical Classifications within regional and national statistical systems to improve the international comparability of data; (d) Assessing the appropriateness of requests for inclusion of classifications into the International Family of Statistical Classifications; (e) Ensuring that international statistical classifications are suitable for use in statistical and data activities in which they are to be applied and that they promote conceptual and methodological coherence, and best practice in the statistical system, as well as reflecting practical considerations in data collection and compilation; (f) Developing a strategic approach towards the future development of the International Family of Statistical Classifications. The Committee will also provide guidance to the United Nations Statistics Division and other custodians of international standard classifications on technical matters of classification revisions or development, as well as strategic planning for classifications work, if so requested.3

3 https://unstats.un.org/UNSDWebsite/statcom/session_52/documents/BG-3k-TOR-UNCEISC-E.pdf Best Practice Guidelines for Developing International Statistical Classifications October 2022 8

Conceptual basis

It is important that an international statistical classification is based upon sound and agreed concepts and principles. The conceptual basis of the classification should be detailed in the explanatory notes and explain why the conceptual approaches taken have in fact been undertaken.

The conceptual basis should be well defined and documented to enable users to understand what the classification is about and should be used for categorising, interpreting and structuring the classification.

The conceptual basis is an essential component in enabling the creation of a sound and logical structure and for making sense of the classification. It may be based on principles or concepts developed through international collaboration, and the production of an agreed international standard, and/or through stakeholder consultation or agreement between national statistical agencies.

Related and supporting concepts should also be well-defined whether or not they are directly related to the classification or form part of a supporting conceptual framework for building a data narrative.

An analytical classification will often be built using different concepts at each level, layered one upon the other using a sequential code pattern

Guideline: The creation or development of a conceptual basis is mandatory for all international statistical classifications. All concepts must be clearly and well-defined.

Classification Structures

Classifications are either structured as a flat classifications (a simple listing of categories) or as an hierarchic classification (with a logical sequential hierarchy of categories ranging from detailed to broad levels).

There are no really hard and fast rules for when to use which type of classification structure. However, the structure should ensure that the most detailed categories are at the bottom or lowest level of the hierarchical classification.

(a) Flat

A flat classification has only one level, i.e. it is a listing of categories. Flat classifications are usually developed when there is no need to group categories into more aggregate groups. However, the categories must be mutually exclusive and the classification must be exhaustive.

An example of a flat classification is:

Sex

A flat classification may also be referred to as a linear classification.

Best Practice Guidelines for Developing International Statistical Classifications October 2022 9

Guideline: A flat classification structure should be used when a simple listing is required or when there is no requirement to aggregate or group categories into categories that may be useful for description or analysis alone in combination with other variables.

A flat classification can also form the basis for disaggregated, aggregated or thematic views of categories and concepts if there is user demand to do so.

(b) Hierarchic

A hierarchic classification is a classification with more than one level of aggregation. These are structured with the most general or broad categories at the top and the most detailed categories at the bottom. Depending on the descriptive and analytical needs, each level can be used when recording a value for the variable e.g., in a survey response or an administrative record.

Examples of hierarchic classifications are:

International Standard Industrial Classification (ISIC); Central Product Classification (CPC).

Guideline: A hierarchic classification structure should be used when there is a requirement to aggregate or group categories into categories that are sufficient for the descriptive or analytical needs, alone or in combination with other variables, or when a classification is to be used by several statistical domains with different needs of aggregation. A hierarchic classification can be input only focussed or output focussed such that the levels of aggregation can be treated differently depending on the collection and output requirements.

Best Practice Guidelines for Developing International Statistical Classifications October 2022 10

Classification Types

There are two types of statistical classifications: international statistical classifications, and national or regional statistical classifications.

(a) International Classifications

An international statistical classification is one developed by an international agency such as the United Nations Statistical Division (UNSD), International Standards Organisation (ISO), International Labour Organization (ILO), United Nations Educational, Scientific and Cultural Organisation (UNESCO), or World Health Organization (WHO).

The aim of international statistical classifications is to provide a common framework for collecting and organising information about a particular statistical system, concept or variable. Their use, either directly or through national adaptations, facilitates the exchange and comparability of official statistics, data and other information between countries. These classifications have generally been developed through extensive international consultation and have achieved broad acceptance and official agreement for use.

Countries should be able to report in international categories at least at the higher levels of the international statistical classification.

Guideline: An international statistical classification may require adaptation to meet country specific conditions as it is not always possible that the classification can be used as it was developed i.e. there may be categories defined for international use which do not apply in country specific circumstances, or there may be country specific circumstances which are not catered for in the international statistical classification.

The definition of correspondences (which map or link classifications together) are mandatory between international statitstical classifications within the same subject/topic/family, as they facilitate international reporting and enable time-series management, for example ISCO88 to ISCO08, or ISIC Rev 3 to ISIC Rev 4. However this is not a prerequisite to the classification being recognised as an international standard.

Correspondences between international statistical classifications in different subjects/topics, such as ISIC to CPC, or ISCED to ISCO, are advisable but not mandatory.

(b) National or regional classifications

National or regional classifications classifications are usually based upon an international
classification. They may be developed by:

• applying the concepts of the international statistical classification in a more rigid or alternative way to produce a different classification hierarchy or structure; • adopting the international statistical classification structure and categories at the higher levels, and then adding additional lower level detail for regional or national purposes; • rearranging or aggregating parts of one or more international statistical classifications to form a new variation of that classification.

Best Practice Guidelines for Developing International Statistical Classifications October 2022 11

Mutual Exclusivity

The categories in a statistical classification need to be mutually exclusive of items at the same level of the classification i.e. each member of the population of primary units must only be classified to one category; and it must be possible to classify all units, or survey responses etc, to a category in the classification.

A classification with categories which are not mutually exclusive will confuse users and not enable the statistical classification to be accurately and consistently used.

Analytical classifications are not bound by this requirement as they are often not for collection or output purposes for official statistics. Examples of an analytical classification are the Classification of Statistical Activities (CSA), or the Classification of Broad Economic Categories (BEC). As a result, an item can be classified in several areas and users can decide where to place it according to their specific need.

Explanatory notes and guidance to users within the supporting metadata, for example the use of definitional text, inclusions and/or exclusions, are important. However, this approach does not represent a break from the Best Practice Guidelines when assessing a classification for inclusion in the International Family of Classifications.

Guideline: Mutual exclusivity is mandatory for all statistical classifications.

Exhaustiveness

An international statisticalclassification should only be exhaustive for all possible values that the variable can take for the primary units for which the classification represents. For coding (recording) it should be noted that surplus or unnecessary categories often hamper the effectiveness and usefulness of the classification, for example not further defined categories which have been added to a classification structure solely for hierarchic coding purposes

7 Statistical Feasibility

The statistical feasibility of an international statistical classification means that it is possible to effectively, accurately and consistently distinguish between the categories in the classification on the basis of the information available, eg as responses to questions that can be reasonably asked in statistical surveys or on administrative forms. To ensure suitability of the classification for use in developed and developing countries the classification needs to be widely discussed and tested. Statistical feasibility is not the same as statistical significance. An international statistical classification may include categories which are not statistically significant for all countries and it is up to individual countries to assess whether to adopt the categories of an international statistical classification for their own need.

Classification Units/Statistical Units

The classification unit is the basic unit to be classified in the classification (e.g. the job in an occupation classification such as ISCO, or the activities in an industrial classification such as ISIC.)

Best Practice Guidelines for Developing International Statistical Classifications October 2022 12

Statistical units are the units of observation or measurement for which data are collected or derived. Statistical units can be people, products, businesses, geographic areas, events, jobs etc. This may or may not be the same as the unit of classification.

Statistical units may also be referred to as units of observation which are entities on which information is received and statistics are compiled..

Note that the term reporting unit is used to mean a unit that supplies the data for a given survey instance.

Time-Series Comparability

In developing and using a statistical classification, consideration must be given to ensuring comparability over time between current and previous versions of the classifications. Important time-series breaks should be avoided but are often necessary when this reflects changes to the reality that the classification should mirror. Time-series can be managed through the use of correspondences (which map or link together different versions of classifications), back-casting or dual-coding.

The use of application program interfaces (apis) for system to system integration can be a way of mitigating time-series issues.

Best Practice Guidelines for Developing International Statistical Classifications October 2022 13

Components of an International Statistical Classification

This section outlines the components required when creating an international statistical classification and these components are equally relevant to national or analytical classifications

Classification Title

The classification title is the formal title associated with the classification. Examples of naming are:

International Standard Industrial Classification of All Economic Activities(ISIC) International Standard Classification of Education (ISCED) Central Product Classification (CPC) Standard International Trade Classification (SITC)

Guideline: Title conventions are as listed below:

“International Standard Classification of....” “Standard International ...... Classification”

Classification Abbreviation

This is the abbreviation associated with the classification usually made up of the first letter of each word in the full title. Examples would be:

International Standard Industrial Classification (ISIC) International Standard Classification of Occupations (ISCO) Central Product Classification (CPC) Standard International Trade Classification (SITC)

Guideline: The use of a year value as part of the abbreviation is discouraged as the versioning for a classification should be sufficient eg CPC V2.0 compared to CPC22

Classification Version

A classification version is a set of mutually exclusive categories representing the classification variable and is valid from a particular date for a period of time that may or may not be specified. For a hierarchic statistical classification, each level of the classification needs to be mutually exclusive and stand-alone. For analytical classifications, levels do not need to be mutually exclusive nor stand-alone.

When a new version of a classification is developed, it supersedes the previous version(s of the classification. Usually new versions are created with a frequency that balances the need to retain comparability over time with the need for the classification to represent the contemporary reality at the time when the observations are made.

Best Practice Guidelines for Developing International Statistical Classifications October 2022 14

It is also important to consider the international reporting requirements for countries when creating new versions of classifications (whether national, statistical, or analytical) and for when they need to be implemented.

Guideline: A new classification version (whether national, statistical, or analytical) will be developed when the scope, concepts, or structure (either in part or in whole) change, or when there is the addition of new or a deletion of old categories, modifications to descriptive definitions or the updating of classification category labels The addition of case law or changes in explanatory notes or inclusion/exclusion text will not necessarily result in the creation of a new version. Versioning protocols will be determined by the custodian of the international statistical classification. Classification Levels

A classification structure can be composed of one (flat classification) or several levels (hierarchic classification) of aggregation. The bottom level of a hierarchic statistical classification is always the most detailed level i.e., has the most precise information (detailed values) for the variable classifying the statistical unit. Categories at this level are aggregated into broader categories in the classification.

For analytical classifications, the structure may not necessarily form an aggregation, more a layering of concepts or reflect a conceptual framework.

There should be sufficient levels in a classification to meet the range of statistical and data needs which the classification is intended to fulfil.

Guideline: The number of levels defined should be kept to the minimum to reflect the detail
needed for different types of description and analysis. In the national context additional levels may be added to an international classification. However, with the increased use of digitalisation of international statistical classifications greater complexity can be more easily managed than with traditional hardcopy publications.

Coding Structure

Classification codes consist of one or more alphabetical or numerical characters assigned to a category in a classification. A code may consist of a combination of alphabetical or numerical characters.

There are no standard criteria for when to use alphabetical codes versus numerical codes. Numerical codes are more useful particularly when creating logical and sequential hierarchical classifications. However leading zeros might be required to ensure a standard code pattern can be stored within computerised classification management systems.

Code patterns need to be consistent and logical for each level they are used. For level one of a hierarchic classification the code pattern should be the first position indicating the most aggregate level e.g. 1 for the first most aggregate group; for level 2 it should be 12, for level 3 it should be 123 - i.e. a logical hierarchic structure. This does not preclude using other patterns but with these it may be difficult to link one level to another (e.g. the use of roman numerals followed by alpha characters followed by alpha-numeric characters would not be advisable.)

Best Practice Guidelines for Developing International Statistical Classifications October 2022 15

Sometimes codes will have '.' in them. An example would be in the Harmonised System Classification. The '.' provides a delimiter at a specific level. It isn't required generally.

The code structure should be robust enough that the addition of new codes can be done in the future. However, the situation may arise that the expansion of the code requires additional levels being added to the classification

In a flat classification the codes may either be sequential numbers or a combination of letters that may serve as easily understandable initials for the category. For example, for a flat list of countries names, two (or three) -character alphabetic codes for representing the names of countries can be used.

Guideline: Every category in a classification must have a code and the code structures need to be consistent and logical for each level they are used. The coding structure and the coding conventions used in the classification should be clearly explained in the classification supporting documents.

Category descriptors or labels

Category descriptors or labels (or titles) are usually a one line text describing the category of the classification. This text serves as the official title of the category.

The descriptor or label should be unique within the classification and meaningful, to illustrate with certainty the exact content of the category. Each descriptor should be meaningful on its own i.e., no further information should be needed to see that this category has content different from all others. When categories at different levels in a classification hierarchy have identical content (for example when a category at level 2 is not further disaggregated at level 3) they should generally have the same descriptor. Descriptors may be repeated in an analytical classification regardless.

Guidelines on the descriptors/labels/ titles? For example the use of commas; semicolon? Should the titles be (to the extent possible) comprehensive? meaning they should not label only part of a category content?

Definitional descriptions/Explanatory notes

Definitional descriptions or explanatory notes provide supporting information about the classification category. Often they are statements which clearly define the category or they may assist users in determining the boundaries of the category.

Explanatory notes may explain the content by giving examples of inclusions and exclusions or provide rules or guidelines for how to use that particular category.

Guideline: Definitional descriptions or explanatory notes are in general recommended as they provide a better understanding of the categories. However, they are not mandatory especially when the content is self-explanatory.

Coding index

A coding index reflects responses obtained from statistical or administrative collections. These are stored with the classification category to which they belong.
Best Practice Guidelines for Developing International Statistical Classifications October 2022 16

A coding index is used to allocate a classification code to a survey response using a computerised coding tool. It usually contains words obtained from a variety of sources which include survey responses and write-ins in administrative forms, through machine-learning or use of AI. A coding index with precise rules on how it should be used in a coding operation should be constructed for computer assisted coding/automatic coding as well as for manual coding.

The coding index should be suitable for the application of exact or fuzzy matching methodologies, However, the use of web services/apis or search and discovery tools may negate the need to produce a coding index or utilise coding tools.

Guidline?

Alphabetic Indexes

An alphabetic index may be included at the discretion of the classification custodian depending on the methodology used for publication or dissemination, usually in support of an electronic copy such as a .pdf. The need for an alphabetic index has diminished with the advances of web search and discovery tools. The index if created should provide a listing of the classification descriptors and related synonyms or phrases relevant to the classification categories. s

Guideline? The creation of an alphabetic index is generally recommended to assist search and discovery of classification content. It may also form the basis for a coding index

Residual categories

Residual categories are designed to classify units that do not fit into the other, fully specified classification categories. They may be required to ensure that all categories in a flat classification, or at all levels in a hierarchical classification, are jointly exhaustive of all units in the population. The need for such categories is dependent on whether or not the other categories specified are exhaustive and is determined at the discretion of the classification developer.

When residual categories are used within hierarchical classifications, they usually inherit the name of a higher-level category qualified by a term such as ‘other’ or ‘not elsewhere classified’. For example, ISCO-08 Sub-major Group 22, Health Professionals includes a Minor Group 226, Other Health Professionals, which in turn includes a Unit Group 2269, Health Professionals not Elsewhere Classified. It is good practice to use a consistent naming convention for residual categories at each level. When using numerical codes, It is general practice to allocate a code ending in ‘9’ (or ‘99’ depending on the number of digit of the classification level) to represent ‘not elsewhere classified’ categories .

Vague and imprecise responses to questions in surveys should not generally be coded to residual categories, but should be coded to the level of detail in hierarchy supported by the information contained in the response. In such cases it may be necessary to create supplementary categories and codes that are not formally part of the classification structure. Categories such as ‘response outside scope’ or ‘response unidentifiable’ may be needed to cover certain instances.

Consideration to including categories in a statistical classification such as ‘refused to answer’, ‘object to answering’, ‘don’t know’ or ‘not applicable’ needs to be undertaken in conjunction with survey questionnaire designers and the classification custodians.

Best Practice Guidelines for Developing International Statistical Classifications October 2022 17

Guideline? The conventions for classifying residual categories excluding ‘not further defined’ or ‘not elsewhere classified’ should be made explicit in each classification, and residual categories should be noted as primarily for operational purposes only. Supplementary or residual codes should reflect the coding structure applied to the classification. Rules may be described for aggregation of residual categories for data output purposes.

Best Practice Guidelines for Developing International Statistical Classifications October 2022 18

Other Issues/Definitions

Correspondence

A correspondence provides a link between different versions of a classification or between different classifications. A correspondence details how a category in one classification relates, or links to, the new/other classification. Sometimes the category doesn't change across classifications, sometimes a category splits into several categories in the new/other classification, and sometimes there is no corresponding category. For the latter situation a decision needs to made as to the presentation in such instances with possible options being to exclude altogether or include and map to ‘no equivalent category’

A correspondence can consist of the following relationships:

One to One (1:1) One to many (1:m) Many to One (m:1) Many to many (m:m)

Guideline: A correspondence is a mandatory requirement between different versions of international classifications e.g. ISIC Rev 3 to ISIC Rev 4, or CPC V1.0 to CPC V2.0.

A correspondence between different international classifications for the same or related variables/characteristics is desirable where this is relevant.

However, the mapping of different international classifications to each other is also dependent on user need i.e. there is no point creating a correspondence unless there is a demand for it .

Units of Measure

Units of measure are often associated with statistical classifications used for the production of trade and/or commodity data. Units of measure are a way of quantifying the weights and measures appropriate to the units being classified, and are part of the basic category definition. The units usually correspond to international standard codes and definitions for weights and measure based on ISO 1000 or the International System of Units (SI). The units may be associated with the classification units or the data produced from the classification.

Coding Decisions/Case Law/Determinations

It is essential to record and make easily available any operational coding decisions, case laws or determinations which may assist users of the classification. There may exist a process whereby users can raise queries that need a resolution which may also assist in clarifying the scope of categories. These decisions should be incorporated into the rules for using any coding index, coding tool or web coding service, the guidelines, standards or manuals supporting the classification, and they may provide an agreed interpretation of:

Best Practice Guidelines for Developing International Statistical Classifications October 2022 19

• how to classify new situations/responses (e.g., jobs with new combinations of tasks and duties, new types of production activities or new products or services) which have emerged since the classification was released (and should be noted for inclusion at the next review), or have resulted from updates of explanatory notes; • how to classify difficult or unusual situations/responses that the existing descriptive definitions do not easily resolve; • how to classify categories for which there have been varied interpretations by users (i.e., to get consistent coding); • relevant administrative or legal interpretations through case law or legislation in a specific country.

Best Practice Guidelines for Developing International Statistical Classifications October 2022 20

Appendix One: Requirements for Developing Classification

Status of the classification Should the classification be an international statistical classification, a regional or national statistical classification or a collection specific classification? What should the name be?
Which data collections will use this classification? Is it specific to a particular statistical or data collection? Will it be used in administrative data collections as well as sample surveys and censuses? What are the possible non-statistical applications?
What are the underlying concepts used in this classification? How are the concepts defined? What are the statistical units being classified? What are the reporting units? Are there other concepts which are closely linked to the classification?
Scope of the classification What population of units does the classification cover?
Primary uses of the classification Is the primary use of the classification as a collection and processing tool? Is the primary use of the classification as a tool for statistical analysis? Will the assigned classification category influence or determine administrative decisions affecting the unit being classified?
User Consultation Should a reference group or task team of key users and/or subject matter specialists be established? (This group should be consulted on content, scope and structure). Should a statistical advisory group of stakeholders be established? (This group should be consulted on the relevance of the statistics that may be produced with the classification). How will conflicting user requirements or applications be resolved?
What are the classification/similarity criteria? Are they compatible? Why were they chosen? What compromises were made to meet special requirements of particular users?
Structure of the classification Does the structure have an appropriate number of levels?

What is the compatibility with other statistical concepts and classifications, and the comparability with international standards?

Are the proposed categories well defined? Best Practice Guidelines for Developing International Statistical Classifications October 2022 21

Are they mutually exclusive and exhaustive when considering the descriptive definitions and explanatory notes, as well as the coding instructions for an international statistical classification? Are the names chosen for categories precise, appropriate, and meaningful?

Appropriateness of code structure Is the code structure appropriate? Are there any special code structures, conventions proposed or required? Are supplementary codes required? Are residual categories needed, specified and used appropriately?
Relationship with other classifications Are there any (other) relevant international standards? What is the relationship between the classification and any other classification or statistical framework?

12