|
|
|
|
| Methodological issues concerning the development, use, maintenance and revision of statistical classifications |
| |
The main tasks of the custodian of a national statistical classification are related to its development, use, maintenance and revision.
Statistical classifications are not necessarily constructed by statisticians but by persons responsible for administering public policies (e.g. custom regulations, criminal legislation), providing public services (e.g. job placement, education, health services) or analysing social, economic or natural phenomena. When adopting and adapting such value-sets for their use in statistical inquiries, statisticians must understand whether and how the various methodological issues have been tackled and contribute to finding improved solutions to them where necessary and possible.
- Definitions
- Development
- Use
- Maintenance
- Revision
Definitions
- A statistical classification is a set of discrete values that can be assigned to specific variables which are to be measured in a statistical survey, or registered in administrative files to be used as a basis for statistics.
- Developing a classification involves those activities needed to create a classification.
- Using a classification involves the recording of the appropriate value of the value set for a particular unit.
- Maintaining a classification involves (a) the correction of errors made in the construction of the classification (value set) and associated coding tools; and (b) up-dating of the descriptions of the value-sets, dividing lines between groups and coding tools of the classification when previously unknown or genuinely new types of primary units emerge or are discovered, or as new information is obtained about existing types.
- Revising a classification involves a complete review of users' needs as well as the conceptual basis and the users' tools. Such reviews should only be undertaken at long intervals (10-15 years) or if there is compelling evidence that revisions are necessary. Such evidence may come from national users or from international developments.
Development
When developing the classification, the custodian should consider
- The users' requirements:
o how to determine who the main users are;
o how to determine what the different users would like the classification to do;
o how to balance the different users' requirements against each other or make a choice between them when they are contradictory.
- The main conceptual issues, given the users' requirements:
o What are the (main) variable(s) for which the value-set is valid? Example: The variable to which ISCO-88 is applied is "main tasks and duties" of work performed, and not other aspects of work, such as exposure to hazardous substances or uncomfortable working conditions. Thus the variable "occupation" can be said to be defined as "the main tasks and duties of work performed".
o What are the primary units for which the we can measure the main variable(s)? e.g. "jobs" are the primary units for ISCO-88. A "job" is defined as a set of tasks and duties (designed to be) performed by one person
o What are the rules for linking other units to the primary units so that these other units can be assigned values of the classification variable? Example: "Persons" can only be assigned an "occupational group" through their link to a past, present or future job. They can only be given an "industry" code by being employed by an establishment.
o What are the conceptual rules for identifying the same value of the variable? Example: For ISCO-88 the rule is "a set of jobs whose main tasks and duties are characterized by a high degree of similarity: they constitute an occupation", which is the most detailed element in the value-set for an occupational classification.
o What are the similarity criteria used to define higher level categories (aggregated value-sets) in hierarchical classifications? Example: In ISCO-88, the main similarity criteria are "skill level" and "skill specialisations" needed to carry out the tasks and duties of jobs.
- The collection of information: how to collect the information which describes the defined categories and the dividing lines between them. Example: for an occupational classification one needs to collect information about the main tasks of jobs in the different occupations for the whole range of work situations that can be found in establishments of different size and in different industries. If one wants to be able to apply different similarity criteria to define alternative aggregated value sets, then all the corresponding information must be collected.
Use
When using the classification, the following issues are important for statistical surveys and administrative records:
- Respondent, interviewer or expert coding:
o With respondent coding the respondent is presented with a value-set and asked to indicate the value which best applies to him/her.
o With interviewer coding it is the interviewer who determines the correct value on the basis of the respondent's information.
o With expert coding, the respondent or interviewer writes down the information from which someone (the expert) later will determine the correct value.
- What question(s) to ask, which depends on the choice of coder:
o Respondents who code, i.e. select a category, need to be told on what basis they should choose one of the possible categories.
o Interviewers who code need to get the information relevant for coding by themselves or by experts. Example: With interviewer or expert coding, the best basis for determining the occupational code is to ask questions about the occupational title as well as about the main tasks and duties.-
- Instructions about which response elements to use and how to use them ensure that interviewers record the most important response elements and that coders use them consistently. The coding process should correctly retain as much of the information provided by respondents as possible. Example: The ILO recommends that when coding "occupation" one should code to the most detailed level in the classification that is supported by the information provided. Title, task etc. should be used according to clearly specified rules.
- How to develop appropriate coding tools? This concerns the development of coding indexes reflecting the type of responses given, as well as with coding rules. The development of automated or computer-assisted coding procedures should reflect the rules to be used for accurate and effective coding.
- Quality control procedures are needed to monitor the quality of the coding process and to provide feedback both to coders and those responsible for the classification itself and the coding tools.
Maintenance
Maintaining a classification should be an ongoing activity of those responsible for the classification, to be combined with training and back-stopping for users of the classification or of the resulting statistics. The main methodological issues are:
- How to best organize maintenance activities.
- Criteria to identify "significant" effects of maintenance activities on comparability of data;
- Development of methods to correct and smooth statistics for the effects of maintenance.
Revision
The activities involved in revising a classification are essentially the same as those needed to develop it. However, additional methodological issues involved are:
- How to determine whether a revision is needed.
- How to determine whether new solutions are better than current practices.
- How to implement a revised classification in on-going statistical programmes, given the need for comparability with past statistics.
|
|
|
