|
QUESTIONS AND ANSWERS ABOUT ISCO-08
This document contains questions and answers on the new International Standard Classification of Occupations 2008 [ISCO-08] and its relationship to previous versions of ISCO and derived occupational status measures such as ISEI, SIOPS, EGP and ESEC.
Last revision: April 29 2013 [HG]
What is ISCO-08 and what can you do with it? What is on the ILO website? How do ISCO-88 and ISCO-08 relate? How does the hierarchical digit system work and how many digits do we need? How is ISCO related to skill level? Can ISCO-08 accommodate crude answers to occupation questions, as often arise in survey responses? How has the classification of managers changed from ISCO-88 and ISCO-08? Has the classification of farmers changed between ISCO-88 and ISCO-88? How do I handle ambiguous and multiple descriptions? Do I need / can I use additional information (on industry or status in employment) to code data in ISCO-08? How can I upgrade ISCO-88 data to ISCO-08 data? How can I compare two datasets with different occupation codes? Is it useful to use ISCO-88 as a bridge to obtain ISCO-08 codes? Can I use automated systems to code ISCO-88 to obtain ISCO-08 codes? How can I upgrade ISCO-68 data to ISCO-08 data? Suppose I would have these ISCO-08 codes, what can I do with them?
What is ISCO-08 and what can you do with it?
ISCO-08 stands for the International Standard Classification of Occupations 2008 and is published and maintained by the International Labour Organization on behalf of the International Conference of Labour Statisticians, who decided on the final form of ISCO-08 in its resolution of December 6 2007. ISCO-08 in its official form is documented on the ILO/ISCO website (http://www.ilo.org/public/english/bureau/stat/isco/index.htm), that also documents the previous versions: ISCO-58, ISCO-68 and ISCO-88.
ISCO is an international classification aimed for producing comparative official statistics and social research. Many countries maintain national occupational classifications that are often more detailed and may have quite different logics of classifications than ISCO. Such national classifications need to be converted to ISCO for comparative purposes (which will imply some loss of distortion of the information). Increasingly in the past, countries have chosen to adopt ISCO or a somewhat refined version as its national classification. International social research projects, such as the ISSP, ESS, EVS, PISA and IALS have chosen ISCO as their preferred occupational classification.
Note that ISCO is used to classify occupations and does not constitute a sociologically meaningful scale as such. In statistical analysis, the ISCO classification is used as an intermediary step to derive various measures of occupational status, such as ISEI, SIOPS or EGP/ESEC.
While ISCO-08 was adopted by ILO in December 2007, it has hardly been used as a classification tool by projects conducted in 2008 or 2009. However, in the near future, international research projects such as PISA and PIAAC, and most likely the 2010 round of national population censuses will use ISCO-08 as their occupational classification tool.
What is on the ILO website?
The ILO website shows the official version of ISCO-08 using one-line titles of the 570, occupations, 120 minor groups, 34 sub-major groups and 10 major groups. There is also: · A 583-page document (dated July 9 2009) with draft definitions of all the groups. This document also lists examples of occupations to be coded or not to be coded into a certain category. · Two correspondence tables between ISCO-88 and ISCO-08, one showing how ISCO-88 maps into ISCO-08, and another how ISCO-08 can be mapped back into ISCO-88. Note that these documents imply a many to many mapping of the two classifications. · Finally, there is a set of discussion documents that comment on definitions by industry or main tasks. These documents were/are used to develop the official definitions.
Note that there appears to be no such thing as an official ISCO-08 manual as of yet. There are many details in the documents at the ILO site that make clear that the final manual is still under construction and that some details have not been settled yet.
The ILO/ISCO website also contains documentation on the three previous versions of ISCO. While there is a pdf version of the complete manuals for 1958 and 1968, ISCO-88 is only documented in an abbreviated form using the definition parts of the manual.
Note that the following document lists a large number of occupations (English language) with their ISCO-68 and ISCO-88 codes: http://www.ilo.org/public/english/bureau/stat/isco/isco88/alpha.htm.
How do ISCO-88 and ISCO-08 relate?
The official aim of ISCO-08 (stated in the preamble to the ILO-resolution) was to produce ISCO-08 as a “minor revision” of the previous version. One reading of that statement could be that the revision basically amounts to how unit and minor groups are organized, as most of the changes indeed occur at these 4- and 3-digit levels. At the major and sub-major level, fewer modifications have been made and they are mostly in the form of further refinement. However, some of the changes at the more detailed level involve that minor groups are shifted between sub-major groups and occasionally between major groups: despite nominal equivalence at the major and sub-major level, this implies changes in underlying contents.
But however minor or major the changes have been, adopting the new classification involves that researchers familiarize themselves with the new classification. It is most important to understand the hierarchical digit system. For any user, it is best to learn the major and sub-major groups system by heart.
For users who are familiar with ISCO-88 it may be helpful to have a list of the most important changes at the 1-digit (major) and 2-digit (sub-major) level.
Major groups (1-digit level): · There are small changes in the names of major groups. · 0000 (Military) is now formally added as a major group. · In the underlying contents the most dramatic change involves that some occupations that were formerly classified in 8000, are now classified in 3000. These involve Manual Supervisors and part of Stationary Plant Operators. · Another major change is that two large groups that used to be part of major group 1000, are now classified elsewhere: Shopkeepers go into 5200 and Farmers into 6100.
Minor groups (2-digit level): · There are now 34 sub-major groups instead of 28. The major changes are:
· In other words: the changes in sub-major groups is certainly not minor. For this reason recoding 2-digits ISCO-88 into 2-digit ISCO-08, without paying attention to the underlying 3- and 4-digit contents is not a good idea.
Two other general observations are important to keep in mind: · Despite its stated raison-d’etre that ISCO-08 adapts to new divisions of labor as they arise in modernizing societies around the world, many changes relative to ISCO-88 are better interpreted as attempts to repair faults of the previous classification and reintroduce elements that were abolished in the transition from ISCO-68 to ISCO-88. Instances are: (A) reintroduction of manual supervisors, (B) separation of shopkeepers from other managers and working proprietors, (C) reintroduction of occupations such as bicycle repairman, kitchen helper, service station attendant and payroll clerk. · Over-all ISCO-08 appears to be a bit more detailed than its predecessor.
How does the hierarchical digit system work and how many digits do we need?
ISCO-08 is conveniently organized in a system of four hierarchical digits that can be displayed as follows:
[1000] Major group 1 [1100] Sub-major group 11 [1200] Sub-major group 12 [1210] Minor groups 121 [1211] Unit group 1211 [1212] Unit group 1212 etc.
(Note that we adopt the convention to list all groups using trailing zeroes. The official ISCO documentation does not. The particular convenience of this is that string sorting and numerical sorting amount to the same thing.)
The hierarchical digit system implies that one can move up levels by simply truncation of trailing digits (and replacing them by 0). This system was first implemented in developing ISCO-88 and was not in place in ISCO-68 and ISCO-58, which creates much trouble in manipulating these codes in computer programs.
Note that the introduction of a major group 0000 may create confusion with missing values codes. Missings are preferably codes as 9999 or negative (-1).
The hierarchical system is a deliberate attempt to group similar occupations in a single branch of the classification in the same way as a hierarchical cluster analysis would do. How similar different members of a branch are, depends of course much on the purpose of an analysis. One approach to answering this question is to adopt a certain scale for all occupations and analyze how much variance is explained at each level. When we take ISEI as our criterion, the following numbers are found for the database of ISSP 2002-2007:
Unit groups 100% Minor groups 95.5% Sub-major groups 89.1% Major groups 82.8%
This result would suggest that any association with occupation is attenuated with the square root of the numbers listed. It needs to be emphasized that these numbers could be smaller or larger when another criterion (such as earnings) would be used.
Still using the ISEI scoring, we can more qualitatively list instances where most heterogeneity would be incurred if only 2 digits were used: o 1100: Legislators and Corporate Managers o 2200: Doctors and Nurses o 2300: University Professors and (Pre-)primary Teachers o 2600: Lawyers, Librarians and Sociologists o 3100: Engineering Technicians and Manufacturing Supervisors, Aircraft Pilots o 3300: Government Officials and Sales Representatives o 3400: Chefs (Cooks) and Professional Sportsmen o 5100: Cooks, Waiters and Hairdressers o 5200: Shop Salesmen and Market Vendors o 7300: Handicraft Workers and Printers o 7500: Food Production Workers and Textiles Workers o 8300: Train drivers, Truck drivers, Sailors o 9100: Office cleaners and Window Washers o 0000: Army Officers and Other Army Ranks
Idem when we use only 3 digits: o 1110: Legislators and Union Officials o 1210: Finance Managers and Human Resource Managers o 1220: Sales Managers and R&D Managers o 1340: Child Care Services Managers and Financial Services Managers o 2160: Building Architects and Graphic Designers o 2210: General Medical Practitioners and Specialist Medical Practitioners o 2220: Nurses and Midwives o 2260: Dentists and Dieticians o 2340: Primary and Preprimary Teachers o 2510: System Analysts and Web Designers o 2610: Judges and Other Legal Professionals o 2630: Economists and Religious Professionals o 2650: Dancers and Sculptors o 3150: Ship Captains and Aircraft Pilots o 3250: Dental Assistants and Ambulance Workers o 3320: Sales Representatives and Buyers o 3330: Conference Planners and Real Estate Agents o 3430: Photographers and Chefs o 4210: Bank Teller and Bookmaker o 4220: Hotel Receptionists and Survey Research Interviewers o 5110: Train Conductors and Travel Guides o 5160: Astrologers, Undertakers and Driving Instructors o 5220: Shop Keepers and Shop Sales Assistants o 5240: Fashion Models and Door-to-door Salesmen o 7110: Bricklayers and Carpenters o 7120: Roofers and Plumbers o 7210: Jewellery Workers and Basket Weavers o 7510: Bakers and Butchers o 7530: Tailors and Shoemakers o 8110: Train Drivers and Taxi Driver o 9333: Freight Handler and Shelf Filler
At the same time it is true that many of the finer distinctions made will be of no practical importance to whatever analysis is done with the occupations.
Many researchers will be tempted to settle for the use of only 2 of 3 digit codes instead of the full version. In doing so they are indeed likely to capture most of the variations in social outcomes brought about by occupations. At the same time, they will bereave (XX) themselves and other users of the data of some extra variance explained, opportunities to compare different ways to scale occupation data and flexibility in uses of the data.
One other element should be brought to the considerations. The novel user of detailed occupation codes may have the impression that coding only 2 digits is a relatively easy task, whereas adding the last 1 of 2 digits is an enormous burden upon the coding process. This is not so, or should not be so. Whereas the coding of major and sub-major groups can (and should) be done initially without consulting the documentation extensively, in the end it is detailed occupations that need to be coded and the choices should be justified using all the available documentation. If one operates coding as a two-step process, in which in a first round the data are classified in a crude way, and in a second round refinement and revision are added, coding the third and fourth digits is not a tremendous amount of work and will in fact add much to the proper coding of the first and second digit.
How is ISCO related to skill level?
Like its predecessor ISCO-08’s primary logic of classification is the skill levels of occupations. However, this point of view is not consistently maintained, and this is a good thing. Major groups 2000, 3000, 4000, 7000, 8000 and 9000 are in clear order by skill levels, but the skill levels of major groups 1000, 5000 and 6000 is ambiguous. There is no connection between skill levels and the ordering of sub-major groups within major groups.
How has the classification of managers changed from ISCO-88 and ISCO-08?
ISCO-88 presented a rather clear definition of how to classify managers and seemed to resolve long-standing issues with this in ISCO-68. First, ISCO-88 distinguished organizations in large and small by using the idea that large organizations have departments and small organizations have not. This distinction was operationalized in a footnote to the classification as small organization having only 1 or 2 managers, while large organizations have at least 2 department managers and at least one ‘corporate’ manager. The managers of small organizations were called ‘general manager’, a rather confusing term. The department managers in large organization were divided into two kinds: those who oversaw the production or operation of the organization itself (mending the ‘core business’) and those in charge of ‘support’ departments, such as finance, sales, research and development, etc. The operations department managers were then subdivided according to industry, which made that the coder was faced with the choices like between “Operations Department Manager in Transport” and “Transport Department Manager”. While these distinctions are actually conceptually clear, this system was in practice hard to implement as no data source contains the number of managers to begin with, and needed an industry classification to be coded properly. Another problem was that the management categories so defined were rather heterogeneous in occupational status.
ISCO-08 has reordered the information in a dramatically different way: · Corporate managers (‘CEO’) are no longer classified with department managers, but with leading government officials. · Managers are now distinguished primarily by level of activity and industry: [1200] Administrative and commercial managers [1300] Production and specialized services managers [1400] Hospitality, retail and other services managers. · Shop Keepers (and Shop Supervisors) who used to be included among managers, are now moved to sales workers (5200). · Farm Managers (& Proprietors) have now been relegated to Agricultural Workers (6000), unless they run very large agricultural enterprises. · Additional categories have been created elsewhere for some occupations that used to be classified as Managers: [0110] and [0210] contain Army Officers and Non-Commissioned Officers, [3120] contains three kinds of Mining, Manufacturing and Construction Supervisors, [3341] contains Office Supervisors. Notice that these additions all arise at the 3- of 4-digit level.
It remains to be seen whether these changes amount to an improvement in practice.
Has the classification of farmers changed between ISCO-88 and ISCO-88?
In ISCO-88 farmers could be coded in five different places in the classification:
[1211] Production Department Manager in Agriculture [1311] General Managers in Agriculture [6100] Skilled Farm Workers [6200] Subsistence Farmers [9200] [Unskilled] Agricultural Labourer
In ISCO-88 it was particularly unclear whether self-employed farmers, running a small farm were to be coded as [1311] or [6100]. Different coders, different researchers and different countries made different decisions on this. Unfortunately, this can be a very consequential choice, as farmers can be an extremely numerous group, in particular among parents and in developing economies, and are very different in socio-economic status (and social mobility) from other ‘General Managers’. Note also that [1311] is a 4-digit code and disappears as a unit when one uses only two or three digits. For this reason it was recommended to avoid [1311] altogether. In ISCO-08 [1211] has been maintained as [1310], but unfortunately the reference to ‘department’ has been removed. However, the ILO draft definitions explicitly refer to “large scale agricultural, horticultural and forestry operations such as plantations, large ranches, collective farms and agricultural co-operatives”, and so this code would be inappropriate for the average farmer.
This leaves only [6100] and its minor and unit groups as an option to code farmers.
Can ISCO-08 accommodate crude answers to occupation questions, as often arise in survey responses?
Crude responses often occur in surveys and occupational classifications should have facilities to accommodate them (crude responses are not necessarily bad of uninformative responses). Unfortunately, ISCO-08 is not very good at this. However, there are some possibilities. These arise because it is not necessary to code every response to the same levels of detail and a higher-level code may be quite appropriate and preserve the information. Here are some examples:
Managers [1000] or [1300] Shop Owner [5221] Independent [1400] Entrepreneur, Businessman [1120] Foreman [3120] Skilled Worker [7000] Semi-Skilled Worker [8000] Unskilled Worker [9000]
How do I handle ambiguous and multiple descriptions?
Very often respondents give information that can be interpreted in multiple ways, because a job has multiple components, or someone has multiple jobs. The Introduction to the ISCO-88 manual provides a set of rules that can be adapted to this situation. Code the information using the following rules in sequence:
· Numerical dominance rule: when one activity dominates, or one interpretation is the far more plausible one given the distribution of activities in the population, code accordingly, · Skill level rule: if a set of activities involves a mix of skill levels, choose the most skilled one. · Production rule: if a set of activities involves production next to sales and/or management, choose the production occupation.
Do I need / can I use additional information (on industry or status in employment) to code data in ISCO-08?
Strictly speaking coding ISCO-08 requires only a description of the tasks and responsibilities to code the occupation, next to the title of the occupation. However, other information may be available and can be useful to determine the most appropriate code. This information may include: · Industry. · Self-employment · Supervisory status.
It is suggested that you provide your coders with this information to expedite their work. · Industry is not only useful because ISCO often refers to industrial distinctions (‘health’, ‘government’) but also because respondents often mix up occupation and industry. · Self-employment is useful to delineate shop-keepers from shop-supervisors. · Supervisory status is useful to categorize responses into one or the other managerial or supervisory codes. However, it is suggested that you do not apply this information mechanically, but judge its application on a case-by-case basis.
It is also recommended that you do not refer to education, earnings, gender or age while coding, as the relationships between these variables and occupational status are often subject to research and they should not be mixed up during coding.
How can I upgrade ISCO-88 data to ISCO-08 data?
It is not possible to upgrade ISCO-88 data to ISCO-08 data without loss or distortion of information. This would only be possible if ISCO-08 at some level would be a many-to-one mapping of ISCO-88, but it is not. Given the greater detail of ISCO-08, the reverse strategy (downgrading ISCO-08 into ISCO-88) might be the more commendable strategy for researchers who need to harmonize data that are coded in the two classifications, but that remains to be seen.
If one is willing to accept the potential loss and distortions, a straight one-to-one recode is available at isco8808.sps. This (spss) recode chooses the ‘best’ possible alternative among options. It also indicates whether there were more options to choose from and how many. The one-to-one recodes were established using the many-to-many conversions that ILO provides on its website.
It is sometimes possible to improve the conversion by taking other data into account. In particular it is sometimes helpful to have additional data on industry, or status in employment to choose a more appropriate ISCO-08 category.
The spss conversion tool will work at all levels of detail of the originating ISCO-88 distribution (if coded using trailing zeroes), but note that the quality of conversion remains detail-sensitive. More detail helps to make it better. Converting ISCO-88 into ISCO-08 using only 2-digit codes on both sides is a different thing as converting at a 4-digit level and then aggregating to two digits!!
How can I compare two datasets with different occupation codes?
A situation that will frequently arise in longitudinal research is that one dataset has ISCO-88 codes and a new one ISCO-08 codes. Comparing these is still possible, but are several ways to go. Three reasonable strategies are: · Convert ISCO-08 into ISCO-88 using automatic conversion and work with ISCO-88. · Convert ISCO-88 into ISCO-08 using automatic conversion and work with ISCO-08. · Construct derived variables (e.g. ISEI) from each measure and work with that. I expect the differences between these alternatives to be minor and would prefer the middle strategy (upgrade to ISCO-08) for practical reasons.
There is also is another way NOT to do this. This would be to aggregate both data-sets to its largest common denominator, the ten major groups, thatare nominally identical. Below this nominal equality there is in fact variable contents. Such a strategy would in particukar miss that all farmers are now moved to farming, while shop-ownerd and managers are now included with sales workers. These decision have fundamentally changed the composition of major group 1000 [Managers].
Is it useful to use ISCO-88 as a bridge to obtain ISCO-08 codes?
If you have new data to code, nothing is gained by first coding into ISCO-88 and then upgrade to ISCO-08. If your aim would be to obtain a double coded dataset, it would be more useful to work the other way around: first code ISCO-08 and then downgrade to ISCO-88.
A different situation arises when you have already coded ISCO-88 but have also additional data (e.g. the original strings or a national classification) available. In these cases it is useful to use the one-to-one recode as a starter and then use the number of options suggested by the recoding module and the many-to-many correspondence tables provided by ILO to decide what the best fitting option is.
A similar situation arises when you use an existing database coded in ISCO-88 to ‘automatically’ code occupations.
Can I use automated systems to code ISCO-88 to obtain ISCO-08 codes?
Increasingly, automated coding systems or support systems are used to code open-answer occupation data. This can happen in-field (respondent / interviewer choose from a database) or in post-processing. It will take a little while before such systems are adapted to the new standard. Meanwhile, in such cases there is an obvious gain in efficiency to first code in ISCO-88 and then convert manually to ISCO-08, i.e. by using the conversion tools and then review the result. Of course, if you do not allow for review (i.e. do not collect or keep the original strings), you are just taking an undesireable shortcut and not adding anything by using ISCO-08.
Can I upgrade ISCO-68 data to ISCO-08 data?
As with ISCO-88, some loss and misclassification will occur when you recode ISCO-68 into ISCO-08. However, there is also a great degree of similarity and constancy, in particular at the most detailed (3-digit) level. Having such a conversion is needed to convert Treiman’s SIOPS scale to a SIOPS-08 scale. There is no support here from the side of ILO: no conversion tables have been published to decide about splits. However, at points in the ISCO-08 classification there is more similarity with the 1968 classification than with the immediate predecessor.
Note that converting ISCO-68 data into ISCO-88 and then to ISCO-88 will not give the same result as the direct recode.
Suppose I would have these ISCO-08 codes, what can I do with them?
At the moment: very little. Most social research involving ISCO codes works with derived occupational status scales, such as ISEI, SIOPS or EGP/ESEC. These scales are not yet available in final format for ISCO-08. The construction and validation of such occupational status scales awaits the availability of large datasets that are coded in detailed ISCO-08.
To fill the gap, at least for the time being, Ganzeboom has created to provisional occupational status scales, one for ISEI and one for ISEC (which incorporates EGP and ESEC). Recodes are available on: http://home.fsw.vu.nl/hbg.ganzeboom/isco08.
The construction of these two provisional scales is discussed in a companion document. |