Field entry, published 2026-11-03T00:00:00.000Z, updated 2026-11-03T00:00:00.000Z. Group: The four levels of ISCO-08

How do correspondence tables work in ISCO-08?

A correspondence table maps groups in one classification to groups in another classification, so that figures produced under one system can be compared with figures produced under the other. In the context of ISCO-08, a correspondence table might link ISCO-08 groups to groups in an earlier version of the classification, or to groups in a national system that a country uses for its own statistics. Each row of the table states a relationship: this group here corresponds, wholly or partly, to that group there. The table is a translation device, and like any translation, it is exact in some places and approximate in others.

Why translation is needed

Different classifications divide work differently, even when they describe the same labour market. One system may split a field that another system keeps whole, or one system may draw a boundary in a different place. Without a correspondence table, a time series that changes classification would break, because figures before the change and figures after the change would not be comparable. A country that reports in a national system would also struggle to compare its figures with international figures reported in ISCO-08. Correspondence tables do not erase those differences, but they make them visible and manageable, by showing where a clean match exists and where a match is only partial.

The three kinds of match

A match can be one to one, one to many, or many to one. A one to one match is the simplest case: a single group in one system corresponds exactly to a single group in the other, and figures can be carried across without adjustment. A one to many match means that one group in the first system splits into several groups in the second, so a figure from the first system cannot be allocated to the second without additional information about how the split falls. A many to one match is the reverse: several groups in the first system merge into one group in the second, so detail is lost in the move, but the merged total remains valid. Recognising which kind of match you are dealing with is the single most important skill in using a correspondence table, because the three kinds support very different conclusions.

How to read a correspondence row

Read a row from both directions, and note whether the relationship is symmetrical. A row may state that a group maps to another group in one direction, while the reverse direction involves a split. Always check the notes that accompany the table, since they often explain partial matches, boundary changes, and cases where a group has no counterpart at all. A group with no counterpart is not an error. It means that the other system does not recognise that division, or recognises it only as part of a larger group. Forcing such a group into a counterpart would invent a precision that the table does not support.

What correspondence tables cannot guarantee

A correspondence table cannot guarantee that two groups with a mapped relationship are identical in every respect, because classifications differ in definitions as well as in boundaries. Two groups may correspond closely yet differ in inclusions, exclusions, or task descriptions. A table also cannot create detail that was never collected. If data were collected only at a broad level in one system, no table can recover the finer breakdown that a second system would have shown. Finally, a table cannot settle a classification decision for a particular case. It compares groups, not people, and a case that sits near a boundary still needs to be judged against group definitions.

A cautious workflow for using a table

Begin by identifying the exact versions of both classifications involved, since a table built for one version pair will not apply cleanly to another. Next, check the match type for every group you intend to use, and set aside any one to many matches that would require data you do not have. Then, compare the definitions on both sides for the groups that matter most to your question, rather than trusting the table alone. Only then should you carry figures across, and when you present the result, state that it rests on a correspondence, with its approximations acknowledged. That caution preserves the value of comparison without pretending that translation is the same as identity.

Sources checked

1. ILO ISCO-08

Back to the The four levels of ISCO-08 group or browse all occupations.