Field entry, published 2026-11-03T00:00:00.000Z, updated 2026-11-03T00:00:00.000Z. Group: From ISCO-88 to ISCO-08, and revision

How should a reader handle a dataset that still uses ISCO-88?

A reader who meets a dataset that still uses ISCO-88 should treat it as a valid record of its own time, read it in its own edition, and convert it to ISCO-08 only through published correspondence material, never by matching titles by eye. Older datasets survive for good reasons: a census may predate adoption, a long running survey may have kept its coding for continuity, or an archive may preserve original coding for research. None of those reasons makes the dataset wrong, but all of them mean that its figures cannot be dropped into an ISCO-08 table without a documented bridge. Handling the dataset well is mostly a matter of labelling, checking, and restraint.

What is the first check to make on arrival?

The first check is the methods note. Look for the edition named, the year or survey round, the level of detail published, and any statement about correspondence or dual coding. A good methods note will say plainly that figures use the older edition, and will point to the correspondence table used if any conversion was applied. If no methods note exists, treat the dataset as unconverted older edition material until evidence shows otherwise, and say so in any reuse. That single habit prevents the most damaging error, which is to present older edition figures as if they were already in the newer edition. A label costs one sentence, while a silent mix can mislead every reader downstream.

How does published correspondence material actually help?

Correspondence material maps older groups to newer groups, recording where one older group splits into several newer groups, where several older groups merge into one newer group, and where a boundary moves only partly. For a straightforward one to one case, conversion is simple and can be stated with confidence. For a split, conversion needs extra information, such as a sample coded in both editions, or a reasoned allocation that is disclosed as an estimate. For a merge, conversion is easier in one direction than the other, because detail lost in merging cannot be recovered without microdata. A reader should therefore ask, for each group that matters to the question at hand, whether the correspondence is exact, partial, or impossible. Groups that are exact can be compared boldly, groups that are partial need a caveat, and groups that are impossible should be described separately rather than forced together.

What should a reader do when no clean match exists?

When no clean match exists, the honest options are to report the older and newer figures side by side without joining them, to aggregate to a broader level where the match becomes sound, or to restrict the comparison to the overlapping portion with the restriction stated. Each option loses something: side by side reporting loses a single trend line, aggregation loses detail, and restriction loses coverage. That loss is preferable to a fabricated trend, because a fabricated trend looks precise while resting on invented allocation. If a chart is unavoidable, mark the edition change with a visible break and a note, so that a reader sees the join rather than inferring smooth change. A dataset that still uses ISCO-88 is not a problem to be hidden. It is a historical record that deserves its own label.

Which groups need the most caution in an older dataset?

Groups that changed most in the revision need the most caution, and service, professional, technical, and technology related areas head that list. Manager and professional boundaries also shifted, so counts in those areas should not be compared across editions without correspondence checks. At the other end, some broad fields changed less, and broad major group comparisons may be workable with a note. Even there, caution remains wise, because a broad total can hide offsetting splits inside it. A practical rule is to check detail before trusting a total: if the subgroups behind a total moved, the total itself may still be comparable, but only a check of the subgroups can show that. Document the check, even briefly, so that later readers can follow it.

How should findings from an older dataset be written up?

Write up findings with the edition in the first mention of the data, not buried in a footnote. For example, state that a figure comes from an older edition dataset, name the years covered, and state whether any conversion was applied and by what method. If conversion was estimated, say that the converted figures are estimates, and give the basis for the allocation. If no conversion was applied, say that comparisons with newer edition figures are limited to the groups where correspondence is exact. That discipline turns an older dataset from a trap into a resource. Historical depth is valuable precisely because it reaches back before the revision, and careful labelling is what lets that depth be used without pretending the past was coded with present tools.

An older dataset handled this way keeps its value intact: it tells us about its own period, in its own terms, and it joins the newer record only where the bridge is sound.

Sources checked

1. ILO ISCO-08

Back to the From ISCO-88 to ISCO-08, and revision group or browse all occupations.