Field entry, published 2026-11-03, updated 2026-11-03. Group: ISCO-08 in statistics and surveys

How do time series handle a classification change?

A time series handles a classification change by making the break visible, by explaining what changed, and, where possible, by providing a bridge that lets readers compare old and new figures without pretending the break never happened. When a statistical office moves from an earlier occupation classification to ISCO-08, or revises ISCO-08 itself, some jobs move between groups, some groups split, some merge, and some gain or lose occupations at their edges. Continuing old numbers under new labels would mislead. Good practice therefore treats the change as an event in the series, not as a footnote to be ignored.

What actually breaks when a classification changes?

Three kinds of break can occur, and they call for different handling. A split occurs when one old group becomes two or more new groups, so the old total can be divided only if additional information shows how its jobs distribute across the new groups. A merger occurs when several old groups become one new group, so the new total can be built by adding the old totals, provided the merger is clean and no jobs enter or leave at the edges. A boundary shift occurs when jobs move between groups that both continue to exist, so neither old total matches its new counterpart, and the size of the shift must be estimated from recoded material. Most real revisions combine all three, which is why a single conversion factor rarely tells the whole story.

The severity of a break also depends on level, so a user should check the break at the level being used.

Which bridging methods do offices use?

Method	How it works in general terms	When it helps most
Dual coding	A sample of jobs, or a full collection where feasible, is coded under both the old and the new classification for an overlap period	When original descriptions are still available and recoding is practical
Backcoding	Earlier descriptions are recoded to the new classification, producing a revised historical series	When earlier material is detailed enough to support new coding
Correspondence mapping	Old groups are mapped to new groups using the structure and descriptions of both classifications	When recoding is not possible and a transparent approximation is acceptable
Break marking	The series is published with a visible break, with old and new segments shown separately and no forced link	When mapping would be too uncertain to support a continuous series

Dual coding is often the strongest approach, because it shows how old and new groupings differ for the same jobs. Backcoding extends that logic backward, but its quality depends on how much detail the old descriptions contain, and vague old answers may only support broad level backcoding. Correspondence mapping is useful as documentation, since it shows the intended relationship between classifications, but a mapping table alone does not prove how many jobs actually move, because real jobs may not distribute as the mapping suggests. Break marking, far from being a failure, is sometimes the most honest option, since a forced bridge built on weak evidence can mislead more than a visible gap.

How should a reader use a bridged series?

A reader should first look for the break marker and the methods note, and should treat any figure that crosses the break with extra caution. If the office provides both an unrevised series and a bridged series, the reader should understand which one is being quoted, and should not mix segments from different versions in a single trend calculation. If the office provides conversion proportions, the reader should check whether they were based on dual coding or on mapping alone, and at what level they apply, since a broad level conversion may not hold at detailed level.

It also helps to compare the bridged trend with related series that did not change classification in the same way. A sudden jump exactly at the revision point, with no movement before or after, should prompt suspicion that the break, rather than the labour market, explains the jump. A gradual movement that continues across the break is more likely to reflect real change, although it still needs methods checking.

What does good documentation look like?

Good documentation states the old and new classifications, the date of change, the groups most affected, the bridging method used, the level at which bridging is considered reliable, and the limits of the bridged series. It also preserves the original series, so that later researchers can see what was published before bridging, and it explains any decision to mark a break rather than bridge it. That transparency lets users choose between a bridged broad trend and separate segments.

Classification change is unavoidable, because work changes and classifications must change with it, and a time series that never changed classification would eventually describe a world that no longer exists. The aim is not to eliminate breaks, but to manage them openly, so that readers can tell how much of a movement reflects work, how much reflects grouping, and how much remains uncertain. A series handled in that way keeps its value across a revision. A series that hides its break loses trust, even if its individual figures were coded carefully.

Sources checked

1. ILO: ISCO-08

Back to the ISCO-08 in statistics and surveys group or browse all occupations.