What is a correspondence table?

A correspondence table for ISIC Rev. 5, UN Statistics Division, maps categories from one classification to categories in another, most often from Rev. 4 to Rev. 5 or from ISIC to a national classification. Each row links a source category to one or more target categories, showing how the structure changed. Correspondence tables are the bridge that lets long time series and cross-country comparisons be built without pretending the structures never changed.

What does one row of the table actually show?

One row shows a structural relationship, nothing more and nothing less. A row with one target usually signals a rename or a category carried over with little change, though even then the notes on both sides should be checked, since a similar title can hide a changed scope. A row with several targets signals a split, and a row whose target also appears as the target of other source rows signals a merger or a partial move. The row does not show how much activity moved, in which year, or in which country: those are facts about data and adoption, not about structure, and they live elsewhere. Reading a row well therefore means reading it as a statement about categories, and then going to the notes, the reference periods, and the adoption history for everything the row does not say. A reader who treats the row as a complete answer will overclaim; a reader who treats it as a starting map will use it correctly.

Why do revisions need a bridge at all?

Without a correspondence table, every revision would strand the past: old series could not be related to new ones, and a reader could not tell whether a change in a figure reflected a change in the economy or a change in the categories. With the table, a statistical office can plan the transition, publishing back series or breaks in series where the structure allows, and a reader can see exactly which rows survived, which split, and which merged. The bridge also serves cross-country work, where a national classification is mapped to ISIC so that countries using different national structures can still be compared at the ISIC level. In both uses, the table’s value lies in making change visible rather than hiding it. A revision handled without a published correspondence leaves readers to guess at the mapping, and guesses about structure produce the same errors as guesses about codes, only spread across whole time series.

How is a correspondence table used in practice?

An analyst building a long series starts by laying the table beside the old and new figures and marking each row by type: one to one, one to many, many to one, or partial. One to one rows can usually be carried through, subject to a check of the notes. One to many rows force a choice between a broader aggregate, finer data, or a break in series. Many to one rows can usually be summed, since several old baskets now form one new basket, again subject to a notes check. Partial rows need the full network of rows that feed each target before any conclusion is drawn. Throughout, the analyst keeps the version labels visible: ISIC Rev. 5, UN Statistics Division, on the new side, and the earlier version named on the old side, so no reader mistakes a mapped figure for an originally coded one. A mapped figure is a constructed comparison, and labelling it as such is part of the honesty of the method, not an admission of weakness.

A long series, bridged row by row

Picture an analyst who must extend a sector series across the move from Rev. 4 to Rev. 5. She opens the correspondence table and works through its rows for her sector, marking the one to one rows that carry straight through and the split rows that do not. For the split rows she checks whether a broader aggregate covers all the targets, and where it does she rebuilds that part of the series at the broader level, footnoting the change. Where it does not, she leaves a marked break and presents the two segments without joining them. The finished series is a patchwork of carried rows, broader aggregates, and one visible break, and it is far more trustworthy than a smooth line would have been, because every segment means what its label says and every break is visible. That patchwork is what a correspondence table is for: not a magic conversion of old numbers into new ones, but a visible map of what survived, what changed, and where continuity ends.

A table read this way does not pretend the structures never changed. It shows the change, row by row, and lets the reader judge what can still be compared.

This guide gives general information about statistical classification only, and it does not settle any reader’s individual matter. It determines no code, no principal activity, and no registration for any reader, and the code belonging to a particular record should be confirmed with the national statistical institute or business register that holds the record.