How do you trace an old Section J code to Rev. 5?

Tracing an old Section J code to ISIC Rev. 5, UN Statistics Division, starts by identifying the Rev. 4 category in words, then following its row or rows in the Rev. 4 to Rev. 5 correspondence table to the new Section J or Section K categories. Where one old category splits into several new ones, the reader must use the definitions and any available detail to decide which new category fits, or aggregate to a comparable level. The trace is complete only when the Rev. 5 explanatory note for the target category has been read and matches the activity described.

Where does a trace begin if the old code is all you have?

A trace begins in words, not in numbers. Write down, in a plain sentence, what the old Rev. 4 category covered, using its own definition rather than a remembered shorthand. That sentence is the anchor for everything that follows, because correspondence tables map definitions to definitions, and a vague starting sentence produces a vague result no matter how carefully the later steps are followed. With the sentence in hand, open the Rev. 4 to Rev. 5 correspondence table and find the row or rows for the old category. Some rows will point to a single new home. Others will point to several, spread across new Section J and new Section K. The table alone does not finish the job in the second case, but it does something just as valuable: it narrows the search to a short list of candidate categories, each of which can then be tested against the activity described in the opening sentence.

What happens when one old category opens onto several new ones?

A split row is not a failure of the correspondence table. It is the table telling the reader, honestly, that the old category covered more than one kind of work in Rev. 5 terms, and that extra detail is now required to place the activity. At this point the reader turns to the Rev. 5 explanatory notes for each candidate category, reading the includes and excludes lists rather than relying on titles. A title can flatter two candidates at once. An excludes list is less polite: it says, plainly, that a particular kind of activity belongs elsewhere. The reader compares each candidate note with the activity in words, and with any detail available about how the work was actually done, and keeps only the candidates that survive that comparison. If two candidates still fit, the reader has learned something important, namely that the available detail is not enough for a fine placement, and the next section explains the honest response to that discovery.

When is aggregating upward the honest finish to a trace?

Aggregating upward means reporting at a broader level where the competing candidates meet, instead of forcing a fine category that the evidence cannot support. A reader might hope for a class level answer and settle, transparently, for a division level one. That is not a defeat. A broad figure that is clearly right serves a comparison better than a detailed figure that a checker cannot reproduce, and it keeps the trace auditable: a later reader can see that the finer detail was unavailable, rather than wondering why a confident fine code rests on thin evidence. The trace is complete, in this sense, when the level chosen matches the evidence available, when the Rev. 5 note for the chosen category has been read and matches the activity in words, and when the version label travels with the result. A trace that ends one level higher than hoped, with its reasoning recorded, is a finished trace.

What does a finished trace look like on paper?

Imagine an analyst preparing a table for readers who did not collect the data. The analyst writes the activity in plain words, follows its correspondence rows, reads the candidate notes, and records the chosen category with its version label visible. A colleague later asks why a neighbouring category was not chosen, and the analyst can point to the excludes list that decided the matter, or to the note explaining why the result was aggregated upward. The published table carries the same label, so a reader in another country can open the same UNSD structure and follow the same path. Four checks run through the whole exercise: version, level, reference period, and source. If any of those four is missing from the analyst’s notes, the trace is not yet ready to share, however tidy the final number looks.

Nothing in this guide settles a particular reader’s position: it deals with statistical classification only, it does not determine any reader’s code, principal activity, or registration, and confirmation for a specific record belongs with the national statistical institute or business register that holds the record.