What do you check before comparing Rev. 4 and Rev. 5 series?

Before comparing Rev. 4 and Rev. 5 series, a reader should check the version labels, the correspondence rows for every category in the comparison, and whether any category was split, merged, moved or retired, using the UNSD material for ISIC Rev. 5, UN Statistics Division. The reader should also check the reference periods, because a revision break and a genuine economic change can look alike in a chart. Only after those checks can a comparison be described as comparable, and any remaining break should be footnoted.

In what order should the checks be done?

The opening paragraph lists the checks, but not their order, and order matters when time is short. Begin with version labels, because a comparison without known versions is not yet a comparison at all. Confirm that one series is Rev. 4 and the other is Rev. 5, from documentation rather than from assumption, and record both labels where the comparison will be published. Next, check reference periods, because a break inside a period, or a period that straddles national adoption dates, changes what the comparison can claim. Only then turn to the correspondence rows, category by category, since that detailed work is wasted if the versions or periods were wrong to begin with.

How do you read a correspondence row properly?

A correspondence row is a small document with a large job: it states how a Rev. 4 category relates to one or more Rev. 5 categories, or the reverse. Reading it properly means attending to its shape, not just its endpoints. A one-to-one row supports direct comparison, subject to checking that the definition did not also change. A one-to-many row, where a Rev. 4 category splits, supports comparison only at the aggregate of the Rev. 5 destinations, or not at all, depending on the data available. A many-to-one row, where Rev. 4 categories merge, supports comparison at the merged level but not below it. A row that ends in retirement, or begins in creation, marks a point where direct comparison stops.

Two further habits complete the reading. First, read the row in both directions when the comparison runs both ways, because a mapping that looks clean from Rev. 4 to Rev. 5 can look messier in reverse. Second, read the explanatory notes behind the row, since the row alone does not say whether a definition changed alongside the mapping. A row plus its notes is evidence. A row alone is a hint.

What counts as a split, a merger, a move, or a retirement?

The four fates named in the opening paragraph need distinct handling, and blurring them is a frequent source of error. A split divides former scope, so the old total cannot be compared with any single new total. A merger joins former scope, so no single old total matches the new total. A move relocates scope without dividing or joining it, so totals may still reconcile once the relocation is followed, but only at the level where the move is visible. A retirement removes a heading, so the question becomes where its scope went, not whether the heading still exists.

Each fate also needs a different footnote. A split footnote should name the destinations. A merger footnote should name the sources. A move footnote should name both the old and new locations. A retirement footnote should state where the scope was absorbed or redistributed, or, if that cannot be established, say so plainly. Footnotes written at this level of specificity let a later reader audit the comparison. Footnotes that merely say a break occurred do not.

Why can a break and a real change look identical?

A chart cannot distinguish its causes. A line that jumps at the revision point may record a booming field, a collapsing one, or a field whose scope was redrawn. The opening paragraph warns about this resemblance because it defeats the most natural way of reading a chart, which is to treat movement as economics. The only way to separate the causes is to bring outside information to the chart: the correspondence rows, the reference periods, and the notes on scope change. Without that information, the honest description of a jump is that it coincided with a revision, not that it was caused by the economy.

This caution applies with special force to the most quotable figures: the largest jumps, the fastest growth rates, and the most dramatic reversals around the revision. Those are exactly the figures most likely to be quoted, and exactly the figures most likely to be classification artefacts. A writer who pauses over dramatic cross-revision movements, checks them against the correspondence material, and tones down any claim the material does not support, protects both the reader and the writer’s own credibility.

This guide gives general advice on comparing statistical series. It does not determine a code, a principal activity, or a registration outcome for any reader, and a specific record should be confirmed with the national statistical institute or business register that holds it.