How do correspondence tables help compare statistics over time?

Correspondence tables help compare statistics over time by showing, for each class, how its scope relates to the scope of classes in the other version, so that a series can be converted, backcast, or tabulated in parallel with the relationship stated. The manual states that any update with structural changes leads to breaks in time series, and that the correspondence tables are important tools for comparing data collected under different classifications. The table is the framework for managing a break. It is not, on its own, a converted series.

What is backcasting, and what is double reporting?

Backcasting, in the sense used on the Eurostat overview page, is the preparation of earlier series on the new basis, so that a longer run can be read consistently under NACE Rev. 2.1. The overview page lists establishing plans for backcasting as one of the roll out steps, alongside publishing guidelines and notes, updating sectoral legislation, and establishing plans for double reporting. Double reporting is defined on the same page as tabulating data by NACE Rev. 2 and NACE Rev. 2.1 in parallel. The two tools serve different needs. Backcasting gives a single series on the new basis, extended backwards. Double reporting gives two series for the same reference period, one on each basis, so that the effect of the change can be seen directly.

The correspondence table supports both tools, because it states the relationship type for each class. For a one to one class, backcasting is direct, because the scope is identical. For a merged class, backcasting requires adding the old classes, because the joint scope is the new scope. For a split class, backcasting requires dividing the old total, which cannot be done from the table alone, and requires underlying data at a finer level, or a parallel tabulation that shows the division. For a many to many group, backcasting must treat the group as a whole, because the joint scope is identical while the individual scopes differ. The table therefore decides which operation is possible for each class, and which operation requires additional data.

Operation	What the correspondence table provides	What else is needed
Direct comparison, one to one	Identical scope stated	Same population and domain, apart from the version
Addition, many to one	Joint scope stated	Old totals for all classes in the merger
Group comparison, many to many	Joint scope identity for the group stated	Totals for the whole group, in both versions
Division, one to many	Split stated, new classes listed	Finer underlying data, or a parallel tabulation, to divide the old total

How is a break in a series recognised?

A break in a series is recognised by checking the version label, the correspondence type, and the metadata for the statistic. The version label states which classification was used for a reference period. Eurostat states that statistics for the period from 2008 to 2024 (FIXED range) use NACE Rev. 2, and that the new version is to be used for European statistics from 2025 (FIXED) onwards. A series that crosses that boundary, without a backcast or a double reported period, will show a break at the boundary, where the classification changed. The correspondence type for the classes in the series states whether the break is a scope break, a division break, or no break at class level. The metadata for the statistic states whether any other break, such as a change in population or method, coincides with the classification break.

The manual states the general principle: any update, particularly one with structural changes, leads to breaks in the time series of the statistics depending on it. That principle means that a break must be assumed where the classification changed, until the correspondence table and the metadata show otherwise. A reader who compares figures across the boundary without checking the table will attribute a classification break to a change in the economy, or will miss a real economic change that is hidden by a classification break. The table and the metadata together separate the two effects, by stating the classification relationship and the other changes that affect the series.

What must a reader not do with a table when converting a series?

A reader must also not treat a double reported period as if it were two independent observations of the economy. Double reporting tabulates the same underlying data by two classifications, in parallel. The two figures for the same period describe the same underlying activity, grouped differently. The difference between them is the classification effect, not an economic change. A reader who adds the two figures together, or who treats the difference as growth, will misread the parallel tabulation. The parallel figures must be read as two views of the same period, and the classification effect must be stated as the difference between the views.

What does a table not provide for a comparison?

A table does not provide the converted figures, the backcast series, or the double reported figures themselves. It provides the framework for producing them. The figures must be produced by the statistical office that compiles the statistic, using the underlying data and the table, and published with metadata stating the version, the method, and any break. A reader who needs a converted series must therefore look for the published backcast or double reported series for the domain concerned, in the Eurostat or national institute publication for that domain, rather than constructing a series from the table alone. The table shows what conversion is possible. The published series shows the conversion that was made.

Check the overview page for the backcasting and double reporting plans, check the table type for the class, and confirm any entry with the national statistical institute or the business register that holds the record.