04 / 20 How to read a correspondence table
How do you read a many to many row?
Published Updated
Statistical classification only
Statistical classification only. This guide does not determine any reader's code. Confirm a code with the national statistical institute or the business register that holds the record. This is general information, not advice on any reader's registration.
A many to many row group shows that the joint scope of two or more NACE Rev. 2 categories is identical to the joint scope of two or more NACE Rev. 2.1 categories, without a one by one correspondence between the individual categories. A many to many group must be treated as a whole. Picking one old class and one new class from the group, and treating them as corresponding, misreads the table, because the table states no such individual correspondence.
How is a many to many group structured?
A many to many group, in the structure of the table, lists several old classes on one side and several new classes on the other side, grouped together. The illustration below shows the structure of such a grouping, using placeholder forms only, without asserting a particular grouping. A real grouping in the published table lists the actual old codes and headings, the actual new codes and headings, and the type. The illustration shows that the grouping is a set to set relationship, not a set of individual pairings. It must not be quoted as a grouping for any particular classes. A verified grouping is found only by reading the rows for the classes concerned in the published table, Version 1.05 as cited in the manual, or the current version on ShowVoc.
| Group element | What the published grouping shows |
|---|---|
| Old side | Two or more NACE Rev. 2 classes, listed as a set |
| New side | Two or more NACE Rev. 2.1 classes, listed as a set |
| Relationship stated | Joint scope of the old set identical to joint scope of the new set |
| Relationship not stated | No individual pairing between one old class and one new class |
| Form illustration only | Old set, new set, one grouping, no pairing implied |
The illustration uses words, not codes, to avoid presenting a placeholder as a real code. The point of the illustration is the set to set form. The content of a real grouping, the codes and headings in each set, must be read in the published table. A reader who needs the grouping for a particular class must find the class in the table, read the whole group that contains it, and note every class on both sides of the group, before drawing any conclusion.
Why does a many to many relationship arise?
A many to many relationship arises where a group of categories is restructured, with partial transfers of content between categories, so that the total scope of the group stays the same while the internal division changes. The manual defines minor changes as including partial transfers of content between classes, and it defines major changes as including restructurings within groups. A restructuring that moves part of the content of several old classes into several new classes, in different proportions, produces a many to many relationship, because no old class maps cleanly to one new class, and no new class is cleanly formed from one old class. The joint scope is preserved, while the boundaries inside the group move.
The size of the many to many group, 225 classes, shows how common that pattern is in the update. It is larger than the merger and split groups combined, and it is the second largest group after the one to one group. A reader who expects most rows to be one to one, and the rest to be clean splits or mergers, will therefore misread a large part of the table. The expectation must be adjusted: many rows belong to groups where the scope was reorganised, and the group, not the individual row, is the unit of comparison. The notes for the classes in the group, read together, show how the internal boundaries moved, and the table shows that the group scope did not change.
How is a many to many group used for a comparison?
A many to many group is used for a comparison at group level. The total for the new set of classes, for a reference period under the new version, can be compared with the total for the old set of classes, for a reference period under the old version, provided that the population and the domain are the same apart from the classification, and that no other break affects the comparison. The table supports the group comparison, because it states that the joint scopes are identical. The table does not support a comparison of individual classes inside the group, because it states no individual correspondence. A comparison of one old class with one new class from the same group, without additional information, will compare different scopes, and the difference will be attributed wrongly to a change in the statistic, when it is in fact a change in the category boundary.
Where an individual class comparison is needed, additional information is required. The explanatory notes for the classes in the group must be compared, to see how the boundaries moved, and the underlying data must be available at a level of detail that allows the old and new scopes to be separated. Where that detail is not available, a backcast series or a double reported series must be used, where the same reference period is tabulated under both versions. The table is the starting point for those operations, by identifying the group that must be treated as a whole, not a substitute for the detailed data or the parallel tabulation.
What does a many to many row not show?
Read the whole group in the published table, compare the notes for the classes in the group together, and confirm any entry with the national statistical institute or the business register that holds the record.