What does a correspondence table not decide?

A correspondence table does not decide a code, an entry, an eligibility, or a converted figure. It records the relationship between categories, with a type for each relationship, so that statistics can be compared and categories can be traced. The table is a statistical tool inside that limit. It does not extend the limit, and it does not decide any question outside it. A reader who treats a table row as a decision has crossed the limit that the summary states.

Which code questions does a table not decide?

The guardrails for this site state the point as a prohibition: a correspondence mapping must never be presented as a one to one guarantee where the table says one to many, many to many, or split. That prohibition applies to every guide, and it applies with particular force to table readings, because the table itself is the source that states the type. A row that is labelled one to many in the table must be described as a split, with the new classes listed and the notes cited. A row group that is labelled many to many must be described as a joint scope identity, with the group treated as a whole. A description that simplifies either row into a single new code, without stating the type and the notes, violates the guardrail, even if the simplified code happens to be one of the new classes in the row.

Table type	What the table states	What the table does not state
1 to 1	Identical scope, both directions	Assignment of a unit to the category
1 to many	Old scope divided among new classes	Which new class covers a particular activity
Many to 1	Old scopes combined into one new class	Old detail inside the new total
Many to many	Joint scopes identical, as a group	Any individual pairing inside the group

Which legal and administrative questions does a table not decide?

A table does not decide tax, social security, collective agreement, licensing, or funding eligibility. The classification is for statistical purposes only, and a code, in either version, does not determine eligibility under any of those regimes. A table row that shows a move from an old code to a new code does not show that a tax treatment, a licence, or a funding eligibility moved with the code, because those treatments are decided under their own rules, by the authorities that administer them, and they do not follow automatically from a statistical reclassification. A reader who cites a table row as evidence for an eligibility question has cited a statistical relationship for a legal question, and the citation does not support the conclusion.

A table also does not decide a registration entry. A register holds an entry under its own rules, on its own timetable, and the entry is confirmed with the register or the institute behind it. A table row may show that the category for an activity moved, but the register entry for a unit moves only when the register recodes it, using the notes and the rules, and the information held about the unit. A reader must not assume that a table row, published by Eurostat, has changed an entry in a national register on the same date. The two events, publication of the table and recoding of an entry, have different dates and different decision makers, and each must be checked in its own source.

Which statistical questions does a table not decide on its own?

A table does not, on its own, decide a converted figure, a backcast figure, or a double reported figure. It states the relationship type that decides which conversion operation is possible, and it identifies the group that must be treated as a whole. The figures themselves must be produced from underlying data, by the statistical office that compiles the statistic, and published with metadata. A table does not state shares for dividing a split total, weights for a conversion, or pairings inside a many to many group, and a series constructed by adding any of those without a source is not derived from the table. The table is the framework. The published series is the result.

A table also does not decide the version label for a dataset. A dataset must state its version, and the table for that version must be used to read it. A table for a different version, or for a different table version, may show different rows, and a row quoted without a version label cannot be verified. The version of the table, the direction of the reading, and the version of the classification for each side of the row, must all be stated beside any row quoted from the table, so that the row can be found again and checked.

Where is each question decided?

A category relationship is decided in the correspondence table, with its type. A category scope is decided in the explanatory note, with its inclusions and exclusions. A classification rule, such as the top down method, is decided in the introductory guidelines, as a description of the method. An entry is decided by the national statistical institute or the business register that holds the record. A legal or eligibility question is decided by the authority that administers the regime concerned, under its own rules. A table reading that keeps those places apart, and that cites each source for the question it decides, stays inside the statistical purposes limit. A table reading that merges them, by presenting a row as an assignment, an eligibility, or a converted figure, crosses the limit.

State the table type for the relationship, state the note for the scope, and confirm any entry with the national statistical institute or the business register that holds the record.