Categories from register codes — Task

Implementation: Categories from register codes (Recipe - QGIS).

Categories from register codes is the Task of turning the codes of an administrative register (BBR, CVR, CPR-based statistics) into the categories an analysis needs: construction period from a construction year, tenure from a rental code, ordinary dwelling from a use code. Registers are made for administration, not for research, so their categories rarely match a research question one to one. This Task is where a concept is operationalised, and where most of the analysis’s assumptions are made.

Used in examples: Dwelling size by construction period in Copenhagen (Example).

Task Schema (tool-agnostic)

1. State the concept (decision point) — what the analysis is about, in words, before looking at the data (e.g. “dwellings built before the Second World War”).
2. Find the codes and read the code list (decision point) — which fields can express the concept, and what each code means according to the register’s own documentation (for BBR: teknik.bbr.dk/kodelister).
3. Count the codes (Operator: aggregate per value) — how many units have each code, including empty values, before deciding anything.
4. Decide the mapping (decision point) — which codes go into which category, and what happens to codes that fit none (left out, or a category of their own).
5. Select the population (Operator: selection by attribute) — the units the analysis is about (e.g. standing dwellings, without student housing).
6. Write the category or indicator (Operator: indicator or recode).
7. Check the counts again — every unit should be in exactly one category, or deliberately left out.

Output: the units with a new category or indicator field, and a table of how many units each code and each category holds.

Epistemic Choices to record

• The code’s real meaning: read the definition, not only the label. A construction year (opførelsesår) belongs to the building, so a dwelling converted from an old warehouse carries the warehouse’s year.
• What the code list cannot say: a register has the categories its administration needs. If the category your concept needs is not in the list, it is hidden inside another category; find out which, and say what that does to the result.
• Empty and residual codes: “unknown”, “not in use” or an empty field are decisions too: in the denominator, or out.
• Who reports the codes: who registers a code, when it is updated, and whether anyone has a reason to keep it correct. A code nobody uses administratively can stay wrong for decades.
• Date: registers change; give the date of the extract.