Methodology

How a figure travels from its source to this page, and what can be checked at each step.

From source to capture

Every source has a sheet stating who publishes it, under which licence, how often and how it is read. The method of reading follows a fixed order: first the machine interface the agency provides, then its open data, and reading the published web page only when nothing else exists, with the reason written on the sheet.

Each download is stored exactly as it arrived and identified by the sha256 fingerprint of its bytes. That fingerprint is the capture shown under each chart: the API uses it to say when the file was captured and from which source, and anyone who downloads the same file from the source can compute its fingerprint and compare the two.

From capture to figure

Every indicator has a versioned methodology sheet. The sheet states what the indicator measures, which source it comes from, how it is calculated, the precision it is published with, its limitations and how it is revised. A published figure stays tied to the version of the sheet it was calculated with, to the capture it came from and to the run that produced it. That path is the lineage the API serves for each value.

The laboratory does not fill in what the source does not publish: a cell with no figure is published as such. When the source marks a figure as provisional or estimated, it is published with that mark, and no estimated, imputed or forecast figure is published without its mark and without the method stated on its sheet. When two agencies publish different figures for the same thing, the sheet cites both and says which one the series follows and why, without reconciling them.

Status of a figure

• Final: the source publishes it without reservation.
• Provisional: the source says it may change.
• Estimated, Imputed and Forecast: the figure is not a direct observation, and the indicator’s sheet states the method used to obtain it.
• Low reliability: the source or the sheet warns that the figure is imprecise.
• Break in series: the way it is measured changed in that period, so the figure cannot be compared directly with earlier ones.
• Not applicable: the figure cannot exist for that cell.
• Not collected: the figure exists, but the source did not collect it.
• Suppressed: the laboratory does not publish it, to protect the people behind it.

Protecting producers

Many figures in the sector come from registers of beekeepers. A cell covering very few producers, or one where a single producer accounts for almost the whole total, would reveal how much one person declared. Those cells are not published, and neither are the cells that would let someone work them out by subtracting from a total. Each indicator’s sheet states which rule applies and, when part of the rule cannot be evaluated with what the source provides, says so.

Corrections

A published version is never rewritten. If a figure changes because the source published a new edition or because an error was found, a new version is released and the previous one is kept. When the new version corrects something already published, the correction is recorded as an erratum, with its reason, on the bulletin page.