The numbers on this site change for two different reasons: the country changes, or we change how we measure it. The second kind is recorded here, with what it did to the numbers.
Every artifact is stamped with the methodology edition that produced it, so an old figure can always be traced back to the rules behind it. How we calculate it describes the current edition.
The map now shows a grade for each of Denmark's 2,096 measured parishes, calculated the same way as the municipal ones. It is a considerably finer resolution, and it changes what the map says: four in ten parishes are graded G, but they hold only about an eighth of the population.
Seven parish boundaries were geometrically defective in the source data from Dataforsyningen, so their outlines crossed themselves and could not be filled. They have been repaired.
The largest method change so far. Four things were corrected at once, and each was measured separately before the edition was published, so it is visible which change moved what.
Workplaces now come from CVR. We previously used town centres as a proxy for work, and that proxy counted towns rather than jobs: a municipality scored higher for consisting of many small towns. Each active production unit is now counted at its registered premises, that is, where work actually happens.
Destinations are now collected in a radius around the municipality instead of being clipped at the municipal boundary. A municipality was previously penalised for having its neighbour's supermarket and hospital just across the line, even though residents obviously use them. That correction moved public transport reach the most.
Public transport quality is now measured on an absolute scale. The 98 municipalities were previously scaled against each other, so a municipality's grade could change because another municipality moved. Each destination group now reaches half its score at a fixed, published constant, and the calculation never looks at where the others sit. A grade can only change if the place's own mobility changes.
Finally, the grade is now given cell by cell and aggregated afterwards, not the other way round. That corrects a systematic overstatement for municipalities with large internal differences. The effect is small, but it always points the same way: Lolland was 2.4 points too high under the old order, Læsø 2.2.
The trend charts used a derived national car-availability figure instead of Statistics Denmark's own. They now use the authoritative series.
The map's public transport lines were drawn with a simplification that cut corners across streets, so neighbouring services could look like unrelated straight lines. The lines now follow the road.
The measure of everyday life without a car compares public transport reach with the car's. That on its own is blind to magnitude: two near-empty reaches look like perfect parity, because the car has no advantage when it gets nowhere either.
There is now a floor. If combined reach falls below it, the measure is pulled down in proportion to how little there is to reach at all. It affects only the far tail, around the first percentile of the national grid, but that is exactly the tail where a thinly populated place could otherwise score like an inner city.
Population determines how much each grid cell weighs when the numbers are aggregated to parish, municipality and region. The first edition used Kontur's grid population both to distribute people within a municipality and to set the municipality's total weight. The distribution was usable, the totals were not: Copenhagen stood at 1,001,310 inhabitants against 673,438 in DST's FOLK1A, Tårnby at 195,504 against 44,230.
DST now sets the municipal total, while Kontur only distributes it across cells. The correction is material: 36 of 98 municipalities and 818 of 2,087 parishes changed grade, and the largest single change was 8.05 points. The mobility model itself is unchanged, and the ranking between municipalities is largely the same, with a Spearman correlation of 0.9915.
The boundaries between the letters were fixed against the distribution from the first run across all 98 municipalities. They are absolute, not quantiles, precisely so that a municipality's grade can never change because the others move.
The criteria were set in advance: at least five of the seven bands had to be in use, and no single band could hold more than 35 per cent of the municipalities. Both held, and all seven bands ended up in use.
The first version of the method: two measures, weighted equally, computed on a grid across the whole country from Rejseplanen timetables, the OpenStreetMap road and walking networks, and Statistics Denmark's definitions.