Four Hundred Districts, Ten Years
Three districts around Berlin grow far faster than their structure predicts. Nine look like they are under housing pressure — seven of them are not. A regional picture of Germany where every number comes with what it can and cannot support.
Somebody asks where the next site should go, or which regions the sales plan should weight, or where demand for a service will be in five years. The answer is supposed to come out of regional data.
The problem is not getting the data. Germany publishes it, openly, in extraordinary detail. The problem is that a district-level table will happily tell you nine districts are under housing pressure, and seven of them are not — and nothing in the table distinguishes the two.
The short version: a namespace covering all 400 German districts over a decade — 8 sources, 27 detectors, 1,775 findings, 8 investigations. It says which districts grew and which only appear to, which are shrinking and ageing at once, where crime statistics measure something other than crime, and where the strongest-looking signal in the data dissolves on inspection.
What follows is four of those investigations and, more usefully, what made each one survivable.
Berlin’s commuter belt, and how to be sure
Three Brandenburg districts ringing Berlin grow far faster than anything about them in 2015 predicts.
The model is an ordinary least squares fit on 2015 age structure, density, East or West, city status and foreign share — it explains about half the variance across the 400 districts (R² 0.48). The three belt districts sit well above it.
The obvious explanation is suburbanisation out of Berlin, and the namespace can check it rather than assert it: Berlin → Brandenburg is Germany’s largest net migration corridor, as its own finding on its own asset.
The more interesting part is the hypothesis that failed. “The belt grows mainly through foreign population” is recorded in the case as refuted — the growth is domestic, not migration-driven. A second hypothesis, that Dahme-Spreewald’s growth is partly the airport, remains proposed: plausible, not yet tested.
Three states — supported, refuted, proposed — on three claims about the same districts. That is what a defensible regional finding looks like.
Nine districts under housing pressure. Seven are not.
A reasonable detector: flag districts whose population grew more than 3% faster than their housing stock. Both sides census-adjusted. Nine districts fire.
Then the namespace checked the detector against the raw, unadjusted figures — and most of the signal turned out to be the census adjustment itself, not a divergence between people and dwellings.
Two of the nine survive: Leipzig and Ludwigshafen are under real housing pressure. For the other seven, the signal was the correction.
We are telling you about this one because it is the outcome most analyses never report. The detector was not deleted; the finding was not quietly dropped. The case holds both hypotheses — “Leipzig and Ludwigshafen are under real housing pressure” and “for seven of nine, the signal is the census correction” — and both are supported, because both are true.
If you are deciding where to build, the difference between nine districts and two is the whole decision.
Where services come under pressure first
The sharpest forward-looking finding in the namespace is also the least surprising, which is how you know the machinery is working.
Districts that shrink and age simultaneously are overwhelmingly East German. That much anyone could guess. What makes it actionable is the second condition: in some of them, Grundsicherung im Alter — basic income support in old age — is already growing faster than in the rest of their state.
Shrinking, ageing, and a rising share of pensioners who cannot cover their own costs is a different proposition from shrinking alone. It tells a regional planner, an insurer or a care operator which districts move first.
A third hypothesis — that care capacity is the binding constraint — sits at proposed, with the test written down and not yet run.
What crime statistics actually measure
Police crime figures per district look like an obvious input to a location decision. Read naively, they are close to useless.
Immigration-law offences are recorded where they are detected. A district with an international airport or a national border therefore accumulates offences generated by people who are passing through, not living there. Those districts top the ranking.
Remove them and recorded crime follows urbanity and social strain — the structural factors you would expect.
The namespace is deliberate about what it does not do here. No detector in it relates crime to migration, and the case states that in writing. It also states the larger caveat: recorded crime is not all crime, and a ranking of recorded crime is partly a ranking of how thoroughly each force records.
Two corrections, both published next to the number, neither hidden in a methods note.
The composite score we demoted
One investigation is closed, and it closed by concluding that its own subject was not worth much.
A composite state ranking — economy 30%, demography 20%, employment 20%, housing 15%, business 15% — is the kind of index that ends up on a slide. The namespace ships it only with a rank band: the range the state occupies across 2,000 random reweightings of those five components.
The result: only the ends survive. Bayern lands between 1st and 3rd whatever the weights. Thüringen between 15th and 16th. Berlin is anywhere from 2nd to 10th — its rank is a statement about the weights, not about Berlin.
The case is closed with the conclusion “answered, and demoted”. The band stays on every state record, and the standing question keeps matching, so the next data release re-tests it automatically.
Publishing an index with the range that survives reweighting is a small change that makes the difference between a number and a decision input.
Districts that only grow through migration
162 of the 400 districts hold their population or grow only because their foreign population grew while the rest shrank. Almost all are in the West.
The investigation is careful in a way worth copying. Naturalisations move people from “foreign” to “German” without anyone moving house, so the non-foreign figure is a lower bound on the German-national side — stated in the case, not discovered later by a reader.
One hypothesis about benefit take-up is marked inconclusive, and the related claim that the 2022 change of eligibility for refugees from Ukraine drives the pattern is supported at state level and explicitly not testable at district level, because the table that would split it does not exist.
“We can show this nationally and not locally, and here is the table that would have to exist” is a more useful output than a confident district map.
What Classifyre contributed
The analysis is ordinary regional statistics. What makes it hold up is structural.
Every caveat travels with its number. A district that carries a census correction, a central-booking artefact or a registration-seat effect has that recorded as a finding on the district itself. Open the district page and it is there — including findings recorded on other assets that point at it, such as an integrity check that is its own object.
Questions outlive the analysis. A standing question — “which districts grow only through migration?” — keeps matching as new releases land, and says what changed. The alternative is a notebook somebody has to remember to re-run.
Hypotheses can fail in public. Across this namespace and its two siblings there are 62 hypotheses, every one with a written testable predicate. Four are refuted, two inconclusive. One of the refuted ones is in this namespace — the Berlin belt’s foreign-population explanation — and killing it is what made the suburbanisation reading credible.
Sources stay current on their own schedule. Regional boundaries change rarely and run quarterly; crime statistics arrive annually; the catalogue is walked weekly. Nobody schedules a refresh by hand.
Who this is for
If you make decisions on German regional data — site and catchment selection, regional sales planning, service and care capacity, insurance and credit exposure by region, public funding applications, ESG and social reporting — this namespace is a worked example of the same decisions made with the caveats attached.
Two things transfer directly regardless of subject: test a detector against the raw figures before believing it, which is what turned nine housing-pressure districts into two; and publish the range that survives reweighting next to any composite score.
See it yourself
The namespace is live, with every source, detector, standing question, case and lineage edge described above:
showcase.classifyre.com/de-regionen
Start with Berlins Speckgürtel and look at the refuted hypothesis first — then follow one belt district back through its findings to the tables underneath.
More case files
The Quarter Before the Numbers Say So: Early Warning From Public Statistics
Schleswig-Holstein's company insolvencies rose 36.6% in one quarter. Building permits collapsed two years earlier. And the signal everyone treats as leading turned out to lag. What monthly public data can and cannot warn you about.
Köln Is Not Shrinking: When the Official Number Is Not the Number
Germany's official population figures say 44 districts shrank over the last decade. Once the 2022 census correction is removed, those same districts grew. Here is the namespace that tells the two apart — and what else it found hiding in plain sight.
Who Am I Actually Dealing With? Austria's Company Register, Made Answerable
Two Austrian companies, both registered in 1992, both "active". One filed 21 sets of accounts, the other has filed none in 34 years. Here is how we built the namespace that tells them apart.