Method and sources

Crimistat shows offences recorded by the police in Belgium: the figures the federal police publish in their quarterly crime statistics (Politiële Criminaliteitsstatistieken). Everything here is a re-presentation of those official reports plus population figures from Statbel — no estimates, no modelling.

Coverage

Geography: Belgium, 3 regions, 11 provinces, 14 judicial districts, 175 police zones, 564 communes.
Years: 2000, then 2015–2025. The police do not publish 2001–2014 — those years are available only on request through BIPOL, which is why every chart here has a visible break.
Categories: 43 offence categories plus the “all recorded offences” total (report part I), and 90 crime phenomena in 12 groups (report part II).
Vintage: reports for 2025 Q4; the source closed the data on 24 April 2026.

How to read the figures

Recorded, not committed

These are offences that reached the police and were registered. Reporting rates differ sharply by offence type and by place, so a higher figure can mean more crime, more reporting, or more police activity. Belgium’s security monitor survey is the source that measures the gap; it is not part of this site yet.

Rates use a fixed population

Rates per 100,000 use the population on 1 January 2026 for every year, because that is the only commune-level population series currently loaded. Comparisons between places in a given year are sound; a rate change over time reflects recorded offences only, not population growth. A per-year population series would fix this and is the next data step.

Phenomena cannot be added up

The crime phenomena in part II overlap: one offence can carry several phenomenon labels. Summing the components of domestic violence for 2025, for example, gives 54,059 against an official combined figure of 50,078 — an 8% overstatement. Each phenomenon is valid on its own and over time, never as a sum.

Parts do not add up to the whole

Some recorded offences are not attributed to a specific municipality. Summing all 564 communes therefore lands 0.6–2.3% below the national total, depending on the year, and summing the 175 police zones lands 0.5–0.7% below. This is a property of the source, not a processing error.

Mergers break the time series

Belgium merged a number of communes on 1 January 2025 (and again the police zones). Merged units such as Beveren-Kruibeke-Zwijndrecht, Merelbeke-Melle or Tessenderlo-Ham carry a series that starts only when the merged entity does, and comparing them with pre-merger years is not meaningful.

Known gaps

Eeklo has no commune report in the published set, so it is grey on the map. Its offences are still inside the Meetjesland Centrum zone total.
The Heusden-Zolder police zone has no zone-level report; the commune figures exist.
Report parts III–VI (ICT detail, place of the offence, goods and vehicles, suspects) are published as PDF and are not loaded here.

Context layers, and what a correlation here does not mean

Besides recorded offences, the map can show who lives in a commune: population density, the share aged 0–14, 15–29 and 65+, median age, population and area. Age structure comes from the Statbel population register (1 January 2026); area is measured from the official commune boundaries, and density is population divided by that area. For police zones these are aggregated from their communes — density from the combined area, age shares weighted by population.

Every commune page also carries a scatter plot of an offence rate against one of those variables, across all 564 communes, with Pearson and Spearman coefficients. Three things to hold on to when reading it:

It is a relationship between places, not between people. A commune-level association tells you nothing about who commits offences — inferring individual behaviour from area averages is the ecological fallacy, and it reverses real relationships often enough that statisticians named it after a 1950 study that found exactly that.
Both columns share the same drivers. Urbanity, income, age structure, and how many people pass through a place daily push crime records and demographic composition in the same direction, so much of any correlation is those factors, not a link between the two variables shown.
Denominators are residents, not people present. A seaside town or a city with a big station serves far more people than it houses, which inflates its per-resident rate.

The dashed line on the scatter plot is an ordinary least-squares fit drawn as a visual aid; it is not a model and carries no controls.

Building the map

Police reports carry no NIS codes, so commune names were matched to the official Statbel geography and every match verified. Each commune is attached to its police zone spatially, from the 2024 police-zone boundaries, and the 2025 zone mergers were resolved by checking that the communes of the merged zones reproduce the published zone totals exactly, in all twelve years. Zone outlines on the map are the union of their communes, so the two map layers stay consistent with each other.

Sources

Federal Police — police crime statistics (DGR/DRI/BIPOL), quarterly reports, 2025 Q4 edition. Reuse with attribution.
Statbel Open Data — population by commune, sex and age on 1 January 2026, and the REFNIS code register. Open licence with attribution.
data.gov.be — commune boundaries (GeoJSON).
geo.be — police zone boundaries, 2024 edition, and the commune ↔ zone reference table.

Crimistat is an independent presentation of open data. It is not affiliated with, or endorsed by, the Belgian federal police or Statbel.