How to read London crime statistics without drawing the wrong conclusion

Start by naming the measure

“Crime in London” can refer to police-recorded offences, crimes reported by survey respondents, calls for service, arrests, charges, prosecutions or convictions. They answer different questions. A police-recorded-crime table shows what entered police systems under the published counting rules; it does not measure every offence that occurred.

Reporting behaviour matters. A campaign can increase reports of an offence even if the underlying prevalence does not change. Recording and classification practices also change. That does not make the data useless; it means a movement needs context before it becomes a claim about safety.

Know which London dataset you are using

The Metropolitan Police publishes detailed borough, ward and neighbourhood datasets through the London Datastore. MOPAC publishes dashboards that add public confidence, victim satisfaction and other policing measures. The Home Office data.police.uk service publishes anonymised street-level points across England and Wales and powers the location pages on this site.

A total is not a rate

Large or busy boroughs can record more offences because they contain more residents, commuters, nightlife or transport hubs. Dividing by resident population creates a rate, but even that can distort central London, where the daytime and visitor population is much larger than the resident count. State the denominator and explain why it fits the question.

Why the map is approximate

Street-level records are moved to anonymised map points to protect people involved. “On or near” is therefore not an address and can represent several nearby streets or a public place. Our location pages count a one-mile radius and say so; neighbouring circles can overlap and must not be added into a London total.

A minimum check before calling something a trend

Use the same offence definition, same area and same-length period. Check seasonality by comparing with the same period a year earlier, not only the previous month. Look at the count and rate, note boundary or recording changes, and test whether a few unusually large incidents explain the movement. If those checks are missing, describe the numbers, not a trend.