Westminster is not London's most dangerous borough

Every crime rate is a fraction. The top half is offences. The bottom half is people. Change your mind about who counts as "people" and the same borough can look like a war zone or a village.

London's official figures put the offences on top and the resident population underneath. For most of the country that's fine. For central London it produces a number that is close to meaningless.

Westminster records 535 offences per 1,000 residents a year (1st highest of 33) — roughly ten times Bexley records 54 offences per 1,000 residents a year.

Nobody who has walked through both would describe the gap that way.

The denominator is wrong

About 210,000 people live in Westminster. On a weekday something like a million are in it — commuters, tourists at Buckingham Palace and Oxford Street, students, theatre audiences, people changing trains. Police.uk records an offence where it happened. The census counts residents where they sleep. Divide one by the other and you have measured, fairly precisely, the wrong thing.

The same applies to the City of London, which is second on that list and has a resident population of around 8,600 against a weekday working population of over half a million. Its rate is arithmetic, not danger.

What the rate is genuinely good for

None of this makes the figure useless. Two things it does well:

Comparing like with like. Between two residential outer boroughs with similar commuter flows, the rate is a fair comparison. Bexley against Sutton against Harrow tells you something real.

Tracking one place over time. A borough's denominator changes slowly. If its rate climbs 20% in two years, that is a genuine signal, whoever is doing the counting.

What it cannot do is rank a tourist district against a suburb.

Why we score the count, not the rate

This is the reason AreaScore's crime score isn't built on the per-resident rate at all.

What it uses instead is the number of offences recorded within a mile of your postcode over the last twelve months, ranked against every other postcode in London. No denominator, so no denominator problem. It answers a question you can actually act on — how much crime is recorded around this address, compared with everywhere else I could live — rather than a question about a borough's daytime population.

It has its own limits, and they're worth stating plainly:

• A mile is a mile. A postcode near a busy high street picks up that high street's offences even if your own street is quiet.
• Recorded is not committed. Under-reporting varies by offence type and by area. Bike theft and phone snatching are notoriously under-reported.
• A count is not a risk. It is not adjusted for how many people are around to be victims. That is the deliberate trade — we would rather have an honest count than a rate whose denominator we don't trust.

The map shows the raw figures for all 33 boroughs and 283 postcode districts without any scoring on top. Every borough also has its own page with all 21 measures and where it ranks on each — Westminster's is a good one to start with, precisely because its crime figure needs the context above to read properly.

The general lesson

Whenever you meet a statistic about a London neighbourhood, the useful question is rarely "is this number accurate?" It usually is. The useful question is "what is underneath the line, and is it the population I care about?"

Recorded crime per resident, school places per child, green space per hectare, GP appointments per patient — every one of them is a fraction, and every one of them can be made to say something ridiculous by picking the denominator carelessly.