Maps The Curious Atlas

Half of London's street theft fits in fifty-seven wards

No open dataset counts stolen phones, so this maps the crime category they live inside — and below the borough line, London's street theft turns out to sit almost entirely in a handful of wards.

Published · Data as of

No open dataset counts stolen phones. Not one. The Office for National Statistics said so in terms, answering a freedom of information request on 8 December 2025: it does not hold police recorded crime data that specifically identify offences in which a mobile phone was stolen.

So this is not a map of phone theft, and I am not going to pretend otherwise in a footnote at the bottom. It is a map of THEFT FROM THE PERSON and ROBBERY OF PERSONAL PROPERTY — the two categories a stolen phone is recorded in, along with the wallet, the bag and the cash. A labelled proxy, and the whole question is whether the label is worth anything.

Here is the test I can run on it.

The proxy moves with the number it stands in for January to May 2026 against the same five months of 2025. The Met counts offences it has flagged as involving a phone; this counts the whole category those offences sit inside.

Over January to May 2026 the proxy category fell 21.6% across London where the Met reported 20.6% for phone-flagged offences, and 45.4% in Westminster where the Met reported 45.8%.

Show the numbers
Area This category Met, phone-flagged
London-21.6%-20.6%
Westminster-45.4%-45.8%

Source: MPS Recorded Crime: Geographic Breakdown (OGL-2.0) — full attribution

The Met published those two figures on 11 June 2026, counting offences its own officers had flagged as involving a phone. I never see that flag. I counted the whole category, in the same boroughs, over the same five months, and got within a point of the first number and within four-tenths of a point of the second.

A proxy that tracks the flagged series to within half a point on a single borough is measuring substantially the same thing.

Interpretation

That is as close to a validation as an open dataset allows, and it is the reason the rest of this piece is worth reading.

The borough is the wrong unit

Every published account of London phone theft stops at the borough. The borough is where the data was easy to get, and it is also where the finding goes to die.

Recorded theft from the person and robbery of personal property, by ward Twelve months to July 2026. Each ward is coloured by its own count, in named bands rather than on a continuous scale — the West End alone records more than the 327 quietest wards put together. The blank shape at the centre is the City of London, which has its own force and so is not in Met data at all.

Theft concentrates in a tight cluster of central wards around the West End, St James's, Bloomsbury and the South Bank; eight wards are in the top band, while 467 of London's 685 wards record under 100 offences in a year. The blank shape at the centre is the City of London, policed by its own force and absent from Met data — not a place without theft.

  1. under 25 143
  2. 25–49 177
  3. 50–99 147
  4. 100–199 121
  5. 200–399 65
  6. 400–999 24
  7. 1,000 or more 8

What this map shows

The 12 wards with the most recorded offences, of 685 in the file. Together they hold 28% of London's total.

  1. West End Westminster 8,970
  2. St James's Westminster 5,538
  3. Bloomsbury Camden 1,638
  4. Waterloo & South Bank Lambeth 1,350
  5. Borough & Bankside Southwark 1,338
  6. Holborn & Covent Garden Camden 1,173
  7. London Bridge & West Bermondsey Southwark 1,172
  8. Vauxhall Lambeth 1,143
  9. Hoxton East & Shoreditch Hackney 902
  10. Hyde Park Westminster 810
  11. Marylebone Westminster 797
  12. St. Georges Southwark 784

The hole in the middle is the City of London, which the Metropolitan Police does not police. Boundaries: ONS Wards, May 2024, super generalised.

Show the numbers
Ward Borough Last 12 months 12 months before
West End Westminster 8,970 16,842
St James's Westminster 5,538 9,120
Bloomsbury Camden 1,638 2,369
Waterloo & South Bank Lambeth 1,350 1,856
Borough & Bankside Southwark 1,338 1,660
Holborn & Covent Garden Camden 1,173 2,010
London Bridge & West Bermondsey Southwark 1,172 1,098
Vauxhall Lambeth 1,143 1,062
Hoxton East & Shoreditch Hackney 902 1,117
Hyde Park Westminster 810 895
Marylebone Westminster 797 974
St. Georges Southwark 784 784
Stratford Olympic Park Newham 761 1,234
North Walworth Southwark 757 657
Chaucer Southwark 753 687
Regent's Park (EK) Camden 736 811
Camden Town Camden 672 984
King's Cross Camden 664 734
Clapham Town Lambeth 628 611
Finsbury Park Islington 594 774
Stratford Newham 518 695
Spitalfields & Banglatown Tower Hamlets 509 544
Bunhill Islington 498 606
Clerkenwell Islington 495 633
East Ham Newham 464 519

The twenty-five highest of 685. Every ward is in ward-totals.json, which the piece is built from.

Sources: MPS Recorded Crime: Geographic Breakdown (OGL-2.0) · Wards (May 2024) Boundaries UK BSC (OGL-3.0) — full attribution

The map is mostly empty, and that is the finding rather than a failure of the map. Sorted by count, the wards fall away so fast that half of London’s total is held by 57 of them.

Half of London's total sits in 57 wards out of 685 Wards ranked by count, then accumulated. The curve is what a borough map cannot draw.

Cumulative share rises almost vertically: the top ward alone holds 10% of London's total, and 57 wards of 685 hold half of it.

Show the numbers
Top wards Offences Share of London
18,97010.0%
214,50816.1%
518,83420.9%
1024,03426.7%
2031,18034.6%
5042,98647.7%
10054,93361.0%

Source: MPS Recorded Crime: Geographic Breakdown (OGL-2.0) — full attribution

One ward, the West End, holds ten per cent of London’s recorded street theft. Add St James’s next door and the two of them hold sixteen per cent. There are 685 wards in the file.

What “Westminster” hides

Westminster is the borough every one of those articles named. Inside it, the spread is eleven to one.

Westminster is not a place where phones are stolen. The West End is. Every ward in one borough, ranked. The borough's total is a number about two of them.

Within Westminster the West End records 8,970 offences and Hyde Park 810, an eleven-fold spread across wards of one borough.

Show the numbers
Ward Last 12 months
West End8,970
St James's5,538
Hyde Park810
Marylebone797
Pimlico North461
Knightsbridge & Belgravia446
Regent's Park340
Lancaster Gate318
Vincent Square209
Church Street170
Bayswater122
Westbourne85
Pimlico South80
Abbey Road79
Harrow Road64
Queen's Park54
Little Venice51
Maida Vale48

Source: MPS Recorded Crime: Geographic Breakdown (OGL-2.0) — full attribution

The fall is real, and it is not displacement

The Met has been running an operation against phone theft, and the numbers have moved a long way. Across the twenty-four months in this file the direction is not subtle.

Twenty-four months, London The window starts in August 2024 on purpose: the Met moved to a new recording system, CONNECT, in February 2024 and is still validating what it recorded either side of that change. Nothing here crosses it.

Theft from the person peaks at 10,413 offences in November 2024 and falls to 5,620 by July 2026, a drop of 46%. Robbery of personal property is flat by comparison, moving from 2,175 to 1,931.

Show the numbers
Month Theft from the person Robbery of personal property
2024087,4032,175
2024097,0602,193
2024109,1332,461
20241110,4132,275
2024129,9562,225
2025017,8452,028
2025027,9711,961
2025038,5822,106
2025047,5782,041
2025057,1082,103
2025067,0002,078
2025076,8702,056
2025086,2211,943
2025095,8481,820
2025106,2741,927
2025117,6102,013
2025127,2521,901
2026015,9581,810
2026025,7501,670
2026036,1421,990
2026045,8481,899
2026055,6451,979
2026065,5962,004
2026075,6201,931

Source: MPS Recorded Crime: Geographic Breakdown (OGL-2.0) — full attribution

The obvious objection is that the thieves moved rather than stopped. They did not. Over the twelve months to July 2026, thirty of London’s thirty-two boroughs recorded fewer of these offences than in the twelve months before. Two did not: Southwark rose 0.4% and Barnet rose 3.8%. Westminster, where the operation concentrated, fell 39.8% — but so did Kensington and Chelsea, by 33.1%, and Camden by 27.6%, and Islington by 14.5%.

The fall is London-wide in direction and extremely uneven in force.

Fact

What is missing from this map

The ward file does not have a row for an offence whose ward is unknown. The borough file, counting the same offences over the same months, keeps it. Over the twelve months to July 2026 the ward file accounts for 93.2% of the borough file’s total, and the shortfall is not spread evenly: it is 9.0% in Westminster, 8.1% in Camden and 2.9% in Sutton.

That works against everything above, not for it. The offences this map cannot place are missing most heavily from exactly the boroughs whose wards top the ranking, so the concentration it draws is, if anything, understated.

The real curve is at least as steep as the one plotted here.

Interpretation

The other thing this map cannot tell you is how many of these offences were phones. It is not proof that the category is mostly phones — although the Met’s own arithmetic implies it is. The same release put the fall at 14,000 offences, or 18%, over the twelve months to May 2026, which puts the phone-flagged count at about 63,800; the whole category over those months was 99,504.

Roughly two in three of these offences involve a phone.

Interpretation

Behind the data

Method. Two files from the Metropolitan Police’s geographic crime breakdown on the London Datastore, ward and borough, each covering the twenty-four months to July 2026. Two SubGroups are read and nothing else: THEFT FROM THE PERSON and ROBBERY OF PERSONAL PROPERTY. Ward totals are the last twelve months against the twelve before. London and borough series come from the borough file, which is the complete one. The choropleth joins ward codes to Office for National Statistics ward boundaries, super generalised, and the pipeline refuses to run if any ward code fails to find a boundary other than the six retired ones described below.

Caveats. The categories are a proxy: no open dataset identifies a stolen phone, and these two hold wallets, bags and cash alongside handsets. Recorded crime measures reporting as well as offending, and phone theft is reported disproportionately because an insurer wants a crime number — a fall in the count is not automatically a fall in theft. The ward file places 93.2% of the borough file’s offences and has no bucket for the rest. Counts are offences, not devices and not victims. The window begins after the Met’s move to CONNECT in February 2024, so no series here crosses that break. Boundaries are Wards (May 2024), super generalised; six retired ward codes still carrying ten offences between them over the last twelve months have no boundary in any current release and are not drawn. Eight offences recorded under a code meaning “another force” are excluded from the ward totals and kept in the borough ones. The City of London is absent because the Metropolitan Police does not police it.

What would change my mind. A release of phone-flagged offences at ward level, by the Met or in answer to a freedom of information request, would replace the proxy at the centre of this and the piece would be rewritten rather than corrected. So would an independent measure of reporting rates by area: if people in the West End report a theft far more readily than people in Barking, then this is a map of who calls the police.

Sources

  • MPS Recorded Crime: Geographic Breakdown — Metropolitan Police Service, via the Greater London Authority (London Datastore). OGL-2.0. Retrieved . Monthly counts by Group and SubGroup at ward and borough level. This piece reads two SubGroups only: THEFT FROM THE PERSON and ROBBERY OF PERSONAL PROPERTY. Neither identifies a mobile phone: no open dataset does, and the Office for National Statistics said so in terms in FOI-2025-3264 on 8 December 2025. The counts are recorded crimes, so they measure reporting as well as theft, and phone theft is reported disproportionately because an insurer wants a crime number. Ward download slugs on the Datastore change with every monthly refresh, so fetch.py resolves them from the dataset page rather than hard-coding a URL. One property of the source matters to the map and was measured on 2026-08-23 rather than assumed: over the same twenty-four months the ward file accounts for 93.2% of the borough file's offences in these two SubGroups, because it has no row for an offence whose ward is unknown while the borough file keeps it. The shortfall runs from 2.9% in Sutton to 9.0% in Westminster, so it is largest where the counts are largest. transform.py prints the figure and refuses to run below 80%.
  • Wards (May 2024) Boundaries UK BSC — Office for National Statistics, Open Geography Portal. OGL-3.0. Retrieved . Ward polygons for the choropleth, super generalised to 200 m and clipped to the coastline. Which vintage was a live question — ISSUES #036 — and it was settled by joining six candidate releases against the 685 ward codes the Met publishes. December 2022, December 2023, May 2024 and December 2024 each left exactly the same six codes unmatched, and those six are the retired wards the Met still carries, so the test cannot separate them: London's wards did not change across that span. This release was chosen because it is the most recent super generalised one that is both populated and queryable. The December 2024 super generalised service is published and returns zero features; the May 2025, December 2025 and May 2026 releases exist only as MapServer and WFS and their query endpoint refuses. The December 2024 generalised-clipped release does work and was rejected on size: 25,843 points for London against 7,156, for detail this map cannot resolve at roughly 100 m per pixel. The envelope covers Greater London with a margin; wards outside the 32 boroughs are dropped in transform.py. The City of London's 25 wards are in this file and carry no data here, because the City is policed by a different force.

Attribution

  • OGL-2.0 MPS Recorded Crime: Geographic Breakdown Contains public sector information licensed under the Open Government Licence v2.0. The GLA cannot warrant the quality or accuracy of this data.
  • OGL-3.0 Wards (May 2024) Boundaries UK BSC Source: Office for National Statistics licensed under the Open Government Licence v.3.0. Contains OS data © Crown copyright and database right 2024.

Data as of . Last fetched . Pipeline code is available on request.