Comparing crime across cities is difficult

Problem 1: Crime definitions vary

A person smashes the window on a parked car and removes money from the glove box, intending to spend the money on drugs and without the permission of the vehicle owner.

London
theft from motor vehicle

Los Angeles
burglary of auto

Problem 2: Different cities publish statistics in different ways

Some cities publish incident-level data …

date_time type address
2025-11-23 23:31 res. burglary 13 High Road …
2025-11-24 07:15 res. burglary Flat A, 6 Gr …
2025-11-24 09:39 attmpt. res. burglary 780 Plane St …
2025-11-24 10:54 res. burglary McDonalds, 1 …

… others publish crime counts

month category count
2025 Jan homicide 2
2025 Feb homicide 4
2025 Mar homicide 3
2025 Apr homicide 3

Solution: harmonise categories to follow England & Wales recording rules

e.g. residential burglary

Barcelona:

  • robos con fuerza en viviendas (robbery with force in homes)

Los Angeles:

  • burglary and attempted burglary in any of residences, hotels, motels, halls of residence, mobile homes, care homes, foster homes or homeless shelters

New York:

  • burglary, residence, day
  • burglary, residence, night
  • burglary, residence, unknown time
  • burglary, unclassified occurring in a residence.

Problem 3: City boundaries affect crime rates

‘Small’ London

364 personal robberies
per 100,000 residents

8% higher than Paris

Actual London

305 personal robberies
per 100,000 residents

9% lower than Paris

‘Big’ London

279 personal robberies
per 100,000 residents

17% lower than Paris

Solution: construct a ‘version’ of each city that matches London

London
admin boundary matches urban area fairly well

Paris
admin boundary much smaller than urban area