Using open crime data from police.uk and map data from OS OpenData, I’ve created a map of violent and sexual crime hotspots in London, based on 135,000 crimes that occurred between April 2012 and March 2013. The map is available as an A0-sized PDF under a CC-BY licence.
The most fun thing about this sort of map is looking at how your neighbourhood compares to others, but lots of other interesting things come up if you have a look around.
The top 20 hotspots
Using the excellent (and free!) CrimeStat III software1, I produced a list of the 20 areas with the highest number of violent and sexual crimes. More than 2,000 violent or sexual crimes were recorded in the number one hotspot between April 2012 and March 2013, and each hotspot had more than 750 recorded crimes within a 1km radius. This showed that violent and sexual crime in London is heavily concentrated in a few areas: half of offences happen in less than 1% of streets while 78% of streets had no violent or sexual crimes in 2012–13.
- Soho and Leicester Square
- Camden Town
- Hackney town centre
- Ladbroke Grove and Kensal Town
- Elephant and Castle
- Peckham town centre
- Croydon town centre
- Stratford town centre
- Shepherd’s Bush
- Kings Cross and Angel
- Edgware Road and Marble Arch
- Kingston town centre
- Wood Green
No hotspot-detection method is perfect: there may not be much real difference between, for example, number 15 and number 16 on this list, or even between 15 and 18. However, this list does provide a starting point for thinking about how violent crime is clustered within London. For me, three types of hotspot stand out: central London, suburban centres and old arterial corridors.
Central London is in a class of its own: seven of the top ten violent-crime hotspots fall within an area between Camden Town in the north, Poplar in the east, Brixton in the south and Hyde Park in the west. The area around Soho and Leicester Square has the highest concentration of violent crime anywhere in London, at-least partly because it has vast numbers of people passing through it to visit shops and, especially, bars. The same applies to the next three hotspots on the list: Shoreditch, Brixton and Camden Town.
Several of London’s suburban centres seem to suffer a lot of violent or sexual crime while being surrounded by low-crime areas. In the cases of Croydon (particularly around the High Street), Kingston (around the train and bus stations) and Romford (South Street), this is probably due to these areas attracting lots of late-night revellers from surrounding suburbs, as anyone who’s spent a Friday night in any of them can attest2. In the cases of Romford and Croydon—but not Kingston—these small areas of crime have given an entire borough a bad reputation, which must be galling if you live in ultra-leafy Hornchurch or Shirley.
Old arterial roads
Before the major road-building programmes of the 1960s, people got in and out of London along long single-carriageway roads. Many of these live on as local high streets, several of which stand out as hot routes. They usually have lots of small shops, pubs and restaurants, so there are plenty of potential offenders and victims around.
Uxbridge Road (top) leads from Shepherd’s Bush Green to, unsurprisingly, Uxbridge in far-west London, and every high street along that route is a mini crime-hotspot. This is particularly true in Acton, where the hotspot extends south towards Acton Town station and the South Acton estate, and Ealing, where the Broadway shopping area is clearly visible.
The A23 road (left) has one top-20 hotspot at each end (Croydon and Brixton), with violent crime remaining high almost all the way through Streatham (where the High Road feels pretty dicey day or night), Norbury and Thornton Heath. Even clearer on the map is the A10, which can be seen as a hot route almost from its start outside Monument station north through Shoreditch, Dalston, Stoke Newington, Seven Sisters, Tottenham and Edmonton.
How I made this map
The crime data released on police.uk is quite difficult to deal with for two reasons. The first is that crimes are only described as falling within one of several very broad categories. The violent and sexual crimes category includes all types of assault, from murder and manslaughter to assaults in which the victim is not injured. Also included are harassment, threatening to kill another person and causing death by dangerous or careless driving, as well as all sex offences (even if they do not involve violence) and rare offences such as kidnapping and people smuggling.
More importantly, although a set of co-ordinates is given for each crime, they are not the co-ordinates at which the crime occurred. Instead, the co-ordinates give the location of the nearest ‘snap point’ to each crime. Snap points can be places of interest such as a hospital or a shopping centre, but the vast majority are found at the centre of 750,000 street segments across England, Wales and Northern Ireland. This means that we do not know the precise location of most crimes.
To make sure the map did not suggest that the crime locations were more precise than they actually are, I used kernel-density estimation3 (KDE) to produce a smoothed estimate of crime incidence across London. I then gave each street segment the average (mean) density value of all the KDE cells that it passed through.
As with everything that is based on police-recorded crime data, it is important to remember that not every crime is recorded by the police. Nationwide, only about 40% of violent crimes are reported to the police. Reporting rates are particularly low for domestic abuse and some sex offences. Nevertheless no other source of crime data provides details about crime at a local level, so police-recorded data is the best that we have.
1 I did this using nearest-neighbour hierarchical spatial clustering with a random nearest-neighbour distance, a standard search radius, two standard deviations for ellipses and a minimum of 667 points in each cluster. Since only the top 20 clusters are used, each one represents more than 750 crimes in 12 months.
2 If you are interested in violent crime and haven’t spent a Friday night in any of these areas, I highly recommend making a trip to any of them.
3 Using a quartic density function with 50-metre cells and a 250-metre bandwidth.