tisdag 4 april 2017

Concentration of gun violence in three Swedish cities

A few months ago the crime analyst Jeff Asher from New Orleans posted some maps highlighting the concentration of gun violence in New Orleans and in Chicago/Baltimore on twitter. Since I've been working on gun violence for a while with Joakim Sturup, Amir Rostami and Anders Sandholm I was intrigued by Jeffs analysis and now I have finally attempted to do something similar for Sweden.

As I will show below gun violence appears to be at least as concentrated in Sweden as it is in New Orleans. More generally there are however some notable differences between the US and Sweden. Most importantly, gun violence is much rarer in Sweden. In our study on near repeat patterns for shootings we identify 948 shootings over 5 years across the three biggest Swedish cities (938 of which we could geocode), and in the case of Stockholm this includes the whole county with over 2 million residents. Our study areas in total cover about 3 million residents, meaning the gun violence rate is around 6 incidents per 100 000 population and year. This includes all confirmed firearm discharges reported to the police - not just incidents where someone actually was hit by a bullet. Counting shootings with a casualty drops the number to 378, about 2.5 shootings per 100 000 residents and year. This can be compared with the 462 shootings reported by Jeff Asher for 2016 alone in New Orleans.

In the city of Malmö 2011-2015 there were a total of 289 shooting incidents, of which 136 involved someone injured or killed. 44% of the 289 incidents took place in 11 neighborhoods comprising 16% of the population or 4% of the municipal area. And 46% of incidents with a casualty take place in 10 neighborhoods comprising 15% of population or 4% of the area. This can be compared with the 16.7% of populated land accounting for 54% of shootings in New Orleans as reported by Jeff Asher. The map below includes land not populated, but as can be clearly seen in the map below (click to see a bigger version) the shootings are also very concentrated within these neighborhoods. Actual concentrations would be much higher if using a smaller level of aggregation - for instance as shown below regarding the Stockholm and Gothenburg examples.
Shootings with casualty in Malmö municipality, 2011-2015
For Stockholm and Gothenburg I have much less neighborhood level data. But to just visualize the patterns I have done maps that are similar to that of Malmö, but which depart from a national area designation used by the police. These areas are smaller than the ones used in the Malmö map above, and the maps are not entirely comparable, but to show the concentration of gun violence they should work fairly well.

In Stockholm the pattern of concentration appear to be similar. The map below shows that 12 areas of the city with at least two gun violence casualties are recorded for 33 out of 70 shooting incidents with injury or death. Notably however the rate is much lower in Stockholm than in Malmö, with a total of 197 shootings, 70 of which with a casualty, and almost 3 times bigger population. In addition the areas with high concentration of shootings are much more spread out than in Malmö. The proximity of disadvantaged neighborhoods with criminal networks in Malmö has previously been suggested to be one of the factors that may help explain the high rate of gang violence in the city.

Shootings in Stockholm municipality, 2011-2015
In Gothenburg 11 areas with at least 2 gun violence casualties are recorded for a total of 33 out of 88 incidents with an injury or death, or 37.5%. While this percentage is small, this is just 11 out of 1110 areas (the map doesnt show the whole municipality). Those 11 areas only account for 0.7% of the land - but this number is clearly somewhat mis-representative as the municipal boundaries used here include large areas of sea to cover the archipelago. As mentioned above I dont have much data over Gothenburg and Stockholm readily available, but while I cannot give an accurate number of just how high the area-level concentration is, I can with some confidence say that the concentration is very high. The rate in Gothenburg is higher than in Stockholm, but lower than in Malmö, and similarly to Stockholm the areas with many shootings are quite spread out across the city.
Shootings in Gothenburg municipality, 2011-2015. 
 The pattern regarding concentration of gun violence is nevertheless quite clear. In all three cities a small part of the city accounts for a large share of the shootings. This can be related to the law of crime concentration which states that a small number of places account for a majority of crimes. That it is applicable to shootings is not surprising, but it is interesting none the less. The fact that concentrations of shootings appear to be fairly similar in otherwise dissimilar countries such as the US and Sweden points to some stability of the patterns of concentrations. To further assess the issue we'd need to do more rigorous calculations however, which lie beyond the scope of this blog post.