A murder map of South Africa [interactive]
I've long wanted to experiment with mapping crime in South Africa but this has largely been impossible since the police statistics don't give us much to work with. Crimes are reported at policing district level and at provincial level making it pretty impossible to do anything particularly useful with them.
This week I asked the incredibily helpful Lizette Lancaster, the project manager of the Institute for Security Studies Crime and Justice Hub, if she could assist my Media24 Investigations team with producing a special query on the police statistics which would show various crime categories at a precinct level by population.
I thought it would be interesting to see which were the most dangerous places in South Africa for murder and so on. The idea came to me from the insanely brilliant Crimehub site that the ISS have built which allows you to see crime information at a very local level and which tracks stats going back to 2004. I highly recommend this site for any journalist who is serious about data; it is probably the best implementation of this concept I have seen anywhere on the web.
The work Lizette did was amazing and produced a set of crime rate numbers for 385 policing districts in SA with populations over 45,000 which would reduce the per capita distortions that precincts with tiny populations would produce.
The results were surprising. Bloemspruit outside Bloemfontein is by far the most dangerous place in SA with a murder rate of 170:100,000 - compared to a national average of 31:100,000. Another surprise was that seven Cape Town precincts were in the top 10, dispelling the idea that Jo'burg is the murder capital of SA.
Based on her data, I was able to produce the map below which gives a pretty alarming picture of murder in South Africa. If you are interested in the murder data you can grab it from a Google Fusion table I created here . We have produced a pile of stories from this data and there are many more to be done.
I used Scribblemaps to produce the map and recommend you fool around with it if you want to do something similar. The geo-coding in this set is probably about 90% accurate so don't be surprised if you see a couple of odd markers.
