Defend Truth


We have to dig deeper into crime statistics to interpret them more meaningfully


Jenny Hillebrand hacks data as a hobby born out of the questions “I wonder what this would look like?” and “Couldn’t we do this better?”. She worked in the IT industry as a programmer and systems designer in the days when COBOL was a thing, during which time she obtained a three-year Diploma in Datametrics through Unisa in the hope that she might finish a BSc degree. Her latest achievement is completing the Stanford University Machine Learning MOOC presented by Dr Andrew Ng (but without paying for the certificate.) In between she graduated to programming in Python and developed her analytical skills through Project Euler. Her day (and often night) job is as a Methodist minister serving churches in Parow and Elsies River. She has a PhD in Theology from UKZN (which she completed and for which she has the certificate).

It is more meaningful to look at the number of murders per number of people than to look at just the number of murders. There are difficulties with this, but if we are not interested in an exact number, but rather patterns and insights, these difficulties may be overcome.

The release of the latest crime statistics has been of much interest. For those with little understanding of how statistics work, an exclamation mark after a figure often means it is very bad and should cause offence.

The word “only”, preceding a figure, may mean that it is not serious and objections should be taken lightly. This sort of statistical illiteracy means that when we are told that Nyanga is the worst place for murder in South Africa, we easily believe it, as we are told that another police station is urgently needed. We are not always told, however, that Nyanga is one of the most densely populated areas in Cape Town and it is thus not surprising that more murders are committed here than elsewhere.

It is more meaningful to look at the number of murders per number of people than to look at just the number of murders. There are difficulties with this, but if we are not interested in an exact number, but rather patterns and insights, these difficulties may be overcome.

The 2011 census counted and published the number of households across South Africa. This is not the same as the number of people, but it is indicative of the number of people. As it is now 2018, and the 2016 Community Survey shows an increase in the population of Cape Town of about 7%, the figures will be somewhat out of date, but they can still give us insight, if not exact numerical answers.

If we use the Small Area Layer census figures we can calculate the approximate number of households in each police precinct in 2011, across Cape Town. These figures will not take into account population growth in the last seven years, and we keep in mind that certain areas will have grown significantly more than others. We then take the number of murders in each police precinct and create a ratio. We are able to plot these onto a map and look for insights.

Click here to go to the interactive map and view police station details. (Works best on a device with a mouse.)

The green flags indicate the lowest murder ratio, with the rate increasing through yellow, orange and then red. The size of the circle around each flag represents the size of the ratio. Perhaps surprisingly, the two red flags are in Philippi, not Nyanga. However, this may be because Philippi is a high growth area and if we were able to use more recent population data the ratio would come down. It would certainly be worth taking a closer look.

Nyanga has an orange flag. It is one of a cluster of orange flags. As can be seen, each of these precincts has a similar ratio value. The value itself has little meaning because of the age of the population data, but unless one area has had disproportionate population growth, the comparison remains valid.

Thus, living in Bishop Lavis is similarly dangerous to living in Nyanga. This doesn’t mean that the murder rate in Nyanga is acceptable – but rather that those areas with orange flags neighbouring Nyanga are also areas of concern.

Looking at the data in this way raises interesting questions. The Gugulethu and Manenberg police stations are very close together – yet the murder rate is still high. Would it help if there was another police station in Nyanga? Perhaps not.

The cluster of orange flags has a ring of yellow flags around it and then a ring of green flags showing layers of decreasing crime. Is the orange centre going to push out, turning yellow flags orange and green flags yellow? Or can the tide be turned and allow the green flags to push in and turn orange to yellow? What would need to be done to achieve this? What do we make of Kraaifontein, which is an orange flag, a little way out of the cluster? Can that area be turned around before it gets any worse?

The pictures and the numbers don’t prove anything – but they give a better idea of what is happening than just the number of murders. It would be necessary to check the population growth in each area, but also very interesting to see if there are creative answers to the questions raised. DM


Please peer review 3 community comments before your comment can be posted

MavericKids vol 3

How can a child learn to read if they don't have a book?

As the school year starts again, thousands of children will not have the basics (like books) to learn from.

81% of children aged 10 cannot read for meaning in South Africa.

For every copy of MavericKids sold from the Daily Maverick shop, we will donate a copy to Gift of the Givers for learners in need. If you don't have a child in your life, you can donate both copies.

Small effort, big impact.