The ANC govt's irreconcilable crime statistics

Chris de Kock says something is seriously wrong with either the SAPS crime figures or StatsSA's VOCS



Just over two years ago I wrote an article which was published in the Daily Maverick of the 11th of October 2017 under the title “Did sexual offences, hijackings and murder really increase respectively by 117%, 93%, and 9% during 2016/2017?”

In that article I suggested that Statistics South Africa (StatsSA) and the South African Police Service (SAPS) should urgently communicate the different uses of the registered crime statistics of SAPS and the results of the victims of crime surveys that StatsSA produce. I warned too that the type of so-called “crime statistics” that StatsSA has been producing for the past three years can be very damaging to the image of South Africa and can cost this country dearly in terms of investment and tourism.

There was no reaction from any quarters (media, government spokespersons, crime analysts and commentators) to that article. Two years and millions of taxpayer’s rand later, StatsSA continue with its business as usual and just last month published the “Victims of Crime. Governance, Public Safety and Justice Survey (GPSJS) for 2018/2019.

For at least the first two thirds of the document, this report, like the 2016/2017 and the 2017/2018 ones, is presented as a “crime statistics” report and, yet again, there is no clear/explicit warning that the so-called crime statistics of StatsSA, Victims of Crime Survey (VOCS), can never replace the reported national crime statistics of SAPS.

It seems that the compilers of the 2018/2019 report succeeded in presenting the results of the survey as crime statistics, because that is exactly how the media, presented it in the three days after it was released by StatsSA.

The exception was an article by Marianne Merten in the Daily Maverick of 4 October 2019, which questioned the huge gaps between the SAPS and StatsSA figures and at least refers to SAPS crime registration as a possible explanation for these gaps.

But what is even more worrying is that the independent commentators and analysts who were asked by the media to comment did so in a non-critical fashion. So, for example, according to Kailene Pillay in a 4 October article in The Mercury, Gareth Newham of the ISS said that the SAPS needed to make greater use of the StatsSA VOCS report and not only depend on their own statistics. He also added that until crime statistics stop being used as a performance indicator of the police there will be a decrease of recorded incidents. If he actually said this, then he:

a) basically, endorsed the validity of the VOCS and elevated its survey status to those of crime statistics, and

b) he questioned the validity of the official reported crime statistics of SAPS.

The SAPS crime statistics of 2018/2019 – as posted on the SAPS website - were certified in 2018/2019 for the first time by the Statistician General, as complying to SASQAF (South African Statistical Quality Assessment Framework) and are now the official crime statistics of South Africa. 

Analysis of StatsSA crime estimates

Column 1 of the table below depicts the StatsSA VOCS 2018/19 estimates for certain categories of crime. Column 2 depicts the estimated reporting rate according to the survey. Column 3 contains the crime figures recorded by the SAPS for 2018/19. Column 4 adjusts these figures to take into account the under-reporting estimates of Column 2. Column 5 is the gap between the StatsSA VOCS estimates and the SAPS figures adjusted for under-reporting.



Column 1

Column 2

Column 3

Column 4

Column 5

Crime category

StatsSA VOCS estimates

Reporting rate according to VOCS

SAPS crime figures

SAPS crime figures adjusted for VOCS estimates of under-reporting

Gap between Column 1 and Column 4

Murder + Culpable homicide.

12 079


32 622

32 622

-20 543

Sexual offences.

26 460


52 420

59 568

-33 108


497 093


332 991

665 982

-168 889

House robbery.

264 054


22 431

37 447

226 607

Street/public robbery.

581 438


131 980

379 253

202 185

Vehicle hijacking.

32 465


17 208

20 245

12 220

Residential burglary.

1 345 196


220 865

458 226

886 970

Theft of motor vehicle.

82 867


48 324

55 995

26 872

The table reveals the following:

Social fabric crime reporting

Social/social fabric crimes against people, like murder, culpable homicide and sexual offences are significantly underreported in the VOCS. The fact that the StatsSA estimate for murder is only 12 079 while the SAPS figure for murder is 21 022 - and 32 622 if culpable homicides are added - raises serious questions about the validity of the VOCS.

Universally murder is seen as the one crime trend where there is no underreporting since, to put it bluntly, there is almost always a dead body. The VOCS indicate that 100% of the respondents/households’ report murder so the SAPS figure of 21 022 / 32 622 can be accepted as correct.

The sexual offence estimate of StatsSA is also less than half of the reported figures of SAPS. StatsSA also found that 88% of the victims of sexual offences report it to the police. This is also not the first time that the StatsSA VOCS found that the SAPS reported figure is basically twice as high as their own surveys.

This analyst can’t remember when in the past 45 years of analysis he ever saw a victim survey where the reporting figure for sexual offences is so high. The estimates over the years fluctuated between 5% and 50% reporting for sexual offences.

332 991 assaults (common and aggravated) were reported to the SAPS in 2018/2019 and StatsSA put their estimate at 497 093. If one adjusts the SAPS figures for under-reporting (using VOCS estimates) then there should have been 665 982 cases reported to the SAPS. This is 168 889 incidents more than the StatsSA VOCS estimate.

It is possible that the respondents in the StatsSA survey had less confidence in the fieldworkers to report intimate domestic disputes, sexual assault and rape within the family and friends circle than when reporting it to the police.

Other crimes

For the other five categories of crime included in the table the StatsSA estimates are between 886 970 incidents (for residential burglary) and 12 220 incidents (for vehicle hijacking) higher when compared to the adjusted SAPS estimates.

So, for example SAPS had 22 431 house robberies reported to them while StatsSA calculated on the basis of the VOCS that there were 264 054. 220 865 residential burglaries were reported to the SAPS while a whopping 1 345 196 were identified by StatsSA.

If the reported SAPS crime statistics are adjusted to account for the StatsSA VOCS estimates of under-reporting, there should have been 37 447 house robberies and 458 226 housebreakings. But even then, there remain gaps of 226 607 cases and 886 970 cases for house robbery and housebreaking respectively. For street/public robbery the gap was 202 185 cases, while it was 12 220 cases for vehicle hijacking and 26 872 cases for theft of motor vehicle.

These gaps in column 5 between the StatsSA estimates (column 1) and the SAPS figures adapted to compensate for under reporting (column 4) need to be explained by other factors than underreporting as underreporting was already accounted for.

Some possible explanations, conclusions and recommendations.

The StatsSA, VOCS report of 2018/2019, as well as the previous two, were produced in the same format and as an alternative crime database to the SAPS reported crime statistics. These are on the internet and thus internationally available. It will in all probability have a negative influence on rating, investment, tourism perceptions and decisions. They will also have a negative influence on the people of South Africa.

If the South African media, analysts, commentators and even the largest and most active crime and policing thinktank in the country, accept the validity and reliability of the StatsSA at face value and recommend that the SAPS use these “more correct” figures, then why should international risk analysts and managers not use it?

Hopefully someone in a leadership role will read this article and at least ask StatsSA, SAPS and/or the relevant Ministers and Parliamentary oversight bodies to do the following:

- StatsSA should explain why for at least the last three VOCS tried to generate crime statistics by calculating from a sample to the population. Why do they not just determine the extent of underreporting of each crime and then use the SAPS crime statistics to calculate the estimated number of crimes (like in column 4).

In all the VOCS before the last three they always warned the reader/user not to see the survey results as an alternative to the SAPS reported crime statistics.

I analysed surveys at the HSRC (Human Sciences Research Council) for nearly twenty years before joining the SAPS in 1995 and then headed and analysed police recorded crime statistics until then end of my official career in 2013. Numerous reasons come to mind as to why national crime figures can’t be calculated on the basis of the victims of crime surveys.

Among other things, there can be confusion on the reporting period and the category of crime, lie factors, differential experience of different categories of crime, the cultural, gender and age characteristics of fieldworkers and many more. StatsSA must explain why there are these huge gaps between their “crime statistics” and the SAPS statistics, even after accounting for under-reporting. According to Merten (2019) the Statistician General, Risenga Maluleke indicated during the release that StatsSA just produce statistics: “The numbers are there. The public must digest them.”

The Statistician General should remember that StatsSA is using millions of rand to do the VOCS every year and many more millions, probably billions, of rand of potential tourism and investment is probably lost as a result of the horrendously higher “figures” of StatsSA.

It is neither acceptable nor responsible to produce confusing statistics without accounting for the following:

- If StatsSA can show that all the variables which could inflate their crime estimates were controlled and that the figures which they produced are at least 95% correct, then very serious questions should indeed be asked about the reported crime statistics of SAPS.

- If StatsSA are correct, then the gaps between the SAPS figures in column 4 and StatsSA figures in column 1 of the table can only be explained by the SAPS manipulating the registration of crime at station level through what Merten, with the support of Eldred De Klerk, refer to as:

a) convincing the victim not to lay a charge because a “crime did not actually occur” or if the victim did not have any particulars of the perpetrator, “then nothing can be done”,

b) “register” the crime in the Incident Book/Register and even supply this number to the victim who will then think it was actually reported and registered. It is even possible that one CAS number can be provided to various victims but it will only reflect one statistic on the CAS. If this is the case it is actually not only a matter of the manipulation of crime statistics, but defeating the ends of justice and fraud.

If the gap in column 5 is mainly as a result of the non-registration of crime by SAPS, the Statistician General should preferably explain why the SAPS 2018/2019 crime statistics were declared official crime statistics, while the most central criteria, (integrity and accuracy) of the SASQAF (South African Statistical Quality Assessment Framework), were clearly not complied with.

Dr Chris de Kock. Independent analyst: crime, violence and crowd behaviour.