Questioning Bheki Cele's lockdown crime stats

Chris de Kock critiques the figures in the three ministerial releases so far

Minister Bheki Cele some of your brag lockdwn crime statistics are incorrect

On Friday afternoon 22 May 2020 the Minister of Police, Bheki Cele, released crime figures for the first 54 days of lockdown compared to a similar period in 2019. This was the third release of crime statistics by the Minister outside of the usual reporting cycle – the previous ones being on the 5th April and 22 April.

I was contacted by the SABC news programme On Point which asked me to listen to the release so that they could interview me afterwards. The Minister only referred to the number of decreases in cases per certain crime categories, but I could do a face value check.

When I compared the figures released on 22nd May with those released on the 5th April for the first week of lockdown, there was clearly something which did not tally.

During my interview I referred in passing to the fact that there was something wrong with some of the figures for trio crimes and that the GBV (Gender Based Violence) figures were confusing.

This article is a full explanation of what I observed.

If one compares the three sets of figures released by the Minister for trio crime figures, on the three different dates, then it is clear that the figures released on the 5th April were in error.

Table 1. Trio-crime figures released for first week of lockdown period (5th of April) first four weeks of lockdown (22 April) and the first eight weeks of lockdown (22 May):


Similar period in 2019, cases

Lockdown period, cases

Case difference

Percentage decreases

Trio crimes for
the first week
of lockdown.

8 853

2 098

-6 755


Trio crimes for
the first 4
weeks of

3 928

1 347

-2 581


Trio crimes for
first 8 weeks
of lockdown.

8 649

3 501

-5 148


As Table 3 shows the trio crime figure of 2 098 cases for the first week of lockdown, compared to 8 853 cases for the same period the year before, are clearly not compatible with the subsequent two releases, which provide lower numbers for both over a much longer period.

The figure of 2 098 cases of trio crime during the first week of the lockdown is, for example, 751 cases greater than the figure of 1 347 cases for the first four weeks of the lockdown.

Now this is clearly incorrect, as is the 8 853 figure. There were in reality only 58 448 trio crimes recorded in 2018/19, according to the SAPS official statistics, an average of just over 1 100 per week.

The comparative trio crime figures for the first week of the lockdown should be publicly rectified by SAPS with immediate effect. The question though is whether this mistake was deliberate or unintentional? Was it manipulated to support a political point or did it slip in accidentally? Even if the latter, why did the Minister not correct the error in his subsequent releases?

The confusion around Gender Based Violence (GBV).

A few days before the lockdown crime statistics for the first week were released on 5 April, Minister Cele referred to 87 000 calls to the GBV hotline. Then with the 5 April release, he admitted that this figure actually referred to all calls in 2019.

He then released a figure of 15 000 calls for the first quarter of 2020, of which approximately 2 300 occurred during the first week of lockdown with only 184 perpetrators charged. Now according to those figures there was in all probability a significant increase in hotline calls in the first week of the lockdown.

Then 17 days later (on the 22nd April) and again a month after that (22nd May) the Minister released the “GBV figures”. See Tables 2 and 3. These figures clearly don’t refer to calls anymore, but according to the Minister to “domestic violence” cases.

Table 2: Cases of “domestic violence” in first four weeks of lockdown:

29 March to
22 April 2019

27 March to 20 April 2020


% change


3 061

-6 929



Table 3: Cases of “domestic violence” in first eight weeks of lockdown:

29 March to
21 May 2019

27 March to 19 May 2020


% change


6 650

-14 383


In his 22 May statement Minister Cele said: “Actually, when we refer to gender based violence, these would cover ALL crimes against women and children as well as the LGBTQI community.… Therefore, for purposes of measuring the impact of the lockdown on homes and families, we use cases of domestic violence.”

Now to use domestic violence as an indicator of GBV is not a very good decision for following reasons:

a) There is a very significant difference between Gender Based Violence (GBV), violence against women and children and domestic violence. Violence against women and children in essence means all violent/contact crime against female victims 18 years and above and children (below the age of 18) of both genders. There doesn’t have to be any relationship between victim and perpetrator. This is a classification of crime purely based on the attributes of gender and age of victims of each crime.

Domestic violence on the other hand includes all violent crimes against members of a household committed by another member of the same household. So, it can include crimes committed by children (of both sexes) against parents of both sexes and crimes committed by parents against children.

It will include crimes committed by children against each other as well by intimate partners against each other (e.g. husband/wife, boyfriend/girlfriend, same sex partners etc.) This can actually not be used as an indicator for GBV - the essence of which is physical, verbal and emotional violence committed by a perpetrator (usually male but it can also be female) against a victim of the opposite sex as a result of a relationship where there is unequal power between the victim and perpetrator.

GBV doesn’t only occur in households. A very good example of GBV is where a male in the position of power sexually and or verbally abuses his secretary. Proper explanatory analysis can only be done on pure categories or subcategories.

So, for example, the lockdown may have caused the conflict between different generations - which may be the larger subcategory in domestic violence - to almost cease, but the abuse of intimate partners to escalate.

Because of the relatively smaller proportion of the GBV subcategory of domestic violence cases, and the relative larger proportion of intergenerational violence, a significant decrease in the latter may produce a significant decrease in the total category of domestic violence even if GBV is on the increase. For this reason, domestic violence is not a good indicator of GBV.

b) In the 18 years that I was in charge of the compilation and analysis of crime statistics in SAPS - as the Head of the Crime Information Analysis Centre - I never trusted, never used and never published domestic violence statistics.

At that stage there were no way of checking how these numbers were generated. The person (not necessarily a police officer) who registered any violence related case would have to make the decision that this was a domestic case and then indicate that by ticking a little box. If this was not done the case would not count as a domestic violence case.

The domestic violence figures in Tables 2 and 3 and the section on domestic violence in the report, South African Police Service.2019. Annual Crime Report 2018/2019. Addendum to the SAPS Annual Report. (p.29), don’t corroborate each other. Personally, I would have used violent crime against woman and children as a closer indicator of GBV.

Selective choice of crime categories.

The question should be asked why the SAPS (Crime Registrar) - hopefully not the Minister of Police (because that would be totally out of line with SASQAF-South African Statistical Quality Assesment Framework ) - selected the specific crime categories released on the 5th of April, 22nd April (Table 4) and 22nd May (Table 5).

Table 4: Crimes in first four weeks of lockdown vs same period the previous year


29 March to 22 April 2019

27 March to 20 April 2020 (lockdown)




1 542


-1 110



2 908


-2 537


Att. Murder

1 300




Assault with intent
to inflict GBH

11 876

1 758

-10 118


Robbery: Aggravated.

6 654

2 022

-4 632





1 146




Robbery - business

1 345




Robbery - residential

1 437





3 928

1 347



***                                                           ***

Table 5: Crimes in first eight weeks of lockdown vs same period the previous year


29 March to 22 May 2019

27 March to 20 May 2020 (lockdown)




       2 970

       1 072

  -1 898



       5 350


  -4 431


Att. Murder

       2 571

       1 132

  -1 439


Assault with intent
to inflict GBV


    22 144

       4 348

-17 796


Robbery: Aggravated

    14 822

       5 397

  -9 425





    2 641


  -1 868


Robbery - business

     2 987

       1 189

  -1 798


Robbery - residential

       3 021

       1 539

  -1 482



       8 649

       3 501

  -5 148


As an analyst for the past 45 years I would have recommended that they should release, at least, the very same 21 crime categories and at least 10 subcategories as in the annual crime statistics release.

The picture provided by the partial release of statistics is an incomplete one. Categories which could provide a better indication of what happened to crime during the different phases of lockdown in the poorer townships, informal settlements, CBD areas and the rural areas were not released. Hopefully, for example, housebreaking (both residential and non-residential) and stock-theft did not increase or stayed stable during the lockdown period and that is why it is not included in the releases.

Chris de Kock
Analyst: crime, violence and crowd behaviour.