Centre for the Study of Violence and Reconciliation

In the Line of Duty?

Shooting incident reports and other indicators
of the use and abuse of force by members of the SAPS

by David Bruce & Gabriel O'Malley

Research Report written for the Independent Complaints Directorate (ICD), October 2001.

David Bruce is a Senior Researcher in the Criminal Justice Programme at the Centre for the Study of Violence and Reconciliation.

Gabriel O'Malley is a former intern at the Centre for the Study of Violence and Reconciliation.

Contents

Acknowledgements

Glossary

Abbreviations

Executive Summary

Section A. Introduction: sources of information on police uses of force

1. Data provided by the South African Police Service (SAPS)
2. Statistics provided by the Independent Complaints Directorate (ICD)
3. Other sources
4. Outline of report

Section B. Evaluation of key data sources

1. Overall analysis of data on shooting incidents
2. Comparison of data from the three provinces

Section C. Fatalities and other aspects of shooting incidents

1. Overall levels of deaths as a result of police action
(a) Accounting for variations in rates of deaths as a result of police action
(b) Evaluating the number of deaths 1 - a proportion of overall homicides

2. A projected national picture of firearm use by the SAPS

3. Incident where people are killed or wounded
(a) Evaluating levels of deaths as a result of police action 2 - lethality index

4. Legality of shooting incidents
(a) Overall Legality of Shooting Incidents
(b) Incidents involving fatalities and woundings

5. Off-duty shootings
(a) Involvement of police reservists in shooting incidents

6. Other general characteristics of shooting incidents
(a) Missed/warning shot incidents
(b) Overall number of shooting incidents
(b) Time of Day

7. Characteristics of police shooters
(a) Number of officers involved in incidents
(b) Age and Experience
(c) Rank
(d) Police officers involved in a large number of shooting incidents

Section D. Indicators regarding other uses of force

1. Complaints submitted to the ICD
2. Criminal and disciplinary cases against SAPS members
3. Civil claims against the SAPS
(a) Claims
(b) Payments

Section E Recommendations

1. Data collection relating to the use of force
2. Greater emphasis on and control of the off-duty use of firearms by police
3. Police Reservists
4. General strategies to control the use of force
5. Further research

Section F Conclusion

Appendix 1: Comparison of existing data sources on use of force by the SAPS - to be completed
Appendix 2: Distribution by year of shooting incidents recorded on SAPS data in all policing areas
Appendix 3: Profile of 4756 shooting incidents in 15 policing Areas in the Western Cape, Eastern Cape and Free State
Appendix 4: Fatal shooting incidents - variations in proportions of incidents declared legal and illegal by policing area
Appendix 5: Criminal and disciplinary cases against SAPS members
Appendix 6: Civil claims against the SAPS

Notes

Bibliography

Acknowledgements

We would like to express out thanks to:

Note: The original draft of this paper was completed in July 1999. However it subsequently became apparent that there was a problem of duplication in some of the shooting incident data on which the report is based. The data has subsequently been rechecked and obvious duplicates removed for the purposes of this October 2001 version.

Glossary

Provinces, areas, and stations - All police stations are part of a police "area" (formerly district) under an area commissioner. All areas, in turn, fall under the provincial commissioner. Note that while all stations fall within a police area there are also units which are linked to the area which are not attached to a particular station. Similarly, there are usually also units which fall under the province which are not linked to a particular unit.

"Deaths as a result of police action" are referred to in ss53(2)(b) of the SAPS Act, 58 of 1995, in terms of which the Independent Complaints Directorate is required to investigate "all deaths in police custody or as a result of police action". These deaths are in general caused by a use of force by the police but would also possibly include a number of deaths caused, for instance, in vehicle accidents.

References to the "three provinces" are references to the Eastern Cape, Western Cape and Free State, the provinces which maintained a reasonably high standard of recording of shooting incidents in the 1995-1998 period.

Missed/warning shots - a missed shot is a gunshot which is fired at a target and which does not hit the target. A warning shot is a gunshot fired with the intention of discouraging a person from a course of action (such as fleeing) but which is not intended to hit that person. In fact "missed shot incidents" are incidents where all shots fired miss the target. In an incident where a person is killed or wounded there may also be a number of missed shots. Incidents where the target hit is not the original target are not recorded in this report as missed shot incidents.

The terms police officer and member of the SAPS are used in this text interchangeably and the word "officer" is therefore not a reference to rank. In fact, the distinction between "officer" and "non-officer" ranks is only relevant in this report with respect to the fact that the person who is responsible for investigating a shooting incident is intended to be of officer rank (Standing Order 251.15.2) unless prevailing circumstances indicate that this is not practical (Standing Order 251.15.3)

Use of force - a use of force in this report is a physical use of force. Some texts on policing also refer to non physical forms of force (sometimes referred to as psychological force).

Abbreviations

ICD - Independent Complaints Directorate
LTL force - less than lethal force. Refers generally to most non-shooting related uses of force.
SAPS - South African Police Service

Executive Summary

This report was written in 1999 using information available at that time on the use of force by members of the SAPS. After problems were encountered with some of the data the report has now been revised with duplicates (roughly 2% of cases) removed from the data on shooting incidents.

The report uses the following four main data sources:

The data from these sources is examined with a view to understanding what available information can tell us about the use and abuse of force by members of the SAPS.

The major part of the report deals with shooting incident data received from the SAPS covering the 1996 - 1998 period. During this period the SAPS system for recording shooting incidents was only operating with a reasonable level of consistency in the Eastern Cape, Western Cape and Free State and in a handful of policing areas in other provinces. The analysis of shooting incidents is therefore based on the data from these three provinces in which 4651 shooting incidents were recorded over the 1996-1998 period, though some reservations are also expressed about the data from all three provinces.

Two major indicators which have been applied in other countries as a way of evaluating police uses of force and particularly firearm usage are (i) the ratio of killings by the police to the overall number of homicides or murders, and (ii) the 'fatality index' - the ratio between persons killed and persons injured by the police. Using these two indexes it appears that police shootings in South African cannot be compared with those in some South American countries, where summary executions by police have been prevalent, and are closer to those in the US. However in a context of massive levels of societal violence the fact that police use of force is not necessarily high relative to societal violence does not justify a lack of concern. Furthermore it does not imply that force, when it is used is necessarily used appropriately, effectively or justifiably.

Other key aspects of the report based on the shooting incident data include:

Other data sources including complaints submitted to the ICD and SAPS data dealing with criminal and disciplinary cases against SAPS members were also examined, partly in relation to whether they reveal anything about non-shooting related uses of force by the police. While these data sources deal with cases where it is specifically alleged that the police actions amounted to criminal conduct, (or at least breaches of police discipline) some degree of caution needs to be applied in using them as indicators of police brutality as, other than in relation to cases on which convictions have been obtained in court, the cases which are recorded amount to cases where police brutality has been alleged rather than proved. A further difficulty is that non of the above data sources distinguish between acts performed by the police while on duty or otherwise performing police responsibilities, and acts involving members of the police in a non-occupational role.

Finally SAPS information on civil claims against the police over the 1995 to 1998 period were also examined. One of the problems here was a level of inconsistency in terms of the categorisation of claims in the different data pieces provided. Key findings from this data were that:

Summary of Recommendations

Key recommendations regarding systems for data collection are that:

Recommendations regarding the control of the use of firearms by police off-duty are that:

Regarding police reservists the report recommends that:

Recommendations regarding general strategies to control the use of force are that:

Conclusion

The report also makes a number of recommendations regarding further research.

Two major causes of concern, the apparent breakdown in the system for monitoring the use of firearms by SAPS members, and the general issue of off-duty firearm use by SAPS members, are highlighted by this report. The report also indicates that existing data sources do not provide much information on non-firearm related uses of force, which may also be abused.

The report does not serve as a major indictment of on-duty firearm use by SAPS members. The information provided indicated that the level of fatalities "as a result of police action" recorded by the SAPS is not necessarily exceptional in relation to that recorded in many US cities, if one takes into account general levels of societal violence. However in a context of massive levels of societal violence the fact that police use of force is not necessarily high relative to societal violence does not justify a lack of concern. Furthermore it does not imply that force, when it is used is necessarily used appropriately, effectively or justifiably.

Section A Introduction: Sources of information on police uses of force

What indicators are there regarding the use of force by the SAPS, how reliable are they, and in so far as they may be reliable, what picture do they provide to us regarding the use of force by members of the South African Police Service? This report sets out to address these issues by examining data provided by the SAPS and ICD on the use of force by members of the South African Police Service.

1. Data provided by the South African Police Service (SAPS)

The main information, which we received from the SAPS, was in the following forms:

2. Statistics provided by the Independent Complaints Directorate (ICD)

Since it started operating in April 1997 the ICD has been producing statistics which, in so far as they are relevant as indicators of the use of force, may be seen to fall into two categories.

3. Other sources

A variety of other data sources were consulted in so far as the provided useful points of reference or comparison. These included research reports and articles that have been written relating to the use of force by the SAPS and a selection of sources from other countries, which it was thought, might provide useful comparative information.

The South African sources are listed in Appendix 1 and compared in terms of the time periods, which they cover, their apparent reliability and other factors. As is apparent from Appendix 1, there is a significant degree of variation between these key data sources:

It is apparent therefore that these different data sources are not comparable in a strict sense. However they are used in this report in an attempt to "build up a picture" of the patterns and variations in the use of force by SAPS members.

4. Outline of report

Using these data sources the report therefore does the following:

Section B Evaluation of key data sources

1. Overall analysis of data on shooting incidents

This report relies primarily on the data provided by the SAPS relating to shooting incident reports recorded on the SAPS centralised database. The original data file covered the period 1996 to 1998 (inclusive) and contained the following fields:

The accuracy of the data is heavily dependent on whether individual police stations reported shooting incidents involving their staff and whether these were recorded by area or provincial offices on the centralised SAPS PERSAP database.

Table 1 provides the distribution of overall shooting incidents by province and compares it to the total population according to the 1996 census and murder rates in 1997 – the middle year of the three years covered by the shooting incident data.

Table 1: Proportion of total recorded shooting incidents in each province over three years compared to proportion of national population and proportion of total number of murders in 1997

 No of Shooting Incidents Recorded:
1996-1998		 % of Total Recorded Shooting Incidents:
1996-1998		 % of National Population:
Census '96		 Number of Murders Recorded:
1997		 % of Total Recorded Murders 1997
Western Cape 2039 27 10 3129 13
Eastern Cape 1587 21 15 4219 17
KwaZulu-Natal 1272 16 20 6200 25
Free State 1130 15 6 1285 5
Gauteng 741 10 18 5645 23
Northern Cape 321 4 2 538 2
North West 237 3 8 1351 5
Mpumalanga 163 2 7 1240 5
Northern Prov 4 0(0.05) 12 981 4
Other/Unspec 72 1


Total 7566 100(99) 100(98) 24588 100

As is indicated in Table 1, virtually 90% of the data reflects incidents recorded in 5 provinces. However:

As is indicated in Table 2 the number of recorded incidents has declined annually since 1996. This appears to be related to a dramatic decline in recording of data from shooting incidents reports on the SAPS data system.

Table 2. Number of reported incidents, by year
Year Frequency Percent
1996 3076 41%
1997 2291 30%
1998 2199 29%
Total 7566 100%

A full breakdown of the data received, by year, is reflected in Appendix 2. This indicates that

i. The three provinces Western Cape, Eastern Cape, and Free State appear to have provided a higher standard of data than other provinces. This is reflected firstly in the fact that the provinces have provided a comparable level of data in all three years and secondly by the fact that there are not dramatic variations in the data indicating a degree of consistency in recording practise in each of the areas over the time period concerned. Even Umtata in the Eastern Cape, appears to have maintained a higher degree of consistency in recording than most of the areas in other provinces. These three provinces which together hold approximately 32% of South Africa's population have accounted for 63% (4756) of the 7566 shooting incident reports recorded.

ii. With the exception of five other areas in different provinces (Umfolozi and Tugela in KwaZulu-Natal, Gordonia and Diamantveld in the Northern Cape, and Marico in the North West), all policing areas show very little consistency in terms of shooting incidents reports recorded. The Northern Cape is the one other province which has a reasonable claim to being included in the group of provinces with more reliable shooting incident data but has been excluded on the basis that 70% of its data is from one area.1 Most of the areas of KwaZulu-Natal show dramatic fluctuations in terms of number of shooting incidents recorded. In particular, however, in Gauteng, North West and Mpumalanga very few incidents (other than in Marico) were recorded after 1996. The reason for the decline in recorded shootings is therefore primarily that the system for recording shooting incidents from these areas stopped operating with the same degree of effectiveness after 1996. In Northern Province a total of four incidents have been recorded over the entire time period.

The implication therefore appears to be that the differences in recorded shooting incidents between 1996 and 1997 and 1998 is as a result of a collapse in the system for recording shooting incidents on the SAPS database in a significant number of policing areas. Overall then, it can be assumed that the differences in total number of shootings recorded in the different provinces over the three year period are more of a reflection of differences in recording practice than of shooting levels.

Furthermore even in 1996 it appears that the system for recording of shooting incidents left much to be desired. Thus, for instance. it would appear reasonable to ask whether the totals recorded in Gauteng in 1996 reflect the true extent of police involvement in shooting incidents in the province. Furthermore, in Northern Province, the incidents recorded in 1996 don't even reflect a tiny proportion of the total number of shootings in that province.

2. Comparison of data from the three provinces

The data from the three provinces Western Cape, Eastern Cape and Free State therefore appears preferable to that from the other provinces and to be more reliable for purposes of building up an overall picture of shootings by members of the SAPS during the three years in question. However closer inspection of the data from these three provinces also indicated that there a number of cases (105) were the recorded incidents were clearly duplicated. After these duplications had been removed the distribution of data from the three provinces over the 1996 – 1998 period was as reflected in tables 3, 4 and 5 below

Table 3: Distribution of recorded shooting incidents in the Free State
Free State 1996 1997 1998 N
Noord-Vrystaat (Welkom) 35% 40% 25% 418
Oos-Vrystaat (Bethlehem) 30% 40% 30% 289
Suid-Vrystaat (Bloemfontein) 36% 35% 29% 384
All police areas FS (N) 368 419 304 1091
All police areas FS (%) 34% 38% 28% 100%
Table 4: Distribution of recorded shooting incidents in the Eastern Cape
Eastern Cape 1996 1997 1998 N
Drakensberg (Aliwal-Noord) 20% 46% 35% 113
Grahamstad 32% 33% 34% 90
Karoo (Cradock) 38% 33% 29% 111
Oos-Londen 30% 33% 37% 384
Port Elizabeth 38% 31% 31% 388
Queenstown 26% 39% 36% 177
Uitenhage 31% 33% 36% 177
Umtata 16% 29% 55% 124
All police areas EC (N) 471 522 549 1542
All police areas EC (%) 471 522 549 1542
Table 5: Distribution of recorded shooting incidents in the Western Cape
Western Cape 1996 1997 1998 N
Boland (Paarl) 30% 39% 31% 538
Oos-Metropool (Bellville) 34% 32% 34% 765
Suid-Kaap (Oudtshoorn) 26% 38% 36% 138
Wes-Metropool (Kaapstad) 32% 37% 31% 577
All police areas WC (N) 643 719 656 2018
All police areas WC (%) 32% 36% 33% 100%

But questions must still remain about to what extent the data accurately reflect overall shooting levels and other characteristics of shooting incidents. Table 6 compares the proportion of shooting incident reports from each of the three provinces with the proportions of population and proportions of average number of murders.

Table 6: Proportion of total recorded shooting incidents in each province over three years compared to proportion of national population and proportion of total number of murders in 1997

 No of Shooting incidents recorded over three years % of total recorded shooting incidents:
1996-1998		 Population as per 1996 census ('000s) % of total population in three provinces Average Annual no of murders recorded:
1996-1998		 % of murder as per total for three provinces
Western Cape 2 018 43 3 957 31 3 129 36
Eastern Cape 1 542 33 6 303 49 4 219 49
Free State 1 091 24 2 634 20 1 285 15
Total 4 651 100 12 894 100 8 633 100

Most significantly, while the Western Cape and Free State both have a higher proportion of the reports relative to proportion of population, the Eastern Cape has a far lower proportion of the total number of reports (33%) relative to its proportion of the population (49%). This might mean for instance, that the Eastern Cape is a less violent place than are the other two provinces. However, the statistic on proportion of murders indicates that the Eastern Cape has a proportion of murders similar to its proportion of the population. In so far as level of violent crime is a predictor of police shootings it may therefore be the case that the Eastern Cape has a higher level of under-reporting than do other provinces.

What types of shootings are least likely to be reported? One possibility is that shootings where no-person is killed or wounded (missed/warning shot incidents) may not be recorded in shooting incident reports to the extent that shootings where a person is killed or wounded are recorded.2 Overall the proportion of incidents where no one is killed or injured was 76% of all recorded shooting incidents in the three provinces (see Appendix 3).

Table 7: Proportion of warning/missed shot incidents in each area
Police area Total number of incidents Number of warning / missed shots % of warning / missed shots
Noord-Vrystaat (Welkom) 418 301 72%
Oos-Vrystaat (Bethlehem) 289 219 76%
Suid-Vrystaat (Bloemfontein) 384 263 69%
All areas in Free State 1091 783 72%




Drakensberg (Aliwal-Noord) 113 82 73%
Grahamstad 90 63 70%
Karoo (Cradock) 111 94 85%
Oos-Londen 384 274 71%
Port Elizabeth 388 315 81%
Queenstown 155 82 53%
Uitenhage 177 149 84%
Umtata 124 27 22%
All areas in Eastern Cape 1542 1086 70%




Boland (Paarl) 538 453 84%
Oos-Metropool (Bellville) 765 638 83%
Suid-Kaap (Oudtshoorn) 138 110 80%
Wes-Metropool (Kaapstad) 577 452 78%
All areas in Western Cape 2018 1653 82%

As can be seen from Table 7 the area in the Western Cape with the lowest proportion of missed/warning shot incidents is the Western Metropole with 78%. By comparison none of the areas in the Free State have a proportion of missed/warning shots higher than 76%. The greatest fluctuations are in the Eastern Cape where two areas in particular (Queenstown and Umtata) have a particularly low proportion of missed shots. As a result of the influence of these areas on the overall provincial statistics the Eastern Cape records an overall lower proportion of missed/warning shots (70%) than does the Free State (72%). The Western Cape statistics are significantly higher than both of the others. It therefore appears reasonable to infer that, unless police shooters in the Umtata and Queenstown policing areas and in the Eastern Cape and Free State generally are significantly more accurate than those in the Western Cape, or that warning shots are used in a higher proportion of incidents in the Western Cape, the reasons for the differences in proportions of missed/warning shots are to do with the likelihood that police in the Western Cape record a higher proportion of their missed/warning shot incidents than do police in the other two provinces. The inference is that the data from the Western Cape is more reliable than the data from the other two provinces in presenting an overall picture of police shootings. Furthermore, the data from the Eastern Cape may be particularly unreliable due to the influence of recording practice in the Queenstown and Umtata areas. Should we therefore exclude the data from the Eastern Cape and possibly also that from the Free State?

A further point of comparison between the two data sets is ICD data on deaths as a result of police action. As indicated, the ICD has been involved in recording deaths as a result of police action since April 1997. The only full year in relation to which the ICD's data and the data on shooting incident overlaps is therefore 1998. To what extent then is there correspondence between the two sets of data? Table 8 compares the total number of deaths recorded in the shooting incident reports with those recorded by the ICD in each province in 1998.

Table 8: Recorded deaths as a result of police shootings in 1998

 ICD – Shooting related deaths as a result of police action SAPS – fatalities in shooting incidents Deaths recorded by SAPS as a percentage of those recorded by ICD
Gauteng 117 1 1%
Northern Province 20 0 0%
North West 13 6 46%
Mpumalanga 32 0 0%
KwaZulu-Natal 110 60 55%
Free State 28 29 104%
Eastern Cape 50 48 96%
Western Cape 39 36 92%
Northern Cape 5 7 140%
Total 414 187 45%

The data reflected in Table 8 appears to give further credibility to the supposition that the data provided from the Free State, Western Cape, and Eastern Cape is reasonably reliable, particularly with respect to shooting incidents which resulted in fatalities.3 Furthermore Table 8 also indicates that the data from the Free State and Eastern Cape may be slightly more reliable in relation to the number of fatal incidents recorded than is the data from the Western Cape. Thus, despite the reservations earlier noted with respect to the number of missed/warning shot incidents recorded in these two provinces, it would appear to be reasonable to include them in this analysis as the data reflects a reasonable degree of reliability.

It is important not to overstate the significance of the correspondence between the data in relation to these three provinces. However it does suggest at least that fatal shooting incidents involving the police that are reported to the ICD are also likely to be recorded in shooting incident reports in these three provinces.4

In summary then:

Despite some reservations then, in what follows we use the data from these three provinces in trying to put together a picture of consistencies and variations in patterns of shootings by members of the SAPS.

Section C Fatalities and other aspects of shooting incidents

As stated, the primary objective of this report is to attempt to clarify what indicators exist relating to the use of force by members of the SAPS and, to identify what information is provided by these indicators with regard to levels and patterns in the use of force by the police. In this section we focus on shootings by the SAPS – the primary cause of deaths as a result of police action.

1. Overall levels of deaths as a result of police action

Table 9 shows the total number of deaths as a result of police action recorded by the ICD in each province. The table also provides the figures for provincial populations from the 1996 national census. In the last column the rate per 100 000 relative to the population of each province is provided. There is therefore a significant degree of variation with provinces like KwaZulu-Natal, Gauteng, the Free State, Western Cape and Mpumalanga recording noticeably higher rates of deaths than the other provinces.

Table 9: Deaths as a result of police action over two years relative to provincial populations

 April '97 -
March '98		 April '98 –
Mar '99		 Total: April '97 – Mar '99 % of national total number of deaths Provincial Population ('000s) % of national total population Rate of deaths as a result of police action per 100 000
Gauteng 135 132 267 25 7 348 18 3.6
N Prov 25 28 53 5 4 928 12 1.1
North W 17 25 42 4 3 354 8 1.3
Mpum 36 37 73 7 2 800 7 2.6
KwaNatal 165 148 313 30 8 417 21 3.7
Free State 36 34 70 7 2 633 6 2.7
E Cape 47 66 113 11 6 302 16 1.8
W Cape 40 62 102 10 3 956 10 2.6
N Cape 5 13 18 2 840 2 2.1
Total 506 545 1 051 100 40 583 100 2.6
(a) Accounting for variations in rates of deaths as a result of police action

The rate of fatalities is potentially influenced by the nature of policies and laws, which restrict firearm use by the police and the degree of seriousness with which such policies are implemented by managers within police departments. An appropriate legislative and regulatory framework is therefore important to reducing the overall level of deaths as a result of police action in South Africa. At the same time, there appear to be other factors which apparently bear some relationship to the rate at which such deaths may occur. In addition to population size, and the crime rate (particularly of types of crime which might justify the use of lethal force by the police in effecting arrests), studies have attempted to identify other variables which may also be useful as "predictors" of variations in the rates of the use of force by the police. The focus of such studies has often been on firearm usage by the police and specifically killings by police officers. Other factors which might in general be expected to correlate with or impact on the number of such deaths include:

In addition to the type of factors listed here, Geller and Scott (1991) point to studies in the USA which have attempted to identify whether other combination of factors "related to levels of violence in communities explain at least some of the variation in police killings of civilians". Factors which have been analysed in this regard include income inequality, the level of food stamp and welfare receipt, measures of social cohesion (such as divorce, unemployment and suicide rates), and levels of gun density within communities. (1991, pp. 450-451) Finally other studies have suggested that political factors are also likely to have an impact on the level at which such killings occur (e.g. Chevigny, 1995; Jacobs and O'Brien, 1998).

(b) Evaluating the number of deaths 1 – a proportion of overall homicides

One important element in evaluating the use of police force is the proportion of the overall number of intentional homicides (i.e. both murders and justifiable homicides) committed by the police. This indicator depends on proportions, rather than actual levels of violence, and therefore reflects the level of police violence in relation to the level of violence in the societal context in which they live and work. Therefore, it can be useful when making international or domestic comparisons. However, when making such comparisons there is a need to observe a certain degree of caution particularly as practices for recording crime differ from country to country. Thus for instance:

In Table 10 therefore the proportion of shooting homicides by the police is compared to overall homicide rates in the three provinces. As can be seen, the proportion of shooting homicides by the police in the Western Cape as compared to the overall number of homicides rose incrementally each year between 1996 and 1998; jumping from .797% of the overall homicides in 1996, to 1.01% in 1998. The Free State saw a higher jump between 1996 and 1997 (from 2.17% to 4.513%), but a precipitous fall from 1997 to 1998 (4.513% down to 2.49%). The Eastern Cape witnessed a slight increase between 1996 and 1997 (1.69% to 1.28%), and an almost imperceptible decrease between 1997 and 1998 (1.28% down to 1.27%). The overall numbers for the three provinces show a decrease in all three years, with a high of 1.69% in 1996 to a low of 1.33% in 1998.

Table 10: Shooting homicides by police as a % of the total number of murders
Province 1996-Total Homicides 1996-Shooting homicides by police (% of total) 1997-Total Homicides 1997-Killings by police (% of total) 1998-Total Homicides 1998-Killings by police (% of total)
Western Cape 3 259 26
(.797%)		 3 129 29
(0.927)		 3 534 36
(1.02%)
Free State 1 339 28
(2.091%)		 1 285 54
(4.202%)		 1 162 29
(2.49%)
Eastern Cape 4 212 41
(1.14%)		 4 219 57
(1.351%)		 3 769 48
(1.27%)
Total 8 810 103
(1.69%)		 8 633 142
(1.64%)		 8 465 113
(1.33%)

To help understand the relevance of the proportions in the three provinces, it may be useful to compare them with those seen in some American cities with relatively large numbers of killings by police, and to the Brazilian city Rio de Janeiro. In making these comparisons however, it should be noted that:

In 1991, police in Philadelphia, Los Angeles, and Houston were responsible for less than 4% and police in Chicago and New York were responsible for less than 2% of the total number of intentional homicides. (Cano, 1997, p.33) By contrast, in 1995, in Rio de Janeiro, police shootings accounted for the death of 358 opponents, 9.3% of those killed intentionally in Rio de Janeiro that year (Cano, 1997, p.33).

The international statistics therefore suggest then that, relative to the overall murder rate the level of deaths as a result of police action in the three provinces is not necessarily particularly high. In particular:

We will therefore assume that existing statistics provide at least a rough approximation of the overall number of people killed by the police. Presuming this to be the case the above data may be seen to provide one indicator of whether or not the SAPS are responsible for a disproportionately large number of deaths. If the overall number of deaths as a result of police action recorded by the ICD during the two year period April 1997 to March 1999 (1051 deaths) is compared to the total number of murders recorded in official statistics over the 1997 and 1998 calendar years (49 463) this provides a statistic of 2.1%. If this and the statistics from the three provinces can in fact be used as a meaningful form of comparison then they indicate that the current overall rates of deaths as a result of police action are not higher than those registered in major American cities and certainly do not indicate a level of police violence comparable with that recorded in cities such as Rio de Janeiro and Sao Paulo in Brazil.7

However there are a number of cautions that should be borne in mind in relation to these statistics. Firstly these province wide, and nation-wide statistics are likely to conceal significant variations in levels of shooting between different localities. Thus, measured relative to murder rates or on any other basis, it is likely that there are significant variations between different areas in terms of numbers of people killed. Secondly, in relation to other variables, and particularly the number of police officers, the number of people killed by members of the SAPS is substantially higher than in a country such as the US. Thus during the April 1997 to March 1999 period when, according to ICD statistics there were 1063 deaths as a result of police action there were on average about 132 500 SAPS members. Calculating the number of deaths per 1000 SAPS members yields a statistic of 8 deaths as a result of police action per 1000 members. By contrast if the average is calculated from figures quoted by Geller and Scott for 11 policing jurisdictions during 1990 these provide a figure of an average of 3 deaths per 1000 members with these varying between 1 per 1000 in Chicago and 7 per 1000 in San Diego, a city which at the time was recording exceptionally high rates of killings by police in comparison with other US cities (figures provided in Geller and Scott, 1992, p. 120).

Comparing South African statistics on people killed by the police relative to the overall rate of murder with statistics from other countries does not necessarily indicate that the SAPS accounts for a disproportionately high number of deaths. But this should not be taken to imply that there is not value in measures being introduced to reduce the overall level of these deaths. Thus, although comparisons with some major US cities are not unfavourable to the SAPS, the overall number of people killed by police in the US is smaller than the total number killed by the SAPS although the population is roughly seven times larger than that of South Africa.8 In 1990 for instance, the US police killed fewer than four hundred people, a number lower than the current yearly rate for South Africa (Geller and Scott 1992 cited in Cano, 1997, p.32). However, it appears that the US figures for 1990 reflect a substantial decline in the overall number of fatalities as a result of police action in the US. Thus in the US it has proved possible to bring about substantial reductions in the number of people killed by the police. This was achieved without reducing the overall effectiveness of the police in dealing with the problem of crime. There is no apparent reason why an effort cannot be made to bring about such substantial reductions in South Africa.

2. A projected national picture of firearm use by the SAPS

As indicated, not all of the recorded deaths as a result of police action are the result of firearm use by the police.9 Table 11 separates out non-firearm related and firearm related deaths using the ICD statistics for the two year period April 1997 – March 1999. A projected average number of shooting incidents per year over the 1996-1998 period is seen in the 2nd last column of table 11.10 The projected provincial average number of shooting incidents per province was then added to get a projected annual average number of shooting incidents nationally for the three years of 6225.

Table 11: Shooting related deaths as a result of police action recorded by the ICD used as a basis for developing a projected annual average number of shooting incidents.

 24 months: non-shooting related deaths 24 months: total deaths as a result of police action 24 months: shooting deaths only % shooting deaths in province Average annual number of shooting deaths as per ICD data Projected average annual number of shooting incidents: 1996-1998 Annual Average Recorded in SAPS data: 1996-1998
Gauteng 22 267 245 26 122 1 627 247
N Prov 7 53 46 5 23 307 1
North W 8 42 34 4 17 227 54
Mpum 5 73 68 7 34 453 79
KwaNatal 37 313 276 29 138 1 840 424
Free State 9 70 61 7 30 400 377
E Cape 10 113 103 11 51 680 529
W Cape 12 102 90 10 45 600 680
N Cape 4 18 14 1 7 93 107

114 1 051 937 100 467 6 225 2 496

In evaluating this table it should be noted that:

It would therefore appear reasonable to assume that the projected average of 6225 shootings a year is in some ways a conservative figure – to what degree, however, must at this point remain a matter for speculation.

Using the projected figure of 6225 shootings as an annual average Table 12 therefore uses a more detailed breakdown of the statistics from the three provinces during the years 1996 – 1998 (see Appendix 3) to provide a projected picture of overall annual national shootings.

The right hand column of table 12 is therefore an attempt to provide a rough overall picture of the shooting incidents in which police were involved during an average year based on ICD data relating to fatal police shootings for the period April 1997 to March 1999 and SAPS shooting incident reports for the three year period 1996- 1998. Note that:

Table 12: Projected total national shootings and their consequences during an average year based on 1996-1998 data from the three provinces
Outcome of shooting incident Data from three provinces: three years % - relative to total incidents Projected national total: three years Projected national total: average year
No of people killed 348 7.5 1 401 467
No of people wounded 982 21 3 921 1 307
No of people arrested 3 843 83 15 500 5 167
No of incidents where person killed 301 6.5 1 215 405
No of incidents where person wounded 859 18 3 364 1 121
No of incidents where person arrested 2 179 47 8 778 2 926
No of incidents where no-one killed or wounded 3 522 76 14 193 4 731
No of incidents where no-one arrested, killed or wounded 2 019 43 8 030 2 677
No of incidents where person hit was not original target 127 3 561 187
Total number of incidents 4 651 100 18 675 6 225
3. Incident where people are killed or wounded

As indicated in Table 12 the proportion of incidents where a person is killed represents 6.5% of the total number of incidents and the proportion of incidents where a person is wounded represents 18% of the total. Table 13 gives a rough indication of the number of people killed or wounded per incident indicating that in 24% of incidents recorded in the three provinces someone was killed or wounded. Of these incidents 21% were incidents were a single person was killed or wounded while in 3% of incidents (135) more than one person was wounded or killed. Of these 135 incidents, two people were wounded or killed in 102 (2%), three people were wounded or killed in 18. Of the remaining 15 incidents, one incident is particularly notable in that an exceptionally large number of people (12) were wounded or killed.

These figures therefore indicate why the number of people killed or wounded exceeds the total number of incidents where people where killed or wounded.

Table 13: People killed or wounded per incident
Number of people killed or wounded Count % Cumulative Percent
0 3 522 75.7 75.7
1 994 21.4 97.1
2 102 2.2 99.3
3 18 0.4 99.7
4 9 0.2 99.9
5 3 0.1 99.9
6 1 0 100
9 1 0 100
12 1 0 100
Total 4 651 100 100
(a) Evaluating levels of deaths as a result of police action 2 – lethality index

Another statistic, which has been used in other countries as an indicator of possible police abuses of force, is the comparison between the number of people killed by the police and the number of people they injure when shooting (see Table 14). A high kill to injury ratio would suggest the likelihood of widespread summary executions where police deliberately "take out" suspects whom they have apprehended. This ratio of persons killed to persons injured in a city is termed its lethality index.

Figures quoted by Chevigny indicate that in the United States in the 1970s and 1980s the Chicago Police Department had a .44/1 lethality index, and the New York Police Department had a .33/1 lethality index. By contrast Buenos Aires (1.5/1), Jamaica (1.9/1) (Chevigny quoted in Cano, 1997, p.34), and Rio de Janeiro (2.7/1) all possess alarmingly high indexes (Cano, ibid). These far outstrip those seen in the Western Cape, Eastern Cape, and Free State as reflected in the shooting incident reports received from the police.

Table 14: Ratio of people killed by police shots to people injured
Province # killed by police shots # wounded by police shots Lethality index Free State
Free State 111 243 0.46/1
Eastern Cape 146 388 0.38/1
Western Cape 91 351 0.26/1
Total 348 982 0.35/1

The total lethality index of the three provinces between the years 1996 and 1998 is .35/1. In other words, according to the data contained in the shooting incident reports in the three provinces, roughly 25% of those who were hit by police gunfire died. This proportion resembles that seen in Chicago (31%) and New York (25%) but is far lower than the indices for Buenos Aires (60% of those hit), and Rio de Janeiro (73%). One conclusion to be drawn from this data is that while summary executions may occur they are not practiced systematically by members of the SAPS who are involved in shooting incidents.12

4. Legality of shooting incidents
(a) Overall Legality of Shooting Incidents

Sub-section 15.5.2.8 of Standing Order 251, the Standing Order which provides for the submission of shooting incident reports, provides that the report must state whether the shooting "was legal or not". It might be expected by outsiders that the police officer responsible for producing such reports would consistently "white wash" the police involved in these incidents and exonerate them from any blame. Instead it appears that the police declare a relatively large number of these incidents to have been "not legal". As is apparent from Table 15 this appears particularly to be the case in incidents where people are killed (41%) and where the person hit (whether killed or injured) was not the original target (43%).

Table 15: Proportion of police shootings declared legal and illegal by category of shooting in the three provinces, 1996-1998

 Legal Not legal Total #
Outcome of incident Count % Count %
Fatality 179 59% 122 41% 301
Wounded 669 78% 190 22% 859
Someone wounded or killed 831 74% 298 26% 1 129
More than 1 person killed 13 39% 20 61% 33
Incorrect target hit 72 57% 55 43% 127
Someone injured or arrested 2 189 90% 252 10% 2 441
No-one wounded or killed 3 175 90% 347 10% 3 522
Arrest during incident 2 052 94% 127 6% 2 179
All incidents 4 006 86% 645 14% 4 651
NB Any incident in which at least one shooting act was classified as unjustified has been classified as not legal in this table.

What is difficult is to evaluate the significance of the fact that a relatively high proportion of shooting incident reports, and particularly reports relating to fatal incidents, contain a finding that the shooting was not legal. After the report has been completed, particularly in incidents where a person has been killed or wounded, the report is sent through to the area or provincial headquarters of the SAPS where the report is supposed to be evaluated by members of the SAPS legal services division. Unfortunately the available data does not indicate what kind of process is involved in such an evaluation and how, for instance, members of the legal services division usually respond to the large number of reports where it is indicated that the shooting was unlawful. Therefore, where a shooting incident report indicates that a shooting was not legal, it is not clear how frequently actual disciplinary or legal steps are taken against the police officer involved.

Shooting incident reports are simply the first step in a bureaucratic process of evaluating shooting incidents and do not have the status of a legal finding. There may be at least three possible tendencies which might impact on the way in which the legality of shooting incidents is evaluated in the shooting incident report:

There is of course another possibility. The provision which requires that the report should state whether the shooting was "legal or not" also indicates that "[If doubtful, this should be specifically mentioned]". The police officer responsible for writing the report could reasonably be uncertain about whether the shooting was legal or not. This might be because of a lack of clarity about the legalities or because of insubstantial or conflicting evidence about what actually happened during the incident. It would therefore be reasonable that in a lot of cases shooting incident reports express uncertainty about the legality of incidents. However, there is no record of the uncertainty in the data on shooting incidents which we received which either records shootings as legal or not legal.

One particular issue on which there is little clarity is why such a high proportion of shootings which have fatal consequences are declared to be not legal while a relatively small proportion of shootings where no one is killed or injured are declared to be illegal. While some of these may be warning shot incidents, part of what is implied is that, in many circumstances, a shooting which would have been declared unlawful had the shooter hit the target is declared legal because the shooter missed.13 Some may argue that it is a positive sign that shooting incidents which have more serious consequences are viewed in a more critical light. However, what is far from clear is whether the evidence at hand shows that a consistent (and preferably rigorous) standard is applied to the evaluation of shooting incidents.14

At base level then, we cannot be sure that variations in proportions of shootings declared to be unlawful (whether by shooting type or by other variables such as policing area) reflect different bureaucratic practices, or real differences in the legality and illegality of shootings. In particular, we suspect that if a more consistent standard was applied to the evaluation of all shooting incidents there would be a higher level of consistency between the proportion of shootings in all categories which are declared to be not legal. Whether this would mean that fewer fatal shootings, or more shootings in which there were no deaths or injuries, would be judged to be not legal is impossible for us to ascertain.

Without a more in depth insight into the significance of these judgments on the legality of shooting incidents the best option may be to take these judgments at face value as an indicator of the opinion of the officer charged with investigating the shooting incidents.15

(b) Incidents involving fatalities and woundings

One of the more alarming statistics in Table 15 is the legality of shooting incidents which resulted in a fatality. Only 59% of all shooting fatalities between 1996 and 1998 were declared to have been legal shootings. If the finding in the shooting incident report reflects a serious belief on the part of the officer charged with investigating the incident that the shooting is lawful or unlawful then this suggests that the police have engaged in haphazard, dangerous shooting. Table 16 provides a breakdown of each province individually. A more complete breakdown by area (see Appendix 4) shows even more dramatic variations with the East Metropole having found 12% of fatal incidents (4) to have not been justified while areas in the Eastern Cape found up to 57% (13 fatal incidents in Queenstown) and 78% (7 fatal incidents in Uitenhage) to have been unlawful.

Table 16: Proportion of fatal incidents declared legal and illegal by province

 Justified Not justified Total #
Province Count % Count %
Free State 52 55% 42 45% 94
Eastern Cape 62 50% 63 50% 125
Western Cape 65 79% 17 21% 82
All 179 59% 122 41% 301

As seen in Table 15, a far higher proportion of incidents in which someone was wounded are found to have been legal than is the case in incidents involving a fatality. The average rate at which incidents in which a person was wounded were found to have been legal was 78% for the three provinces. Nevertheless some area statistics do stand out. In the Southern Free State, of a total of 78 incidents where people were wounded over the course of three years, 33 of those incidents (42%) were categorized as illegal by the police. Both Grahamstown (50%) and Karoo (57%) have high rates of illegal wounding too, although each region reported fewer than twenty-one incidents.

The proportion of all shooting incidents in which someone was wounded or killed which were found to have been legal or illegal is seen in Table 17. The three province total of legal shootings that wounded or killed someone stands at 74% of all such shootings. In the Western Cape, 88% of these shootings were legal, while in the Eastern Cape and the Free State the percentage was below 70%.

Table 17: Proportion of incidents where someone was wounded or killed declared legal and illegal by province

 Justified Not justified Total #
Province Count % Count %
Free State 208 68% 100 33% 308
Eastern Cape 301 66% 155 34% 456
Western Cape 322 88% 43 12% 365
All 831 74% 298 26% 1 129

According to the shooting incident reports therefore in more than one quarter (27%) of the incidents where someone is injured or killed and substantially more than a third of incidents where someone is killed (41%) the shootings were judged not to be lawful. Particularly in the Free State and Eastern Cape even higher proportions of shootings where a person is killed or injured are declared to be not legal.

As seen in Table 15 in 42% of the cases where the police hit a person who was not the original target the action was judged to be legal.16 The proportion of these cases which were judged not to be legal therefore resembles the proportion of cases where a person was killed which were judged not to be legal. While cases where the incorrect target was hit do not necessarily point to unlawful police conduct – the shooting could be a reasonable mistake – one would perhaps expect a higher proportion of these shootings to be judged to be illegal. Of all the shooting incidents between 1996 and 1998, one hundred and twenty seven (3%) hit an incorrect target. These represent 11% of incidents were someone was killed or wounded.

These statistics therefore provide cause for serious concern. In a large number of incidents where a person is killed or injured the shooting is not at face value legally justified. If what the numbers say is to be believed, then there is a widespread pattern of unjustifiable gunfire in circumstances where this causes a death or injury.

As we will show however, the data relating to the legality of fatal shootings becomes even more disturbing in relation to shootings involving off-duty police officers.

5. Off-duty shootings

The data in Table 18 is a sobering reflection on the state of off-duty policing.17 Although off-duty police officers participated in 16% of all recorded incidents, they accounted for 23% of all incidents where someone was wounded, 35% of all incidents where the incorrect target was hit, and a staggering 42% of all fatalities. According to the data reflected in Table 15 when they discharged their guns, off-duty police officers were twice as likely as their on-duty counterparts to wound or kill someone, three times as likely to hit the wrong target, and almost four times as likely to kill someone.18

Table 18: Involvement of on and off duty officers in incidents with incorrect target hits, woundings and killings

 All incidents Incorrect target hit Wounding Fatality
Status Count % Count % Count % Count %
On duty 3 888 84 82 65% 659 77% 174 58%
Off duty 763 16 45 35% 200 23% 127 42%
Total 4 651 100 127 100% 859 100% 301 100%

As seen in Table 19, while only involved in 16% of the total number of shooting incidents, off duty police accounted for 42% of incidents in which fatalities where involved. These statistics are particularly pronounced in the Free State where off-duty police officers were involved in 28% of all shootings incidents but accounted for 51% of all fatal incidents and in the Eastern Cape, where off duty police accounted for 18% of shooting incidents but 46% of all fatal incidents in the province. In the Western Cape, although off-duty police officers were involved in only 9% of all shooting incident, they were accountable for 27% of the shooting fatalities.

Table 19: Provincial Breakdown of on/off duty police incidents involving a fatality
All incidents Fatal incidents
Province Status Count % Count % of fatal incidents % of all incidents
Free State On duty 784 72% 46 49% 6%
Off duty 307 28% 48 51% 16%
Eastern Cape On duty 1 271 82% 68 54% 5%
Off duty 271 18% 57 46% 21%
Western Cape On duty 1 833 91% 60 73% 3%
Off duty 185 9% 22 27% 12%
Total On duty 3 888 84% 174 58% 5%
Off duty 763 16% 127 42% 17%
All 4 651 100% 301 100% 7%

What is most disturbing about the data relating to off-duty shooting is the high contribution which these shootings make to the overall level of fatalities with 127 out of the total of 301 fatal incidents having involved off-duty police officers. Of these incidents 57 were in the Eastern Cape, 48 were in the Free State and 22 in the Western Cape.

While 17% of the recorded incidents involving off duty police resulted in a fatality, in the Eastern Cape 21% (one in five) of these incidents were fatal. Potentially, one reason for such a high fatality rate is that a smaller proportion of the non-fatal incidents involving off-duty police officers are recorded in the shooting incident data.

In all likelihood, off-duty police officers discharge their guns more frequently than is seen in the data here. It is however, also possible that the circumstances in which off-duty police officers use their guns might tend to be quite different from the circumstances in which police use their guns while on duty. For instance, if it were the case that off-duty shootings tended to happen at closer range than the average on duty shooting, or warning shots tended to be less of a feature of off duty shootings, then this might also contribute to there being a lower proportion of missed/warning shots recorded for off duty shootings.

Overall, roughly 28% of the time that a police officer from the three provinces was involved in an incident where someone was injured or killed he was off-duty. When the three South African provinces are broken down individually, it becomes clear that there is a real discrepancy in patterns. In the Western Cape in 14% (52 out of 365) of the incidents when someone was shot and killed or injured by a policeman, that policeman was off-duty. This number rises to 30% (135 out of 456 incidents) when in the Eastern Cape, and escalates to 42% (128 out of 308 incidents) in the Free State. In a disturbing 34% of the incidents where more than one person was killed or wounded in an incident (46 out of 135 incidents) the police officer was off-duty.

One point of comparison is a study of on- and off-duty shootings in Michigan state between 1976-1981, where off-duty police officers contributed to 17% of incidents where persons were killed or wounded (Horvath, 1987, p. 231). By comparison with these data it seems that the proportion of off-duty shootings involving a fatality or injury in the three provinces – and potentially throughout South Africa - is relatively high. Several factors might be relevant to understanding the issue:

First, members of the SAPS (like the police in many other countries) are expected to place themselves on duty in certain situations even when they are off-duty. As far as we are aware the exact nature of the obligation of a police officer to "place himself on duty" is not clearly defined anywhere. What exists is a general expectation, backed up by the potential for reprimand or discipline, that police officers will do so particularly where they are faced with a situation of a more serious crime in progress. It is not known, for instance, to what extent training deals with the potential dynamics of a situation which a police officer might encounter while off duty but our impression is that the issue is not covered in police training. While police officers may deal with some such situations relatively effectively this is not necessarily generally the case. As noted by Geller and Scott,

"Police officers traditionally have considered their responsibility for off-duty action to include aggressive intervention (termination of crimes and arrest of suspects) and not just careful observation and notification of on-duty police. The arming of off-duty police, a practice which has been severely criticised by some commentators … facilitates such aggressive action. But police frequently are at a tactical disadvantage while off duty, being out of radio communication with other officers and usually not having the opportunity to plan a course of action as they travel to a suspected crime scene … . [O]ff–duty officers who happen to be socialising in a tavern when armed robbers enter may be in no condition to outdraw the holdup men, yet the presumed imperative to take police action may prompt an ill-advised confrontation.

Geller and Scott go on to note that off-duty police officers who are out of uniform do not possess the same legitimacy to intervene that on-duty officers have. An attempt to interfere in a situation may often be perceived as aggression, and therefore received with more resistance.

A further possible explanation for the high rate of off-duty shootings is that police officers may live in high crime neighborhoods themselves, and therefore be exposed to situations where the use of force appears to be necessary when off-duty. Thus, disproportionate off-duty lethality might also be explained by the mere chance that off-duty police officers in these areas are frequently found in situations which require them to shoot and kill.

Therefore there may be legitimate reasons for off-duty shootings. In fact, the overall proportion of incidents where a person is injured or killed which occur off-duty (28%) is only slightly higher than that in some US cities. According to sources cited by Geller and Scott, studies conducted in a number of major American cities indicate the proportion of civilians shot by off-duty officers has tended to lie somewhere between 17% and 26% (Geller and Scott, 1991, p. 460).

The proportion of persons who are shot by off-duty officers in the US is regarded by some observers as quite large and is