South African Review of Sociology 2005, 36(2) 141 Interpreting the Body Count: South African statistics on lethal police violence David Bruce Senior Researcher, Centre for the Study of Violence and Reconciliation PO Box 30778, Braamfontein, 2017 dbruce@csvr.org.za This paper looks at the incidence of lethal police violence since, and prior to, the transition to democracy in South Africa in the mid 1990s. Independent Complaints Directorate statistics on ‘deaths as a result of police action’ indicate that they have declined to their lowest levels since 1997 in five of South Africa’s provinces, though two provinces also demonstrated dramatic increases in deaths in the last year. The possible impact of new legislation on the use of lethal force for arrest, implemented in South Africa in 2003, on the number of deaths, is briefly considered. Statistics on killings by police from the apartheid period are then examined in relation to the question of whether levels of killings by police have changed since apartheid. It is suggested that there is reason to doubt the reliability of official figures on people killed by police from the apartheid era. While deaths have declined since 1997, and it cannot be said that current levels of deaths are comparable to apartheid era levels, figures, such as those on innocent bystanders killed by police, and on the legality of shootings, provide cause for concern. Keywords: police violence, lethal force, police reform Introduction During the apartheid period, South Africa had a worldwide reputation for police brutality. With the adoption of the Constitution, and the efforts to establish greater accountability and to ‘transform’ the police, many hoped that dramatic changes in police behaviour would be achieved. It therefore came as something of a surprise to those concerned when, following its establishment in April 1997, the Independent Complaints Directorate (ICD)1 started reporting what appeared to be an exceptionally high rate of deaths in police custody and as a result of police action. ICD reports covering its first year of opera1. The Independent Complaints Directorate was initially provided for in Section 222 of South Africa’s ‘interim’ Constitution (Act 200 0f 1993) which provided for ‘an independent mechanism under civilian control’ to be established to ensure that ‘complaints in respect of offences and misconduct allegedly committed’ by police be ‘investigated in an effective and efficient manner’.The legislative provisions establishing the South Africa Independent Complaints Directorate (Chapter 10 of the South African Police Services, 68 of 1995), require that the ICD investigate ‘all deaths in policy custody or as a result of police action’ (Section 53(2)(b)). In addition police are also required to report all of these deaths to the ICD (Section 53(8)). While the ICD is defined by name as a ‘complaints’ body, most deaths are reported to the ICD in terms of section 53(8) rather than being the subject of complaints. 142 South African Review of Sociology 2005, 36(2) tion, from April 1997 to March 1998, indicated that there had been 737 police action and custody deaths. In the following year, ending March 1999, the ICD recorded 756 of these deaths. From the start these figures were a source of confusion. One tendency was for people, including the ICD, to use the abbreviated term ‘deaths in custody’ to refer to all of these deaths. Most of the deaths which were reported though were not deaths in custody. During the period following the establishment of the ICD in 1997 numerous newspaper articles were published purporting to provide figures for ‘deaths in custody’ whilst using figures for all deaths recorded by the ICD. This confusion was reinforced by the tendency to equate these supposed ‘deaths in custody’ with the phenomenon of ‘deaths in detention’ which had been the cause of so much anger during apartheid. But the concern with ‘deaths in detention’ had been a concern with the death of political activists in custody. These were mostly people held in terms of security legislation that authorised the police to detain people without being charged or tried for extended periods of time – in effect indefinitely.2 During apartheid there was hardly any public awareness or information about the deaths of criminal suspects, arrested under the normal arrest provisions of the Criminal Procedure Act. The bulk of deaths recorded by the ICD are not however deaths in custody. Deaths in custody are mostly, if not entirely, deaths from causes other than police violence, including: injuries inflicted by other people, suicides and natural causes. In its statistics the ICD records ‘deaths in custody’ alongside a category of ‘deaths as a result of police action’. ‘Deaths as a result of police action’ are mostly deaths as a result of the use of force or violence by the police, and constitute 65 per cent of the more than 4,600 deaths recorded by the ICD since 1997. It is these 65 per cent of deaths that provide the focus of this paper. The purpose of the paper is to clarify what is known about levels of deaths resulting from the use of force or violence by the police in post-apartheid South Africa. ICD statistics show that in the majority of provinces there has been a decline in the number of these 2. The Human Rights Committee records 73 deaths of people, held in terms of the detention without trial provisions of security legislation, in the period 1963 – 1990. (Coleman, 1998:57) A large number of these deaths occurred in suspicious circumstances though inquest courts generally exonerated the police. However in 7 cases police were found to be criminally responsible for the deaths, 6 of these involving homeland police agencies (Coleman, 1998:6367). During the apartheid period there was not much of an interest in the deaths in custody of people who have been arrested by the police for alleged criminal offences. In so far as there was concern about ‘deaths in custody’ (as opposed to ‘deaths in detention’) this also tended to be a concern about the deaths of political activists who had been charged with offences (including offences under the Terrorism or Internal Security Acts) and therefore were being held in terms of the ordinary provisions of the Criminal Procedure Act rather than in terms of the ‘detention without trial’ provisions of security legislation. The Human Rights Committee also recorded 32 deaths of political activists held in police custody between 1984, when monitoring of these deaths began, and 1990 (Coleman, 1998:67). Fernandez (1991:20) quotes figures from Amnesty International indicating that more than 300 deaths in custody were reported between the beginning of 1980 and the end of 1982 and from the Institute of Race Relations indicating that 105 deaths in police custody were said to have occurred in 1987. Beyond this, little statistical information appears to be available on the number of non-political deaths in custody during the years of apartheid. South African Review of Sociology 2005, 36(2) 143 deaths, since the ICD first started recording these deaths in 1997, though in two provinces there was a dramatic increase in deaths in the 2003-2004 year. The paper also attempts to shed some light on what conclusions may be drawn about the ‘lethality’ of the police in democratic South Africa, as compared with the SAP during apartheid.3 The paper concludes that there is reason to doubt the reliability of information on killings by police from the apartheid period. As a result there is no basis for making comparisons between past and current statistics. But despite the declines in the majority of provinces, ICD statistics themselves, including data indicating that a high number of ‘innocent bystanders’ have been shot by police in recent years, and data on the legality of shootings, provide reason for serious concern about killings by police in current day South Africa. ICD statistics on deaths in police custody or as a result of police action Overall levels and trends in police action and custody deaths since 1997 The SAPS is required in terms of Section 53(8) of the SAPS Act (68 of 1995) to report all police action and custody deaths to the ICD. Despite this statutory obligation, the experience of the ICD has been that these provisions are sometimes ignored.4 Some reports on deaths which the ICD receives originate from sources other than SAPS reports. But while there are indications that under-reporting may occur there is no clear evidence that this has a significant impact. As far as is known therefore ICD statistics provide a relatively reliable record of the total number of police action and custody deaths.5 According to the ICD statistics, reflected in Table 1, during the period April 1997March 2004 – the first 7 years of operation of the ICD – there have been a total of 4,688 police custody and action deaths. 3. 4. 5. Apart from acknowledging the possibility that figures on deaths in custody might include some deaths as a result of police violence, this paper therefore does not examine questions to do with deaths in custody. The issue of deaths in custody is also a subject which needs much greater attention from researchers and authorities partly as these deaths are now at the highest level recorded since the ICD started operating in 1997 (see for instance Dissel & Ngubeni; 1998 & Bhana, 2003). See for instance ICD, 1997/98 Annual report, p.8 and 1998/99 Annual report, pp.25-26 Bruce and O’Malley (1999:10) compared ICD statistics on deaths as a result of police shootings in 1998 with raw police shooting data for the same year. In three provinces in which the police data appeared to be reasonably reliable the discrepancies between number of deaths recorded were as follows: Free State (28 deaths in ICD statistics, 29 in SAPS data); Eastern Cape (50 deaths in ICD statistics, 48 in SAPS data); Western Cape (39 deaths in ICD statistics, 36 in SAPS data). In addition to the Free State more deaths were also recorded in police data in the Northern Cape (7 deaths) than in ICD statistics (5 deaths) for that province. It may be observed therefore that, relative to 122 deaths recorded by the ICD in these four provinces during 1998, ICD statistics appear not to have included at least 3 (2 in the Northern Cape and 1 in the Free State) killings by the SAPS that are reported in SAPS statistics. 144 South African Review of Sociology 2005, 36(2) Table 1 Number of deaths recorded by ICD, April 1997– March 2004a Deaths as a result of police action Deaths in police custody Police action and custody deaths combined 1997/1998 518 219 737 1998/1999 558 198 756 1999/2000 472 209 681 2000/2001 432 255 687 2001/2002 371 214 585 2002/2003 311 217 528 2003/2004 380 334 714 Total: 7 year period 3,042 1,646 4,688 a.The ICD started operating in April 1997 and has generally reported its data over the financial year from the beginning of April to the end of March. This paper will usually simply refer to these years as e.g the year 1997/ 1998. On average over the seven year period representing the middle and latter years of South Africa’s first decade of democracy, the ICD has recorded 670 deaths a year, or 56 a month. Overall of these 56 deaths a month, almost 20 (35 per cent) have been classified by the ICD as ‘in custody’ and 36 (65 per cent) have been deaths classified as the ‘result of police action’.6 During the first and the second year of operation of the ICD the monthly average of these deaths increased slightly from 61 to 63 a month. While there was a steady decline in these numbers over the following four years (to end March 2003) the death toll rose again in the 2003/2004 year with 714 of these deaths being recorded, a dramatic 35 per cent increase on the previous years death toll of 528. The overall trend also conceals much greater variations with recorded deaths as a result of police action having declined fairly substantially overall up to the year ending March 2003 (after increasing initially) and recorded custody deaths having remained in the region of 200 to 250 per year, before rising dramatically to 334, their highest level ever, in the April 2003- March 2004 year (see Figure 1). Uncertainty regarding the exact distinction between police action and police custody deaths In its statistics on deaths the ICD records ‘deaths as a result of police action’ and ‘deaths in custody’ separately. Most deaths ‘as a result of police action’ recorded by the ICD are the result of actions by the police against people who are not in custody, whether the person dies instantly, or at a later point. It is also assumed in this paper that a death which was 6. Note that there are sometimes inconsistencies in the number of deaths reported by the ICD and figures used in this paper are those provided in the ICD annual reports. South African Review of Sociology 2005, 36(2) 145 Figure 1 Trends in deaths in custody and as a result of police action, April 1997 to March 2003 clearly related to an assault, or other violence by the police, where the assault or violence occurred in custody, will also be classified by the ICD as a death ‘as a result of police action’. In this paper it is therefore assumed that deaths recorded by the ICD as ‘deaths as a result of police action’ include all deaths where there is reasonable evidence that these deaths are directly caused by the actions of the police, irrespective of whether the actions of the police, or the deaths themselves, occur in, or out of custody.7 By implication where the ICD records a death as ‘in police custody’ (rather than ‘as a result of police action’) this implies that this death is from causes other than the use of force or violence by the police (police action), or at least that when the death is recorded, there is no clear evidence that the death is directly caused by the police.8 It should be noted that the ICD does not itself have a clear way of distinguishing between these two categories of deaths and this way of distinguishing deaths in custody from deaths as a result of police action may be incorrect.9 Deaths are categorised in each of the ICD regional offices and, in the absence of a clear directive setting out the distinction between ‘custody’ and ‘action’ deaths it is likely that there will be differences in the approach applied by different ICD offices. 7. 8. 9. It is also possible that the death of a person after their release from custody might be the result of injuries or ill-treatment at the hands of the police which occurred prior to or in custody. The term ‘in custody’ is here understood to mean ‘under the control of’ the police. Custody therefore commences as soon as a person submits to arrest, or as soon as the police gain control over the person by forcible means where the person does not willingly submit to arrest. The ICD sub-categories referring to causes of ‘deaths in custody’ include the category ‘injuries in custody’. It is believed that deaths in this category generally exclude known deaths resulting from assaults by the police in custody and mostly relate to deaths resulting from assaults by other prisoners or other non-self-inflicted injuries. 146 South African Review of Sociology 2005, 36(2) In addition some deaths which are recorded by the ICD as deaths in custody may be ‘disguised’ deaths as a result of police action. For instance independent forensic pathologists have come across cases which were ostensibly the result of suicides, but where examination of the body during the autopsy reveals injuries, believed to have been inflicted by the police, which may have been the primary cause of death.10 While it is therefore possible that some deaths as a result of the use of force or violence by the police, are recorded in ICD statistics as ‘deaths in custody’, as far as is known ICD practice is that, where deaths are identified as being the result of the use of force in custody, they are classified as ‘deaths as a result of police action’. Being concerned with ICD statistics on lethal police violence this paper therefore focuses only on ICD statistics on deaths ‘as a result of police action’. Lethal police violence in South Africa, 1997 -2004 Deaths as a result of police action and lethal police violence As reflected in Table 2, ICD statistics on deaths as a result of police action indicate that: • Overall 91 per cent of the 3042 deaths are recorded as having been the result of shootings and 9 per cent of deaths are not recorded as shooting deaths. • During the first four years the ICD recorded 11 per cent of deaths as ‘other’ though it provided no breakdown as to these ‘other’ deaths. • Of 998 shooting deaths recorded during the 3 years April 2001-March 2004, 37 (4 per cent of shooting deaths) were deaths of innocent bystanders. These were in fact recorded by the ICD under the ‘other’ category in the 2001/2002 and 2002/ 2003 years (they have been moved to the ‘shooting’ category in the table above on the basis of the assumption that they are shooting deaths). This suggests that the ‘other’ deaths recorded in the April 1997-March 2001 period might also have included some deaths in this category. 11 • During the 3 years from April 2001-March 2004 the ICD recorded 64 non-shooting deaths. These constituted 6 per cent of all deaths as a result of police action. Of these 45 (70 per cent) were the result of incidents involving police vehicles while 19 (30 per cent) were allegedly the result of physical assaults of one kind or another including ‘beatings’ or ‘torture’. 10. Workshop of the Independent Medico-Legal Unit, Johannesburg, 10-11 September 1999. 11. The ICD also sub-divides shooting deaths into several other sub-categories namely: ‘during arrest’, during crime’, ‘during investigation’, ‘other intentional’, ‘possible negligence’. In some years deaths were also recorded in an additional category of ‘negligent handling of a firearm’ while in 2001/2002 they added the categories ‘during escape’ (16 deaths in that year). However the ICD has never provided a detailed explanation as to what the distinction is between these categories which are not meaningful in terms of the law which primarily distinguishes between shootings for purposes of defence and for purposes of arrest. It is difficult to ascertain, to give just one example, whether the ICD would record the death of a person who is fleeing from the scene of a crime as ‘during crime’ or ‘during arrest’. It is also unlikely that there has been consistency in applying these classifications considering that case analysts and monitors first received training in November 2002 and the ICD does not have formal definitions which distinguish these categories from each other. South African Review of Sociology 2005, 36(2) 147 Table 2 ICD classification of causes of deaths as a result of police action: April 1997 – March 2004 Shootings Other Total 1997/1998 458 60 518 1998/1999 501 57 558 1999/2000 405 67 472 2000/2001 402 30 432 2001/2002 345 26 (includes: 20 – innocent bystander (18 – struck by police vehicle, 7 - beaten with shot from ‘other’ category) hands, fists or other object, 1 - tortured 371 2002/2003 293 (includes: 9 - innocent bystander shot from ‘other’ category ) 18 (9 – struck by police vehicle, 8 – assault/ beating, 1 –toture) 311 2003/2004 360 (includes: 8 – innocent bystander shot) 20 (18 - vehicle accidents, 2 – beaten with hands/fists) 380 Total: 7 years 278 2,764 (includes: 37 – innocent bystander (includes: 45 – struck by police vehicle, 17 – shot April ‘01 – March ‘03) assaults/beatings, 2 – torture from period April ’01 – March ’04)) 3,042 In the 3 years April 2001-March 2004 therefore shootings and other uses of force (excluding police vehicles) accounted for 1017 (96 per cent) of the total of 1062 deaths. For the purposes of this paper therefore, it will also be assumed that throughout the 7 year period April 1997 to March 2001 shootings combined with other uses of force by the police account for around 96 per cent of deaths as a result of police action, with the balance being vehicle related deaths.12 Apart from ICD statistics and ICD reports,13 as well as occasional press reports, there are a couple of other studies which have been conducted which may be used as additional information in illustrating the nature and causes of deaths as a result of police action. In 1998 the author conducted a study of ICD dockets covering deaths as a result of police action or in police custody for Gauteng province for the period April – December 1997. The 165 deaths covered by the study included 116 deaths as a result of the use of force and 3 vehicle deaths (the balance were deaths in custody). The 116 deaths as a result of 12. In this paper vehicle related deaths are not therefore recorded as uses of force by the police. Note however the remark by Geller and Scott that some US police agencies ‘report as instances of deadly force only events involving police firearms, excluding for instance, high speed pursuits in which the police intentionally ram the suspect’s vehicle’ (ibid) while ‘[o]ther agencies explicitly acknowledge the potentially lethal consequences of certain high-speed pursuit tactics…By some, although not most, police agency definitions, fatal chokeholds, fatal TASER shocks or fatal attacks by police dogs might be classified following investigations as official uses of deadly force’ (Geller & Scott, 1992:24) 13. In addition to statistics the ICD annual reports usually give brief summaries of some of the cases which the ICD has been investigating. 148 South African Review of Sociology 2005, 36(2) the use of force included: shootings of people who ostensibly posed a direct threat to police officers (59), during the course of arrests (34) and during escapes from custody (3). The remaining 20 include: 6 cases potentially related to an assault of a person in custody, 1 cases of a person allegedly beaten to death in other circumstances, and 13 shootings in other circumstances such as ‘shot person in the head while drunk and showing of his pistol’, ‘murder in shebeen using service pistol’ ‘accused of neglect of duty by member of the public who he shot’, ‘intentional family killing using firearm’, possible accidental shootings and some in uncertain circumstances. At least 6 of the use of force incidents occurred while the police officer was off duty. One of the shooting related deaths occurred during the resettlement of a community. Overall then of 119 ‘police action’ deaths, 109 (91.5 per cent) were the result of shootings, 7 (6 per cent) the result of other uses of force, and 3 (2.5 per cent) related to incidents involving vehicles (Bruce, 1998). Another study, which was conducted during the 3 years from January 1998 – December 2000, provides a profile of 102 ‘police action’ and 15 ‘police custody’ deaths. The study was conducted using postmortem records, and interviews with SAPS or ICD personnel attending autopsies, at a mortuary in Durban. Of the 102 police-action deaths 88 (86 per cent) are attributed to shootings, while 11 (11 per cent) are attributed to assaults by police, and 3 (3 per cent) are attributed to ‘assaults by police dogs’14 (Bhana, 2003). These two studies confirm the broad outline of ICD statistics with uses of force accounting for the vast majority of police action deaths, and firearms accounting for the bulk of these. (This account does not deal with the variation over time, and by locality and region, of the broad trends revealed by national level ICD statistics). It is also interesting that the Bhana study records police dogs as a small, though significant, contributor to deaths during the 1998-2000 period. No mention of police dogs is made in ICD statistics, even in the slightly more detailed statistics which the ICD started providing from 2001/2002 onwards. The cases recorded by Bhana would probably have occurred prior to the 7 November 2000 screening, on the South African Broadcasting Corporation’s programme Special Assignment, of the notorious video depicting members of the North East Rand Dog unit setting their dogs on three ‘suspected illegal immigrants’ in a sadistic training exercise (details of the case are in Bruce, 2002a). It would be comforting to believe that the outcry following this incident contributed to an overhaul of the use of police dogs throughout the SAPS, and that the absence of statistics on dog related deaths, in the post 2000 period, is indicative of a profound change in how dogs are being used. However a report compiled on the basis of interviews with persons in custody conducted during late 2001 and the first half of 2002 also contains a number of allegations of the use of police dogs for the purpose of torture (KwaZulu-Natal Campaign Against Torture, 2002). This suggests that the absence of incidents involving the use of police dogs from ICD data on deaths may merely reflects the limitations of detail and explanation of their statistics. 14. Vehicle deaths are not recorded in the sample. South African Review of Sociology 2005, 36(2) 149 Provincial variations Table 3 Deaths as a result of police action in each province, April 1997 - March 2004 1997/ 1998/ 1999/ 2000/ 2001/ 2002/ 2003/ Total 1998 1999 2000 2001 2002 2003 2004 Total Provincial Percentage percentage Distribution of national population Gauteng 135 132 104 94 82 99 73 719 24 18 Limpopo 25 28 27 24 22 21 24 171 6 13 North West 17 25 30 24 22 15 11 144 5 8 Mpumulanga 36 37 32 38 39 28 28 238 8 7 KwaZulu N 165 148 155 108 93 70 147 886 29 21 Free State 39 34 13 17 20 27 12 162 5 6 E Cape 51 66 61 65 49 16 48 356 12 15 W Cape 43 62 45 49 38 34 32 303 10 10 N Cape 7 13 5 8 6 1 5 45 1 2 Total 518 545* 472 427a 371 311 380 3,024 100 100 a.The discrepancies with respect to years ending March 1999 and 2001, as opposed to data in Tables 1 and 2, reflect inconsistencies in reporting of figures by the ICD. The data in Table 3 illuminates variations in the trends in number of deaths in each province. Broadly the differences in the overall numbers of deaths between provinces are fairly consistent with differences between the provinces in relation to factors such as population size and levels of violent crime. Limpopo, for instance, which record levels of police action deaths significantly lower than its proportion of the population, also had the lowest levels of murder and violent crime in the 2002-2003 year (Leggett, 2004:16). At a glance though there are some apparent anomalies: the Western Cape for instance recorded the highest murder and violent crime rate in 2002/03 (Leggett, 2004) but doesn’t record a rate of killings by police significantly higher than it’s proportion of the national population.15 Similarly the Northern Cape has high murder (4th highest) and violent crime (3rd highest) rates (ibid) but a lower rate of police killings than its proportion of the national population.16 KwaZulu-Natal, which consistently records the highest overall number of killings by police (except in 2002/2003), and which has the highest level of killings by police relative to proportion of the population, had the third lowest overall level of violent crime in 2002/2003, though it did have the third highest homicide rate. (Gauteng had both the second highest violent crime and murder rate in that year.) (Leggett, 2004). 15. The rankings provided in this paragraph are from Leggett (2004). 16. However the high murder rates in these provinces are related to high levels of violence between people who are acquainted with each other. Perhaps violence of this kind is less of a predictor of police violence than other more ‘predatory’ forms of violent crime. 150 South African Review of Sociology 2005, 36(2) Differences between provinces are therefore not purely a reflection of different population and crime levels. They may reflect other differences, including differences in the relative efficacy of internal police organisational controls on the use of force, differences in the attitudes of police leadership, or variations in the impact of ICD investigations on police practise in the different provinces. Impact of the amendment to Section 49 The provincial data also provides insight in relation to the sharp increase in the number of killings by police in the 2003/2004 year. This is of particular interest as the amendment to Section 49 of the Criminal Procedure Act, the law regulating the use of lethal force for purposes of arrest, was finally brought into operation on 18 July 2003, after being promulgated by parliament in 1998.17 Evidence from other countries, notably the USA (Geller & Scott, 1992:257-267), predicts that reforms of the law restricting the use of lethal force by police should be associated with overall declines in the number of shootings, and fatal shootings, by police. On the contrary however South Africa saw a substantial increase of 23 per cent in the number of deaths resulting from police shootings in the 2003/2004 year. In fact in 5 of the provinces (Gauteng, North West, Mpumalanga, Free State, and the Western Cape) the figure recorded in the year ending March 2004 is the lowest figure since 1997, with the trend in most of these provinces reflecting a fairly consistent decline since the late 1990s. In Limpopo the figures have remained consistently in the 20s throughout the 7 year period while in the Northern Cape the 2004 figure is one of the lowest recorded in that province. Two provinces however recorded dramatic increases in the year ending March 2004. The increases recorded in ICD statistics in 2004 primarily reflect an unexplained and dramatic increase of 109 deaths in KwaZulu-Natal and the Eastern Cape, as compared to figures for the year 2002/2003. The new law has come into operation subsequent to a period of several years of uncertainty regarding the legal position, during which deaths as a result of police action have nevertheless declined overall. Considering that the new law finally only came into operation when the overall level of killings had already been in decline for several years, it will be difficult to understand whether law reform has directly contributed to reducing the overall number of killings by police. The two provinces where changes in the law were followed by sharp increases in the number of killings also directly contradict this possibility. Comparing current rates and those during the apartheid period What can we say about the impact of police transformation on the use of force by the police? Early in 2000, for instance, a New York Times editorial stated that statistics on killings by police indicate that police in South Africa ‘are more deadly today than they were during most of the last 25 years of apartheid’ (New York Times, 2000). But can one in fact compare current figures to figures on the number of people killed by the police during the 1980s or other years of the apartheid period? 17. See Bruce (2003) for a detailed discussion. South African Review of Sociology 2005, 36(2) 151 Lethal police violence during the apartheid period The key source of information about the number of people killed by the police during the apartheid period was the Minister of Law and Order, who would provide a set of figures each year in response to questions in parliament. These figures are recorded in the annual surveys of the South African Institute of Race Relations (SAIRR). A scan of the SAIIR reports indicates that in 1985, a year in which a State of Emergency was declared in response to heightened resistance, 763 people were reported as killed by the police (SAIRR, 1986:484). This figure appears to be the highest reported number for persons killed by the police in any single year in South African history. The figure of 763 represents an increase of 166 per cent on the figure of 287 for the previous year (1984) (SAIRR, 1985:788), and is also higher than the 716 killings reported in 1986 (SAIRR, 1987:862). A survey of SAIRR reports for the entire period 1970 to 1985 indicates that even the 1984 figure of 287 persons killed is significantly greater than that reported for any previous year (Foster & Luyt, 1986). The survey points out that these figures exclude those killed during the civil unrest of 1976 and 1977, but include persons killed in unrest during 1984 and 1985. The figures given for persons who died in non-unrest related circumstances are 202 in 1976 and 149 in 1977. According to the official enquiry into the upheavals in Soweto and other townships during the period 16 June 1976 to 27 February 1977, the total number of people who died at the hands of the police was 451. If, for the sake of this analysis, we assume the vast majority of these people died in 1976, we might add the entire figure to the 1976 figure for non-unrest related deaths. This gives a total of roughly 653 people who died at the hands of the SAP during 1976, in both unrest and non-unrest related circumstances. Using this method of calculation it would appear that, according to official figures, the three years in which most people were killed by the SAP were 1985 (763 people), 1986 (716 people) and 1976 (an estimated 653 people). According to the Minister of Law and Order a further 400 people were allegedly killed by the police in 1987 (SAIRR 1988:599). There is also evidence that the number of killings by the police in 1996 might have been in the region of 380.18 Therefore apart from the years 1976, 1985, 1986, 1987, 1996 (Table 4) and as suggested by the TRC report in 1990,19 all the formal evidence indicates that prior to 1997 the highest figure for people killed by the police was the 287 persons recorded as having been killed in 1984. 18. According to figures provided to the Human Rights Committee by the SAP at that time there were a total of 550 police action and custody deaths. (The email message is not explicit on this question but it appears that these incorporate both police custody and action deaths.) Assuming that people killed by the police accounted for 65-70 percent of this number (this was the pattern in the late 1990s) this suggests that police killed 360-380 people. (Source of figure of 550 - Email received from Jeremy Sarkin, Professor of Public Law and Deputy Dean, University of the Western Cape, 21 October 2000). Figures provided by Sarkin for 1994 and 1995 are 379 and 260 respectively. The 1996 figure was provided to the HRC by the SAPS but then denied by the SAPS when the HRC publicized the figure. 19. A graph in the TRC report indicates that after peaking in 1985, killings by police declined over the years 1987 -1989, before peaking again at about 300 deaths in 1990 (TRC, 1998, Vol.3:9). 152 South African Review of Sociology 2005, 36(2) Table 4 Official figures on people killed by police in South Africa prior to creation of ICD in 1997 5 highest figures Year Number allegedly killed Original Source 1985 763 Minister of Law and Order 1986 716 Minister of Law and Order 1976 653 (if all 451 ‘unrest’ related police killings recorded by Cillie Commission are allocated to 1976) SAP + Cillie Commission Report 1987 400 Minister of Law and Order 1996 380 Estimate based on figures provided to Human Rights Committee, and then denied, by SAPS While we may doubt the accuracy of these official figures, there is no source which clearly contradicts them. The TRC report for instance, based as it is on victim statements, does not contain any systematic attempt to calculate the number of people killed by members of the SAPS, or the security forces generally, during the period (1960-1994), which it was authorised to investigate. While no figures are provided, rough graphs indicate that 1985 and 1986 were the years the TRC received reports of the greatest number of killings (roughly 500, and 450 respectively) by members of the SAP (TRC, 1998, Vol.3:9). The graph links the SAP to more deaths during 1990 (roughly 300) than in 1976 (just over 100).20 The imprecise figures suggested in the TRC report are significantly lower than those provided by the SAP or Minister of Law and Order. However, we need to bear in mind that the TRC was only concerned with incidents of a ‘political’ nature while the figures provided by government at the time, did not distinguish ‘political’ from ‘criminal’ deaths. While the SAP figures purport to record all killings by SAP members, the TRC figures only covers cases where family members or other people made a statement to the TRC. There can be little doubt that this did not lead to comprehensive coverage, even of killings of a political nature.21 Nevertheless the TRC report confirms the trends reflected in the official statistics, including that 1985 was the year in which the highest number of people were killed by members of the SAP. 20. The TRC by contrast recorded 8,901 killings in South Africa, and 342 outside South Africa through the entire period 1960-1994 which was the period of its mandate (TRC, 1998, Vol.3:4). A rough graph in the report indicates that the TRC recorded in the region of 2,600 killings by police for this 35 year period. By comparison the Inkatha Freedom Party are linked to about 4,400 killings, the African National Congress to 1,300, the South African Defence Force to about 350, the KwaZulu Police to about 250, and the Ciskei security forces and ‘comrades’ each to about 100 (TRC, 1998, Vol.3:9). 21. As opposed to the 8,901 killings recorded in South Africa by the TRC (see previous footnote), Coleman cites figures indicating that, in the period July 1990-April 1994 alone, 13,933 deaths in political violence were recorded (Coleman, 1998:163). South African Review of Sociology 2005, 36(2) 153 Lethal police violence, 1997-2004 Since the ICD started operating in April 1997 it has recorded an average of almost 690 deaths a year with these deaths peaking at 756 in the year to 31 March 1999. As demonstrated the category ‘deaths in police custody or as a result of police action’ however cannot be equated with persons killed by the police particularly as most people who die in custody are not killed by the police. Instead this paper has suggested that in the region of 96 per cent of ‘deaths as a result of police action’ (i.e. excluding all deaths in custody) may be the best figure for estimating the number of people killed by the police (lethal police violence) for the four year period April 1997 – March 2001 (Table 5). Table 5 Estimates for people ‘killed by the police’ derived from ICD statistics April 1997 – March 2004 Persons killed by police Notes on calculations 1997/1998 497 518 x 96% 1998/1999 536 558 x 96% 1999/2000 453 472 x 96% 2000/2001 415 432 x 96% 2001/2002 353 345 shootings + 7 - beaten with hands, fists or other object, 1 – tortured 2002/2003 302 293 shootings + 8 assault/beating + 1 torture 2003/2004 362 360 shootings + 2 beaten with hands/fists Total: 7 years 2,918 This method of calculation is intended to exclude deaths in motor vehicle accidents from the statistics. Applying this method to the ICD figures of 558 deaths as a result of police action in the 1998/1999 year therefore provides a figure of 536 people killed by the police during this 12-month period. According to available information the number killed in the year 1998/1999, and apparently in each of the three years from April 1997 to March 2001, was greater than the number killed by the SAP in any year except for 1985, 1986 and 1976. Comparable statistics? What then should be made of this? First we should be clear about what exactly is at issue. Are we strictly concerned with comparing the ‘lethality’ of the SAP with that of the SAPS? We may delineate the discussion as a comparison of the number of people killed by state agencies inside South Africa. To do this would be to exclude killings by the South African government of South African nationals outside South Africa, victims of the South African Defence Force (SADF) during the Namibian and Angolan wars, as well as victims of the campaigns of destabilization in Mozambique and elsewhere, and of the South African arms 154 South African Review of Sociology 2005, 36(2) industry. If the discussion were to be widened to killings by state agencies inside South Africa then killings by the SADF should also be taken into account.22 Particularly during the mid and late 1980s the SADF played an extensive internal role. On the basis of victim statements the TRC recorded roughly 350 killings by the SADF during the 1960-1994 period. However it is not clear from the report in what years these killings took place, or what proportion took place inside, as opposed to outside, South Africa. Second, it is clear that official figures provided during the apartheid period excluded certain categories of killings by the police. Thus official involvement in the killings of people in detention or custody, assassinations, ‘disappearances’, and other covert murders were not acknowledged. As many people were aware at the time, and as the TRC has confirmed, these practices were not uncommon during apartheid. The TRC also received substantial evidence pointing to collusion between the SAP and Inkatha (later the Inkatha Freedom Party, or IFP). The TRC report details involvement by the SAP, and other state security structures, in the provision of weapons to the IFP, particularly during the 1990s (TRC, 1998, Vol.2:605-610). Arguably therefore, in providing a tally for killings for which the SAP were responsible it might be reasonable to add to this figure many of the killings which were attributed to the IFP (in the sense that the SAP shared responsibility for these killings with the IFP). The TRC information indicates that killings by IFP were at their highest levels during the 1990 to 1994 period with roughly 3,500 people killed by IFP members in this period (TRC, 1998, Vol.2:176).23 Then there is a third consideration. The statistics issued by the Minister of Law and Order, particularly for the later years of the apartheid period,24 exclude deaths in ‘independent homelands’. These geographical areas are now, again, integrated into South Africa, and members of the former homeland police are now part of the SAPS. If deaths at the hands of homeland police services were added to those for the apartheid period these figures would obviously be higher.25 22. In addition unless the discussion is defined in relation to ‘non-judicial killings’ people sentenced to death and executed would also have to be included in these calculations. From the creation of the Union of South Africa in 1910 until the end of 1988 over 4 200 people were hanged in South Africa. From 1978 until the end of 1988 a total of 1,335 people were executed in South Africa (excluding the nominally independent ‘homelands’), the number exceeding 100 each year except for 1981 and 1983. (Coleman, 1998:82). Coleman provides a ‘partial list’ of 49 executions in the 1963-1989 period which were identifiable as ‘political’ of which all but two took place in the years 1963-64 and 1983-1989. In late 1989 a moratorium was declared on executions and the death penalty was declared unconstitutional by the Constitutional Court in 1995. According to the TRC over 2 500 people were hanged during the 1960-1994 period. Political executions during the 1960s included ‘almost one hundred Poqo activists hanged for involvement in acts of violence’ (TRC, 1998, Vol.2:169). 23. The actually death toll at the hands of Inkatha may have been higher (see footnote 21 above). 24. The first ‘homeland’ to be granted ‘independence’ was Bophuthatswana in 1977. 25. See footnote 20 for TRC figures on persons killed by the KwaZulu police and Ciskei security forces. South African Review of Sociology 2005, 36(2) 155 Reliability of SAP & SAPS statistics on deaths Even if we simply focus on the SAP and SAPS and, in relation to the apartheid period, restrict the discussion to ‘formal’ (rather than covert) killings, there is substantial reason to doubt the reliability of statistics on killings by police provided by the SAP, who would have been the source of the statistics in question prior to 1994. In early 1997, just prior to the opening of the ICD, the SAPS produced a document which was distributed under the letterhead of the office of the National Commissioner and which purported to record all police action and custody deaths over the 30-month period January 1994 to June 1996. The document recorded a total of 500 of these deaths during the period in question. Soon afterwards, when the ICD began, it became apparent these deaths were occurring at an entirely different rate, with roughly 1,800 recorded in the first 30 months of the ICD’s operation. If this report were regarded as reliable then this would indicate that between the years 1994-1996 and the years 1997-1999 there was a sudden, dramatic and completely unexplained increase of three to four times in the number of these deaths. Assuming this to be highly unlikely, it therefore appears clear that the SAPS report’s seriously under-represented the total number of these deaths, possibly to the extent of only recording between a quarter and one third of the deaths which occurred in the 1994-June 1996 period. But was this a deliberate misrepresentation? Further evidence suggesting that a tradition existed of ‘manipulating’ statistics emerged when the SAPS reported to a non-governmental organisation that 550 people had been killed by police or died in their custody during 1996. When the non-governmental organisation then reported these figures the SAPS then disputed that these were in fact correct. The figure of 550 is a highly plausible figure, consistent with the figures which were reported, from 1997 onwards, by the ICD. It appears fair to infer that in this instance the SAPS released ‘unfalsified’ information by mistake and then engaged in an exercise in damage control.26 In 1995 papers submitted to the Northern Cape Division of the Supreme Court in the Raloso matter on behalf of the then SAPS National Commissioner, provided what were alleged to be statistics ‘on civilians killed during arrest’ which were clearly fabricated, apparently in order to provide support for certain contentions made in the papers.27 Some evidence therefore exists that statistics on police related deaths provided by the SAPS in 1995-1996, were not only unreliable, but also subject to manipulation. It is reasonably to assume that this continued a tradition which had been established years previ26. See footnote 18 above for details. 27. The papers contended that a ‘sharp decline in the injuring and killing of persons during arrests, or attempted arrests is partly related to the fact that policemen are hesitant to use their firearms, mainly as a result of the uncertainty regarding the provisions of Chapter 3 of the Constitution and the human rights provisions contained therein’.See affidavit of John George Fivaz, paragraph 10.5 and annexure GF3(a)-(j) to this affidavit in the matter of Raloso v Wilson and Others reported as 1998 (4) SA 369 (NCD);1998 (1) BCLR 26 (NC). An argument demonstrated that the information provided in the affidavit is not authentic is provided in the supplementary affidavit of Robert David Bruce, paragraphs 33-44 in the matter of S v Walters and Another, Constitutional Court Case No. 28/01. 156 South African Review of Sociology 2005, 36(2) ously, by the South African Police and government. Even if there were not evidence which suggested that statistics may have been falsified, it is also possible that, in the profoundly racist society that was apartheid South Africa, there was often a dismissive attitude towards the need to actually record deaths of black people. As a result it is quite likely that many deaths simply did not make it into statistics. Situations of intense mass based political violence, such as those which characterised South Africa in various periods, may also make it more difficult to record all deaths, assuming that there were the will to do so. If one accepts that statistics on people killed by the police, provided by the SAPS or SAP up to 1996 are of doubtful reliability, then one is left with no reliable and comprehensive information on killings by the police in South Africa prior to this point. Since 1997 however the police have been obliged to inform the ICD about each death,28 and the provision of this information is no longer centrally controlled. The ICD operates independently from the SAPS, and its figures on deaths may be regarded as more reliable and accurate than those previously available. Conclusion When the ICD started releasing statistics on deaths as a result of police action these were interpreted by some as indicating that police behaviour had not really changed in South Africa and that people were being killed by police in South Africa at rates comparable to those during the time of apartheid. But statistics cannot be used to prove or disprove these perceptions as it is reasonable to assume that official statistics on killings by police before the ICD started operating in 1997, were not reliable. While this has not been done here it is possible also to compare South African statistics, on people killed by the police since 1997, to statistics from other countries. In relation to the indicators that tend to be used for comparison29 (killings by police compared to: the overall murder rate, the rate of killings of police, numbers of policel; numbers of people wounded in shootings as opposed to killed in shootings by police) it should be remembered that the time at which the highest number of police killings was being recorded was also the time at which South Africa was experiencing its highest murder rates, and a notably high number of killings of police.30 South Africa since 1997 is therefore not an example of the type of phenomenon, characteristic of the Brazilian city of Sao Paulo in the early 1990s, of extreme uncontrolled police violence.31 Furthermore the years since 1997 have seen a decline in the number of killings by police with five provinces recording their lowest levels, since 1997, in the 2003/2004 year. This trend has matched other declines in violence related deaths with both murder over28. 29. 30. 31. Section 53(8) of the South African Police Service Act, 68 of 1995. See for example Geller and Scott (1992), Chevigny (1995), Cano (1997). See Bruce (2002b) on the killings of police in South Africa. In the years 1991-1992 levels of police violence in Sao Paulo reached extreme levels with 1,470 civilians killed by the police in the state of Sao Paulo in 1992 alone (Chevigny, 1995:148), 1,190 of these in the Sao Paulo metropolitan area (Human Rights Watch, 1997:51). South African Review of Sociology 2005, 36(2) 157 all32 and killings of police33 having declined during the same period. Notwithstanding these positive trends however, one cannot assume that killings by police have been brought satisfactorily under control. In KwaZulu-Natal and the Eastern Cape, police killings have escalated dramatically in the last year, despite the introduction of new lethal force legislation. In addition, while information on the overall number of shootings may not demonstrate that South African police are killing an unjustifiably high number of people, existing information on the quality of the use of firearms by police suggests cause for serious concern. ICD statistics over the 3 years ending March 2004 indicate that 1 in every 27 (37 out of 998) people who is killed in a police shooting is an innocent bystander who is presumably shot in error. Shooting policies in responsible police agencies should place a premium on the protection of human life, and particularly the lives of innocent civilians.34 It seems however that these are not concerns which have broadly permeated through to police in South Africa. If error takes such a high toll of innocent bystanders (a number of bystanders are also likely to be injured in shooting incidents) it is also possible that a significant number of supposed ‘suspects’ who are killed are also the victim of ‘mistaken identity shootings’. Such shootings would not necessarily be picked up by ICD investigations. While a number of unjustified shootings may not be picked up by the ICD, there is still significant evidence of a problem of illegal killings by police. The latest ICD report, for instance, appears to indicate that 13 police were convicted of murder, and 2 of culpable homicide, in the year ending 31 March 2004. Considering that an investigative body such as the ICD, which faces significant resource and capacity constraints, would not necessarily be able to uncover all illegal killings, and would be unlikely to secure convictions in relation to all cases where there is evidence of criminality, this suggests a significant problem of unlawful police violence.35 The number of killings of bystanders, the number of criminal convictions, and other problems such as that of apparently unlawful off duty killings (on the latter see Bruce and O’Malley, 2001) therefore suggest that there is a significant problem of misuse of lethal 32. Official statistics record a decline in murder from 24,486 in 1997/98 to 21,553 in 2002/03, a decline of 13 per cent during this six year period. (Crime Statistics as released on 2003-09-22, www.saps.gov.za). 33. Police statistics indicate that 244 police were killed in 1997, and 163 in 2001, a decline of 50 per cent during this five year period. (Bruce, 2002b). 34. For instance the Philadelphia Police Department’s deadly force policy states that ‘Above all, the safety of the public and the officer must be the overriding concern whenever the use of firearms is considered’. The Washington D.C. Metropolitan Police Department indicates in its lethal force policy that lethal force may only be used for arrest if ‘the lives of innocent persons will not be endangered if deadly force is used’. The Los Angeles Police Department’s policy manual states that ‘Officers shall not fire under conditions that would subject bystanders or hostages to death or possible injury, except to preserve life or prevent serious bodily injury (all quoted in Police Assessment Resource Centre, 2003:26, 35) 35. An examination of police shooting incident data covering the 1996-1998 period provides evidence that an exceptionally high number of unlawful killings by police take place when police are off duty (Bruce & O’Malley, 2001). 158 South African Review of Sociology 2005, 36(2) force by SAPS members. Since the advent of democracy in 1994 the SAPS has at no point engaged in a comprehensive re-examination of its systems for supporting and controlling the use of lethal force by police members.36 As a result, in so far as the SAPS has dealt with the issue, this has been in an ad hoc and inconsistent way. But police in South Africa are supposed to uphold the South African Constitution and this implies that they need to give high value to the protection of human life, including the lives of police officers, suspects, and innocent civilians. The SAPS, and perhaps other police agencies,37 should therefore revisit their policies, and improve their systems for controlling the use of lethal force, if they are more fully to meet these standards. The continuing availability of information on police killings, and evidence of some success in bringing police responsible for unlawful killings to justice, may also be seen as an endorsement of the value of an independent investigative mechanism such as the ICD, which was initially provided for in South Africa’s interim Constitution. Acknowledgments Thanks are due to Muff Anderson for editing an earlier version of this paper. References Bhana, B. 2003. Custody-related deaths in Durban, South Africa 1998-2000. The American Journal of Forensic Medicine and Pathology, 24(2):2002-2007. Bruce, D. 1998. Towards a strategy for prevention: the occurrence of deaths in custody or as a result of police action in Gauteng, April - December 1997. Report for the Independent Complaints Directorate. Johannesburg: Centre for the Study of Violence and Reconciliation. Bruce, D. 2000. The management of the use of force in the South African Police Service. Masters of Management research report. Johannesburg: University of the Witwatersrand. Bruce, D. 2002a. Police brutality in South Africa. in Mwanajiti, N., Mhlanga, P., Sifuniso, M. NachaliKambikambi, Y., Muuba, M and Mwananyanda, M (eds.) Police brutality in Southern Africa: a human rights perspective. Lusaka: Inter-African Network for Human Rights and Development (Afronet). Bruce, D. 2002b. Stopping cop killing: Lessons and limitations of South African research. SA Crime Quarterly, 2:13-17. Bruce, D. 2003. Killing and the Constitution: Arrest and the use of lethal force. South African Journal of Human Rights, 19(3):430-454. Bruce, D. & O’Malley, G. 2001. In the line of duty? Shooting incident reports and other indicators of the use and abuse of force by members of the SAPS. Report for the Independent Complaints Directorate. Johannesburg: Centre for the Study of Violence and Reconciliation. Cano, I. 1997. The use of lethal force by police in Rio de Janeiro. Rio de Janeiro: Instituto de Estudos da Religiao (ISER). Chevigny, P. 1995. Edge of the knife: police violence in the Americas. New York: New Press. 36. See Bruce (2000). Intensive steps were taken in relation to specific units, notably Public Order Police, but not more broadly. 37. The deaths as a result of police action recorded in 2003/2004 include 16 deaths linked to municipal police services as follows: Durban City Police – 6, Cape Town Metropolitan Police – 4, Ekurhuleni Metropolitan Police – 5, Johannesburg Metropolitan Police – 1. South African Review of Sociology 2005, 36(2) 159 Coleman, M. (ed.) 1998. A crime against humanity: Analysing the repression of the apartheid state. Claremont: Human Rights Committee, Mayibuye Books, David Philip Publishers. Dissel, A. & Ngubeni, K. 1999. A lonely way to die: An examination of deaths in police custody. Report for the Independent Complaints Directorate. Johannesburg: Centre for the Study of Violence and Reconciliation. Fernandez, L. 1991. Police abuses of non-political criminal suspects: A survey of practises in the Cape Peninsula. Institute of Criminology, University of Cape Town. Foster, D. & Luyt, C. 1986. The blue man’s burden: Policing the police in South Africa. South African Journal on Human Rights, 2:297-311. Geller, W. & Scott, S. 1992. Deadly force: What we know. Washington DC: Police Executive Research Forum. Human Rights Watch / Americas 1997. Police brutality in urban Brazil. New York: Human Rights Watch. Independent Complaints Directorate. 1998. 1997-98 Annual Report. Pretoria: Independent Complaints Directorate. Independent Complaints Directorate. 1999. 1998-99 Annual Report. Pretoria: Independent Complaints Directorate. Independent Complaints Directorate. 2000. 1999-2000 Annual Report. Pretoria: Independent Complaints Directorate. Independent Complaints Directorate. 2001. 2000-01 Annual Report. Pretoria: Independent Complaints Directorate. Independent Complaints Directorate. 2002. 2001-02 Annual Report. Pretoria: Independent Complaints Directorate. Independent Complaints Directorate. 2003. 2002-03 Annual Report. Pretoria: Independent Complaints Directorate. Independent Complaints Directorate. 2004. 2003-04 Annual Report. Pretoria: Independent Complaints Directorate. KwaZulu-Natal Campaign Against Torture. 2002. Report for the KwaZulu-Natal Provincial Office of the South African Police Services on the findings of the pilot project to monitor six police stations in the Durban area for incidents of torture. Unpublished Report. Leggett, T. 2004. What’s up in the Cape? SA Crime Quarterly, 7:15-20. New York Times. Democracies and their police. New York, 1 May 2000. Police Assessment Resource Center. 2003. The Portland Police Bureau: Officer-involved shootings and in-custody deaths. Available online at: www.parc.info. SAIRR. 1985. Race Relations Survey 1984. South African Institute of Race Relations: Johannesburg. SAIRR. 1986. Race Relations Survey 1985. South African Institute of Race Relations: Johannesburg. SAIRR. 1987. Race Relations Survey 1986. Johannesburg: South African Institute of Race Relations. SAIRR. 1988. Race Relations Survey 1987/88. Johannesburg: South African Institute of Race Relations. SAIRR. 1989. Race Relations Survey 1988/89. Johannesburg: South African Institute of Race Relations. South African Police Service , Crime Statistics as released on 2003-09-22, www.saps.gov.za. TRC. 1998. Report of the Truth and Reconciliation Commission of South Africa (Vols.1-5). Cape Town: Truth and Reconciliation Commission, Department of Justice.