Factors Associated With Rape Case Attrition in the South African Criminal Justice System: A National Cross-Sectional Study

Abstract

BACKGROUND

Attrition of rape cases after being reported to the police is a global problem (Lea et al. 2003; Jewkes et al. 2009; Morabito et al. 2019b). Research from diverse countries shows that the ideal of rape perpetrators being arrested, charged, taken to trial and convicted of rape with a hefty sentence is only a path traversed in a small minority of cases. The Tracking Justice Study which investigated attrition of reported cases in one South African province in 2003 found that out of every 100 complaints of rape registered with the police, 50 arrests were made, 43 accused were charged, 17 trials commenced and 6 perpetrators were found guilty (Vetten 2008; Vetten et al. 2008). It also found that DNA results were only available for 2 per cent of cases and made no difference to trial outcomes (Vetten 2008; Vetten et al. 2008). However, finding injuries and documenting them on medicolegal forms (referred to as J88 forms) significantly enhanced the likelihood of conviction, even after adjusting for other factors (Jewkes et al. 2009). The conviction rate reported in Tracking Justice (6 per cent) was very similar to the 6.5 per cent convicted in six jurisdictions in the United States (Morabito et al. 2019b) and 9 per cent reported in England and Wales in 1997 (Harris and Grace 1999). In these three countries, cases were particularly vulnerable to attrition at the initial police investigation phase (Lea et al. 2003; Artz and Smythe 2007a; 2007b; Vetten et al. 2008; Jewkes et al. 2009; Morabito et al. 2019b) Whilst in some cases the perpetrator is unknown, research in multiple countries has shown that a combination of factors, only some of which relate to the law influence the progress of rape cases across the criminal justice system (CJS) (Lea et al. 2003; Artz and Smythe 2007a; 2007b; Jewkes et al. 2009; Alderden and Ullman 2012; Spohn et al. 2014; Smythe 2015).

There have been legal reforms in many countries in response to the identified problem of rape case attrition. This includes changes to the administration of the criminal justice system and development of more victim-friendly services with better approaches to gathering forensic evidence. South Africa has introduced similar measures to reduce case attrition including legislation in the form of the Criminal Law (Sexual Offences and Related Matters) Amendment Act 32 in 2007 (Republic of South Africa 2007), training for health professionals (Jina et al. 2008) and rolling out one stop centres (Thuthuzela Care Centres) and Sexual Offences Courts to improve services for survivors and conviction rates (National Prosecuting Authority 2014). These measures are important for rape survivors and it sends a key message to society about a desire to end impunity, however there is limited evidence of impact on the overall outcomes of cases within the CJS because, to our knowledge few studies focusing on rape case attrition have been conducted since the law reforms. The National Prosecuting Authority commissioned research aimed to better understand rape case attrition and its drivers in the South African CJS. We report findings of this research and will describe attrition in rape cases reported to the police in 2012 and factors underlying attrition at the stages of police investigation, prosecutorial case enrolment and trial outcomes.

South African Sexual Offences legislation and directives

The 2007 Criminal Law (Sexual Offences and Related Matters) Amendment Act 32 provides definitions of sexual offences, directs an integrated and multi-sectoral approach to handling cases and seeks to provide justice for victims in a supportive victim-friendly environment (Republic of South Africa 2007). Sexual offences include a broad range of acts and rape or compelled rape are a distinct category of these. This legislation has a broad, gender-neutral definition of rape that encompasses any sexual penetration with a body part or object into the mouth, anus or vagina of another without consent. The legal definition of rape also encompasses a wide range of circumstances in which sexual acts are definitionally coerced including when the victim is threatened, is intoxicated or drugged, unconscious, when the perpetrator impersonates another or obtains consent by fraud, and when sex is with a minor (age under 12, and at 12–15 years with some caveats if the other party is also under 16) or a person with intellectual disabilities (Republic of South Africa 2007). The category of compelled rape is defined as when a person is coerced to sexually penetrate another, in which case the perpetrator of rape is defined as the person forcing the sexual act (Republic of South Africa 2007).

There are a set of accompanying National Directives and Instructions on examining and caring for survivors. When a case is opened, the South African Police Service (SAPS) is required to arrange a medico-legal examination of the victim, with findings documented on the J88 form and the health care provider must complete a Sexual Assault Evidence Collection Kit (SAECK) which must be submitted by the police to the Forensic Scientific Laboratory (FSL). According to the National Directives for the SAPS, detectives must conduct a thorough and professional investigation of every reported case in an appropriate, efficient and sensitive manner (Republic of South Africa 2007; South African Police Services 2008; National Prosecuting Authority South Africa 2014). Detectives must visit and compile a description of the place of rape, collect forensic evidence and send it to the FSL for analysis (South African Police Services 2008). They must obtain testimonial evidence from the victims, witnesses and identified perpetrators, captured as a written statement under oath. Detectives must liaise with the prosecutor during the investigation and hand the case file over to commence pre-trial proceedings (South African Police Services 2008).

On receiving a case file (commonly referred to as a docket) a prosecutor should assess the evidence collected and decide whether it is adequate for case enrolment at court and trial (Republic of South Africa 2007; National Prosecuting Authority South Africa 2014). To enrol a case, the prosecutor has to be satisfied that a prima facie case exists and that the evidence links the perpetrator to the crime (Republic of South Africa 2007; National Prosecuting Authority South Africa 2014). If it is not, the prosecutor should advise the detective to collect further evidence, and where this is not possible, decline case enrolment (Republic of South Africa 2007; National Prosecuting Authority South Africa 2014). The case might then be presented in court; however, cases can be withdrawn at any stage up to the commencement of the trial and re-enrolled later if further evidence becomes available.

Attrition points within the South African CJS

Attrition of rape cases within the criminal justice system is inevitable. Some cases cannot feasibly progress to an arrest, as there are no clues to the accused’s identity. For other cases, the case opened by a complainant is assessed at various points on its merits, both by the complainant or by officials. A decision is taken either to proceed within the system, or if the benefits of this are doubtful, a decision is made not to proceed. Consequently, the potential attrition points include: at the reporting stage, where a police officer may use their discretion not to open a case if they are not convinced of the reliability of the complainant’s statement; at the police investigation stage, when either the suspect remains unidentified and chances of locating or arresting the suspect are deemed nil or very low, or when the victim is subsequently untraceable or expresses disinterest in pursuing the case leading to police closing the case as undetected or closed due to victim withdrawal; at the prosecution stage, when the prosecutors decline to prosecute for different reasons, including the low strength of evidence collected; prior to the start of trial, the prosecutor may withdraw the case due to the lack of co-operation from the complainant, the disappearance of the perpetrators or other reasons; or after the trial has started there is failure to establish a prima facie case or other reasons leads to the case being discharged by the court.

Framework for assessing factors associated with rape case attrition

Rape case attrition occurs due to the interplay of multiple factors related to the victims’ characteristics and their accounts about alleged perpetrators and circumstances of the rape incident, which are assessed at the different stages of the criminal justice by actors including the police, prosecutors and the judiciary. Evidentiary assessments are supposed to be standard and based on set investigative, prosecutorial and judicial guidelines. However, individual criminal justice actors’ perceptions and motivations, can influence the success, or otherwise, of a case. The highest rape case occurs at the police investigation phase driven by victim characteristics and circumstances of the incident, investigative rigor, police member attitudes towards gender norms and acceptance of rape myths, availability of resources and staffing in police stations and cooperation between different police clusters and other actors in the criminal justice system (Watson 2015). Victims’ participation, which is affected by their recall and testimonial evidence of the incidents, their own agency, confidence in the criminal justice system to deliver justice or discouragement by significant others, also impacts case attrition (Smythe 2004; 2015). Perpetrator characteristics may also pose challenges for investigations—many non-arrests of stranger suspects and case closures as ‘undetected’ are because the complainant cannot identify the perpetrator, or the police lose contact with alleged perpetrators. Forensic evidence collection, analysis and case linkage may mitigate the challenges of perpetrator non-identification. However, this is dependent on investigator adherence to policing guidelines. The availability and use of forensic evidence in trials is also dependent on timing and the personnel’s ability to collect evidence, the quality of the evidence collection and the capacity of the forensic science laboratory to process the samples.

Prosecutors and judicial officers may also hold misogynistic stereotypes, blame victims, be skeptical of complainants’ reports and view victims as not credible, which negates the progression and outcomes of cases at the respective system stages. Specific beliefs regarding rape victims and their behaviour during and following rape, also referred to as rape myths, may form a basis for not proceeding with cases. Prosecutors may err towards selecting cases that fit into the stereotypical ‘real rapes’ with ‘genuine victims’ (Du Mont et al. 2003). In this way, the victim characteristics, circumstances of the rape and rape stereotypes are key to prosecutors’ views about the cases’ chances of conviction. Prosecutors may screen out cases that are unlikely to result in a conviction due to questions about the victim’s character, the victim’s behaviour at the time of the incident and the victim’s credibility. While victim characteristics play a role in attrition at the prosecution stage, the decision about whether to institute criminal proceedings depends on the investigative rigor, quality and strength of evidence collected by police. The strength of a case depends on the extent to which there is sufficient and admissible evidence to provide a reasonable prospect of successful prosecution (Watson 2015). Cases where procedural and evidentiary requirements for trial cannot be met are either dismissed or culminate in non-conviction (Figure 1).

Fig. 1.

Framework for factors associtated with attrition. Alderden and Long, 2016, Alderden, and Ullman 2012, Artz and Smythe, D. 2007a, Artz and Smythe, 2007b, Bougard and Booyens 2015, Du Mont et al. 2003, Forr et al. 2018, Frohman 1997, Gregory and Lees 1999, Grubb and Turner 2012, Hansen 2019, Jina et al. 2011, Kaiser et al.,2017, Koss 1985, Koss et al. 1988, Krahe et al. 2008, Lynch et al 2013, Maddox et al. 2011, Morabito et al, 2019a, Nagel et al. 2005, Nishith et al. 2001,O’Neal and Spohn, 2017, O’Neal et al. 2015, Santtila et al. 2004, Siller 2018, Smythe, 2004, Smythe 2015, Snyder 2000, Spohn et al. 2001, Spohn et al. 2014, Spohn and Tellis, 2019, Swemmer, 2020, Taylor and Gassner, 2010, Ullman et al. 2005, Van der Watt et al. 2015, Vetten et al. 2008, Vetten and Haffejee, 2005, Waterhouse et al. 2016, Watson 2015, Woodhams, and Cooke, 2013, Woodhams et al. 2012, Woodhams et al. 2019, Wong, and Balemba, 2016, *Point of highest attrition (Artz and Smythe, 2007b, Munro and Kelly, 2009, Daly and Bouhours, 2010, Burman et al., 2009, Vetten 2008; Vetten et al. 2008, Lea et al., 2003);**Point of second highest attrition Coloured arrows depict the linkages between factors and impacts on attrition at the CJS stages. Notes: In Some instances, a rape victim may be seen at a health facility first and have evidence collected before reporting to the police for the purpose of this paper, we used a guilty verdict as the outcome and did nor include sentencing outcomes in the framework.

Open in new tabDownload slide

METHODS

Study design and sampling

The study used a retrospective multi-stage random sampling study design so that cases were identified in provinces, clusters and in individual police stations. The retrospective study design and use of rape cases reported at police stations across the country in 2012 provided the opportunity to track the progression of cases over time without interfering with processing of cases. By the time of the study, the majority of cases had been closed for investigation and processing. From a total 1,134 police stations across nine South African provinces, we randomly selected 172 police stations from a national list using a strategy of probability proportionate to size, stratified by province and police station rape case load (Machisa et al. 2017). The sample size calculation was based on selection of a sample of 4,300 cases with an estimated 500 cases in each of eight provinces and 300 cases in the ninth (least populous) province. At each selected police station, researchers obtained the list of all rape matter cases (referred to as ‘rape’ in this paper) reported for the period 1 January to 31 December 2012 and systemically selected 30 case numbers (Machisa et al. 2017). Systematic selection is a probability sampling approach that involved, choosing a random starting point on a case list and then selecting cases with a fixed, periodic interval. The sampling interval was calculated by dividing number of cases on each police station list by our desired sample size of 30 cases per police station. In police stations where the total rape cases for 2012 were less than 30, all case numbers were included into the study (Machisa et al. 2017). Some of the selected dockets could not be found in docket archives and substantial parts of the dockets were not available so that they could not be included into the sample (N = 348). The final sample of cases that we studied was 3,952, which is 92 per cent of the targeted sample.

Data collection

Members of police assisted trained research assistants to access available dockets from archives. The research assistants used a structured paper-based forms to capture detailed information from these. Dockets typically contained background information about the victim including race, age and biological sex, circumstances of the assault contained in a sworn statement, and often characteristics of the perpetrators documented in statements. Against a backdrop of long history of apartheid and racial segregation, South African official race classifications for the population are still divided into five major racial categories: Black Africans, Whites, Indians, Coloreds, who have mixed White and Black descent, and ‘Other’ for race groups who don’t fit neatly into the four main race groups (Statistics South Africa 2012). Dockets also contained case management information including the detective’s investigation diary, detection and arrests of perpetrators, the medical examination or J88 form, facial composites, use and processing of the Sexual Assault Evidence Collection Kit (SAECKs), and DNA evidence by the Forensic Science Laboratory (FSL), charge sheet, court hearing dates, notes showing case progression and decisions made in respect of the prosecution and trial stages, and trial outcome (if any). As mentioned, dockets must contain charge sheets which outline the various crimes the perpetrator will be prosecuted for, otherwise these were collected from the Courts where the trial took place. Data were collected from different documents in the docket. Table 1 shows the data variables that we collected and their source documents.

Data management and analysis

In some of the selected dockets, information was missing. To avoid line deletion in the analysis when data was missing, we created a ‘missing data’ category for each key variable used in analysis. Particularly a third of the medical examination forms (J88s) were missing in dockets (N = 1,142), yet our findings suggest that these may have been more commonly absent among cases that were enrolled for prosecution. The original J88 forms part of the evidence that is admitted to court and detectives make copies that are kept in the docket. It is possible that the missing copies of J88 forms were due to the original having been attached to the charge sheets used by the courts and copies not having been made or other docket filing discrepancies.

Data were entered into EpiData and imported into Stata 15 for cleaning and analysis. We used svy commands which took account of the sampling design with cases clustered within police stations, stratified by province. The data were self-weighting. First, we described the percentage of cases at each stage in the CJS i.e. the police investigation, prosecutor enrolment and trial stages. We then describe the social and demographic characteristics of victims and perpetrators, and documented circumstances of the rape and aspects of the investigation by whether the stage was reached or not. We compared the proportion of cases with the characteristic that did or did not reach the stage using Chi squared tests and assessed P values. We built four logistic regression models to determine factors associated with attrition at four main points i.e. perpetrator not arrested and/or charged in court, police closure without referral for prosecution, prosecutor declines to prosecute and convictions at trial of being guilty of rape. The multivariable logistic regression models of factors associated with the perpetrator not arrested and/or charged in court and that with the outcome of case closure at the police investigation stages both included all reported cases and the independent variables tested were victim, perpetrator characteristics and different investigative tasks. The multivariable logistic regression model of factors associated with prosecutors declining prosecution was conducted with cases that were referred by police (i.e. all cases not closed by police), and independent variables were victim and perpetrator characteristics, the different investigative tasks and the availability of statements and forensic evidence. The last regression model focused on factors associated with a guilty verdict among those cases that went on trial. Explanatory factors included in the model were victim and perpetrator characteristics, the different investigative tasks and the availability of statement and forensic evidence. In each multivariable logistic regression model, we included only independent variables that were associated with the outcome with P values < 0.2 in bivariable analyses.

Ethical considerations

The South African Medical Research Council’s Ethics Committee (EC020-11/2013) approved the study and permission to access dockets was granted by National and Provincial Commissioners of the South African Police Service and by the National Director of Public Prosecutions. Each selected docket was allocated a random anonymised study ID, and no victim or perpetrator identifying information was recorded on any of the tools. Police stations were assigned station codes, and the linked information was kept securely by the primary researcher so that cases we completely de-identified. Researchers underwent obligatory weekly individual debriefing sessions with a specialist psychologist to minimise impacts of vicarious trauma.

RESULTS

Table 2 shows both the overall and stage specific attrition rates. Three per cent of cases were withdrawn early by victims i.e. before an arrest could be made. The greatest single point of case attrition occurred at the police investigation stage. While the police made arrests or charged perpetrators in only 58 per cent of cases, they referred 65 per cent of all cases for prosecution. Arrest did not occur in these 7 per cent of cases, because the perpetrators were children (under 18 years of age), and the police may have referred other cases where no perpetrators were identified, and arrests were not made. Prosecutors declined 47 per cent of the cases referred by police (i.e. 31 per cent of all cases). About half (46 per cent) of the enrolled cases (i.e. 16 per cent of all cases) were either withdrawn or struck off’ the court roll prior to the trial commencing so that trials commenced in only 54 per cent of all enrolled cases (i.e. 19 per cent of all reported cases). Perpetrators pleaded guilty in 11 per cent of cases taken to trial (i.e. 2 per cent of all cases). About half (53 per cent) of cases taken to trial (i.e. 10 per cent of all cases) were concluded with acquittals and 47 per cent of cases proceeding to trial (i.e. 9 per cent of all cases) were concluded with guilty sexual offence verdict. Seven cases in which there was a sexual offence charge and a second charge of a non-sexual offence (e.g. robbery), were finalised with a conviction for a non-sexual offence. The attrition rates reported in this study were comparable with the Tracking Justice study which used similar methods but was conducted using rape cases opened in 2003 in the single province of Gauteng (Vetten et al. 2008).

Tables 3 shows factors associated with non-arrests and prosecutors declining cases from bivariate logistic regressions analysis Victim age, perpetrator age, victim race, perpetrator relationship, number of perpetrators, victim resistance, victim injuries, tactics used by the perpetrator, time to reporting and conducting investigative tasks were associated with non-arrests.

Factors associated with police case closure without referral to prosecutors in bivariate regression were victim age, victim sex, victim disability, victim race, perpetrator relationship, number of perpetrators, tactics used by the perpetrator, time to report, completion of investigative tasks, victim injuries and SAECK completion. Sex of the victim, victim disability, victim race, perpetrator relationship, tactics used by the perpetrator, completion of investigative tasks, victim injuries and DNA isolation were significantly associated with the prosecutorial decisions to decline cases in bivariate analyses. Tactics used by the perpetrator, victim race and visiting the crime scene were significant associated with guilty verdicts in bivariate analyses.

The first multivariate logistic regression model (Model 1 in Table 4) shows the factors associated with the perpetrator not being arrested compared to being arrested. The likelihood of police not arresting was higher with increasing age of the perpetrators. Perpetrators involved in the cases of White, Indian and ‘Other’ victims were more likely to be not arrested compared to perpetrators involved in cases of with Black African victims. Perpetrators who were intimate partners (aOR 1.41 [95% CI: 1.02–1.94]), acquaintances (aOR 1.44 [95% CI: 1.04–2.00]), strangers (aOR 12.81 [95% CI: 9.40–17.44]) and the ‘other category’ perpetrators (aOR 6.70 [95% CI: 4.56–10.12]) were more likely to not be arrested compared to cases that involved perpetrators who were family members. Perpetrators involved in multiple-perpetrator incidents were more likely not to be arrested compared to perpetrators involved in single perpetrator incidents (aOR 1.59 [95% CI:1.30–1.95]). It was less likely that perpetrators were not arrested when there was evidence of victim resistance compared to where there was none (aOR 0.62 [95% CI: 0.53–0.72]). The accused were less likely to not be arrested when investigative tasks were undertaken including visiting the crime scene (aOR 0.49 [95% CI: 0.42–0.58]), taking first report statements (aOR 0.5 [95% CI: 0.48–0.73]) and SAECKs had been done on victims (aOR 0.5 [95% CI: 0.36–0.64]).