Protest injuries A situational analysis of injurious protests in Gauteng South African

In this article, we investigate contextual and situational circumstances of protest events that record injurious outcomes for civilians and examine how these differ from protests which do not record such outcomes. Using the IRIS database, we examine how contextual factors, including protest period, protest location, reason for protest, and situational factors, such as type of protest, damage to property, arrests and police response contribute to civilian injury. Using logistic regression analysis, it was found that: 1) protest-related injuries were more frequent during the late-2000s than the 2010–15 period; 2) protest location was not a significant predictor of protest injury; 3) protests which recorded arrests and damage to property were more likely to report injurious outcomes; and 4) the addition of an aggressive police response was significant in determining protestor injury outcomes. Our findings have implications for public policing strategies, highlighting the role of different modalities of police response in the mitigation or escalation of injuries at protest events.


Introduction
South Africa has been described as a "protest nation" and the "protest capital of the world", with research indicating that protest action is increasing in frequency. 2 Although protest action in South Africa is mostly non-violent, protest events are sometimes accompanied by Pascal Richardson, Lu-Anne Swart, Rajen Govender and Mohamed Seedat 1 richap@unisa.ac.za swartl@unisa.ac.za govender@iafrica.com seedama@unisa.ac.za https://doi.org/10.17159/2413-3108/2022/vn71a12891 violence, which threaten the safety and wellbeing of civilians. 3 A growing body of research on protests in South Africa has offered explanations on the magnitude, geography, contexts, and drivers of protests. 4 Existing research offers macrolevel factors (including broader socio-economic and political factors), and individual-level factors (such as protestor characteristics and motivations) to explain the occurrence of (non) violent protest events. Fewer studies have looked at specific characteristics of protest events (such as the type of protest, specific activities at the protest and the police's response to the crowd), which appear to be an important set of factors to consider when understanding violence (and injuries) at protests. 5 This study expands on the literature by examining the characteristics of protest events, including situational context and dynamics.

Defining violence at protests
Overall, the research evidence on violence at protests in South Africa remains equivocal. Alexander, Runciman and Maruping have identified that the vast majority of crowd incidents in South Africa were peaceful, while only one in ten were classified as 'unrest'. 6 This is contrasted with results from Powell, O'Donovan and De Visser, who assert that protests are becoming increasingly violent, reporting that up to 80 percent of protests in 2014 involved some form of violence. 7 Critical here are the distinctions in conceptualisation of protest action and classifications of violence. Operational definitions of violence at protests have variously been described to include protestor intentions, protestor actions, protestor consequences, police perceptions, and police responses, or a combination of these factors. As a consequence, protests have been variously classified as 'peaceful' and 'violent', with others instead introducing more nuanced notions of 'disorganised' and 'disruptive'. 8 Such differences in conceptual and operational definitions of violence and protests are problematic when synthesising literature on the topic, as research findings vary greatly, depending on how violence is understood.
In this article, we use the word "violence" to describe any injury outcomes to a person, whether the injury was intentional or accidental. 9 It is important to note that while not all injuries are a direct consequence of direct violence, they may nevertheless indicate the presence of disorganisation at a protest event, with such disorganisation being a higher risk for potential injuries. Injuries may occur through an ordinary object (e.g., rocks), a defensive or harmful device (e.g., tonfa, firearm) or physical actions (e.g., beatings, punches). We recognise that using injury as a proxy for violence is problematic as we cannot identify the cause of each injury. However, injuries during protest action are attributed to the protest event as a whole, therefore there is merit in discounting the need to recognise specific causes of each injury or to have this serve as a limitation to analysis of such injury.

Characteristics of protest events
In this article, we examine protests by focusing specifically on the characteristics that typify the protest event. Available literature has demonstrated that an investigation into the characteristics of protest events, such as the situational context and dynamics, may be useful in predicting future protest-related injury, and therefore may be used to prevent injurious protest outcomes. 10 The following section will conceptualise and highlight these factors, indicating how they are understood to interact with and contribute to the occurrence of violence at protest events.

Situational context
Research has demonstrated that explanations of violence at protests should include a close examination of the contextual factors within which protests are located. Situational contextual factors include the geo-spatial and temporal period of a protest event, and the reason for the protest. 11 Protest action typically occurs in overcrowded and under-resourced communities as they often experience marginalisation from socioeconomic services. 12 Urban areas are therefore particularly vulnerable to protest action due to a rapid influx of people leaving peri-urban and rural communities in search of employment opportunities. In South Africa, protests are mostly located in historically impoverished and marginalised communities in urban and peri-urban areas and are a manifestation of a dissatisfaction with the status quo, typically relating to service delivery (including access to electricity, sanitation, housing, and education), and labour-related grievances. 13 These longstanding grievances, culminating in protest action, see protestors become emotionally driven, fuelled by anger and frustration in search of an opportunity to assert their rights to inclusion and dignity. 14 Although our focus is on non-violent protests, it is important to note that rage has been widely acknowledged as one of the driving forces of protest violence. 15 Protest action may also turn violent when the local authorities are regarded as being inept or indifferent. 16 This is a sentiment raised by participants in a South African study, which highlighted that communities resort to protest action due to "unfulfilled promises" and a lack of "openness and transparency" between the people and the government. 17 Protest participation is constituted by petitions for basic socio-economic and political transformation and participation, including demands by marginalised communities for inclusion in decision-making processes and structures. 18 While contextual factors are necessary to explain why protest action occurs, it does not sufficiently account for why some protest events are accompanied by violence, while others are not. 19 For this, we look further to situational characteristics for protest action.

Situational dynamics of protest events
Research has highlighted the importance of including situational characteristics of protest action when examining how and why violence manifests at specific protest events. 20 Examining situational characteristics of protests typically entails an analysis of events and behaviours that are specific to a given protest, and which are critical to understanding how the protest event develops. 21 Situational factors are multiple and include, but are not limited to, the type of protest, damage to property, arrests, and specific policing strategies or behaviours which may indicate that the protest event may turn violent. 22

Type of protest
Protest action can be enacted through various strategies, ranging from forms that are typically regarded as being legitimate, such as petitions, demonstrations, marches, and boycotts, to a host of activities, which are often denounced by ruling institutions, such as barricades and riots. Similarly, protests can be divided into those that are protected, and those which are not. Although the South African Constitution affords citizens the right to mobilise, applications for protected protest action are often rejected, leaving communities with no alternative other than participating in an unprotected protest. Such events are more likely to turn violent, as they are considered disorganised or unplanned. 23 This increases the likelihood of injury, as a general sense of disorientation and confusion combined with frustration creates fertile ground for violence.

Damage to property
Previous research has demonstrated damage to property to be an important factor in explaining protestor violence and injurious outcomes. 24 This has been described by Nassauer as resulting from two related factors: first, police tend to regard destruction of property as a precursor to threats to human safety therefore justifying police force; secondly, destruction of property creates intragroup tensions between those who refrain from damaging property, and those who do not. These factors, along with an uncertainty, experienced by both the authorities and other protestors, facilitate a sense of unease and threat to self, which precipitates violent outbursts. 25 Arrests Closely related to police use of force, is their reliance on forceful mass arrest or detention as a strategy through which to manage unrestful crowd incidents. 26 Here, arrests are more likely to occur when authorities identify an incident as violent, which has an increased likelihood if the incident is also perceived as disorganised. This creates a sense of threat from the unpredictable nature of protest participants, resulting in an increase in forceful police intervention. 27

Police response to protests
Existing literature referenced throughout this paper highlights the impact of the police response to a protest incident on the protesters and onlookers. 28 In South Africa, protest events are attended by various groups of authority, including station-level police, metro police and private security forces. The South African Police Service (SAPS) is responsible for all crowd-related incidents, having an established specialised taskforce, the Public Order Policing (POP) units, which are primarily tasked with managing such incidents. These units have specific policing actions available to assist with protest management, including setting up official barricades, extinguishing fires, and using various strategies to disperse crowds (such as rubber bullets, teargas and watercannons). POP units have undergone considerable changes since their inception, and successive restructuring processes have been motivated and directed by several factors of influence, specifically; new mandates, varied deployment strategies, incident management strategies, professional training and changes in manpower. 29 Actions of POP units have come under considerable scrutiny in recent years following widely publicised examples of excessive use of force. 30 Although most protests are not marked by injury, escalating tensions between community members and the authorities sometimes result in civilian injuries. International research has demonstrated that police are more likely to use force when they perceive themselves to be hopelessly outnumbered, therefore resorting to varied strategies to regain authority. 31 Furthermore, it has been found that police officers often have negative perceptions of group behaviour, regarding crowds as "inherently irrational and dangerous". 32 Similarly, in South Africa research has shown that officials harbour preconceived notions of the 'inherent nature' of violence at protests, with Brooks finding that some POP officers held notions that protests are a guise for other acts of criminality -therefore justifying early and unwarranted use of force. 33 An increase in protest frequency in South Africa, and the resultant increase in potential risk for injury, warrants further investigation. This research addresses a paucity in South African literature by investigating injurious outcomes in protest events. The current study specifically aims to understand how situational contextual factors (including protest timeframe, protest location, and protest reason) and situational dynamics (including protest type, damage to property, arrests, and other police responses) are associated with injurious outcomes to civilians. Therefore, the goal of this paper is to examine the protest characteristics that contribute to injury at protests, with the intention of contributing to the growing body of protest work in South Africa, with a view to applying this knowledge by way of recommendations for injury prevention in future protests.

Data
To investigate the contribution of contextual and situational factors to injurious outcomes in protests in South Africa, we draw our data from the Incident Registration Information System (IRIS) database. The IRIS database is perhaps the most comprehensive database on crowdincidents in South Africa, capturing information on various types of incidents, including protest action and social gatherings. 34 The IRIS database was accessed through a Promotion of Access to Information Act (2 of 2000) (PAIA) application, approved by the SAPS Research Division. 35 As received from SAPS, the IRIS dataset listed 25 607 crowd incidents in Gauteng for the period 1 January 2005 and 31 December 2015. The database records all crowd incidents that had some form of POP intervention. As the focus of this paper is on protest action, incidents which were not considered to be protest action, such as funerals, sporting events and other recreational and cultural events, were removed from the dataset. This was accomplished by analysing the accompanying explanatory notes for relevant variables. 36 By systematically analysing these data entries the team identified 12 004 protest incidents. To rationalise the dataset for analysis, cases with missing data were identified and removed. Cases were excluded if they contained missing values on key variables. An analysis of excluded cases indicated that cases with missing values were randomly distributed across the key variables. Accordingly, the excluded cases were not deemed to be qualitatively different from those cases retained for the analyses. The final analysis dataset comprised of 8 888 protest incidents.

Variables
The dependent variable was constructed by examining whether any injury was recorded for a given protest incident. 37 As data was not available on the actual number of injuries for an incident, it was not possible to record the frequency of occurrence of injury. The outcome variable was thus a dichotomous variable, coded as "no injury" and "injury". The outcome variable was constituted of all forms of injury recorded on the IRIS, including accidental and intentional blunt and sharp force injuries, firearm-related injuries, beatings, burns, explosions and rape. Note that recorded injuries in this dataset are not solely attributable to police action and may therefore also represent accidental injury sustained during the protest. In line with the research aims, during analyses, "no injuries" was used as a reference category.
Several explanatory variables were created based on existing variables and incident notes captured on the database by POP officials. The construction of these variables was informed by literature on contextual and situational factors of protests. "No damage to property" was used as a reference category during analyses.

Arrests
A variable for arrests was constructed by analysing recorded arrests for each protest incident; a decision was made to transform this variable into a dichotomous variable, "arrests made" and "no arrests made". We used "no arrests made" as the reference category, as it is indicative of police intervention at the protest event.

Police response
Police response recorded the type of interventions employed by the police to manage the crowd incident. This was coded with three categorical values: "only non-aggressive", "both non-aggressive and aggressive", and "only aggressive". Non-aggressive police action included attending the scene, investigating the incident, negotiations and extinguishing fires. In contrast, aggressive action included using rubber bullets, teargas, or water cannons to disperse crowds. We used "only non-aggressive" as the reference category.

Data analysis
Frequency analyses were used to examine and describe the frequency of protest-related injuries. Logistic regression analysis was conducted to examine how the various protest circumstances differentiated the two categories of the outcome variable. IBM SPSS Statistics for Windows (version 27.0) was used for all analyses. All statistical tests' results were assessed at a significance level of p=0.05 and all regression parameter estimates were assessed using the 95% confidence interval (CI).

Ethics
This research is part of a broader study on protests in South Africa, which has received ethical clearance from the University of South Africa's College of Graduate Studies. 40 Table 1 displays the descriptive statistics for the variables analysed in this study. The vast majority of the 8 888 protest events analysed in this study did not report any injury outcomes, with injuries being recorded in 125 incidents (1.4%). We observed a difference in protest incidence across the two defined time periods, with almost two-thirds of recorded protests occurring between 2010 and 2015 (66.4%). Most recorded protests occurred within a metro municipality (86%) with more than a third of incidents occurring in Tshwane (35%), followed by Johannesburg (34.2%), and Ekurhuleni (16.8%). The remaining protest incidents (14%) occurred in non-metro areas of the province.

Logistic regression analysis
Logistic regression analysis was performed to measure the impact of situational context and situational dynamics, independently and concurrently, in differentiating the risk for protests with injurious outcomes to civilians. The logistic regression modelling was conducted sequentially, first by testing the impact of three variables measuring situational context (year of protest, location of protest and reason for protest), and thereafter by entering variables measuring situational dynamics (type of protest, damage to property, arrests made and nature of police response). The results of the sequential logistic regression modelling are presented in Table 2. All reported odds ratio values have been adjusted for other variables in the model.
Situational context and situational dynamics were analysed in model 1 and model 2, respectively. Results indicate that, when analysing injurious outcomes by examining situational context alone (model 1), 'year' and 'reason for protest' were statistically significant predictors of injurious outcomes. With regard to reason of protest, 'service delivery', 'labour related', and 'student protests' were more likely to result in injurious outcomes as compared to 'socio-political protest' events. In model 2, assessing situational dynamics, 'damage', 'arrests' and 'police response' were statistically significant, while 'protest type' was not statistically significant.
Analysing the situational context and situational dynamics together, the -2 Log Likelihood test indicates that model 3 was statistically significant at predicting injurious protest outcomes (χ 2 = 378.356, p = 0.000). Therefore, this specific model fit the data, which is supported by the estimate of the Nagelkerke pseudo-R 2 at 0.317.  Reference category is non-injurious outcomes 1 Adjusted odds ratio 2 BCa 95% confidence interval for AOR ** p ≤ 0.01; * p ≤ 0.05 aggressive and non-aggressive responses were up to seven times more injurious than events where only non-aggressive strategies were employed (AOR = 7.990, 95%, CI = 4.628-13.799).

Discussion
In this study we examined how situational contexts and dynamics of protest action is associated with injurious outcomes in Gauteng. Analyses revealed that situational explanations of violence at protests contribute meaningfully to explaining why injury occurs at some protests but not others. Protests with injuries were more likely to have occurred between 2005-2009 than in the years thereafter. Protests in Johannesburg metro were associated with greater probability for injurious outcomes than other metro and non-metro protests in the province, however, this difference was not statistically significant. Protest reason and protest type did not significantly influence injurious outcomes.
Protests where damage to property, arrests and an aggressive and non-aggressive police response style was recorded were more likely to have injurious outcomes for civilians.

Situational context of protest injury in South Africa
We found that year of protest was significantly associated with protest injuries, with protest action being less injurious between 2009-2015, compared to the timeframe preceding it, when controlling for other variables. This may be reflective of both the POP mandates changing from being orientated to crime prevention to crowd-management in the late-2000s, 41 and a general increase in the number of POP units. For example, decreases in recorded incidents between 2006 and 2009 coincide with governance restructuring, which drastically reduced the number of POP units, and correspondingly, protest incidents appear to rise again as POP units increased in anticipation of the FIFA World Cup in 2010. 42 Changes in unit mandates and associated crowd-management strategies, particularly leading up to the World Cup, and again after legislative reform and unit restructuring following several high-profile incidences involving POP units and accompanying forces, may account for fluctuations in reported and attended protest events. 43 In line with findings elsewhere, the majority of protest action was located in metro areas. 44 This may be explained by rapid expansion leading to a growing number of marginalised and underresourced communities that engage in protest action as a form of political participation to address grievances with service (non)delivery and a lack of employment opportunities. 45 The frequency of such protests is representative of chronic dissatisfaction with local government, creating heightened tensions, which occasionally result in injurious outcomes.
Contrary to expectation, protest reason did not bear a significant influence on the injurious nature of the protest. We expected service deliveryrelated protests to record more injury than other protest causes, as these are often manifestations of longstanding frustrations, prolonged disenfranchisement, and have spontaneous origins. Based on previous literature, these factors should create a space where injury is more common, as the emotionality, disorganisation and unplanned nature of such protests should make them vulnerable to aggressive police responses, and disorderly crowd conduct. Interesting here is the injurious nature of student protests -what would typically be considered organised, with strong group ties. The injurious nature of student protests may be explained by the accompanying police response, which serves to protect educational infrastructure, thus eliciting strong police responses in a situation that may otherwise be left unchecked.

Situational dynamics of protest (non)violence
In our analyses, the type of protest only bears a significant influence on the injurious nature of the protest, if it was a strike. In our analyses, strikes were more likely to result in injurious outcomes than other forms of protest. Due to inconsistency of data captured, IRIS did not allow for analysis of group size, however other research shows that strikes are often attended by larger numbers of people. Additionally, it has been suggested that labour strikes have a differed organisation than other forms of protest, with worker committees using violence as a means to prevent other employees from dissenting from the cause. In such situations, violence is used to prevent worker fragmentation and therefore enhance solidarity among staff, thus strengthening the group's position of power over the employer. 46 We found that protest events that were accompanied by damage to property are more likely to record injuries among civilians, which is in line with findings reported elsewhere. 47 This is attributable to several reasons. Protests characterised by longstanding grievances are more likely to turn violent as tensions heighten, and anger and frustration is reified as destruction of property. This in turn elicits defensive and aggressive responses from policing units, resulting in the use of varied crowd-dispersal strategies, which may cause injury to civilians. 48 In line with findings elsewhere, 49 we found that injurious protests were also associated with those that recorded arrests. This is congruent with the notion that protests which are perceived to be violent are met with specific types of police action, aimed at minimising continued violence, such as mass arrests.
Our analyses revealed that aggressive and nonaggressive force action was significantly more likely to be accompanied by injury than other forms of police action, including action coded as 'only aggressive'. This may be explained by the sample size rather than a true difference, as 'only aggressive' made up less than 1% of police responses. The addition of 'aggressive' police responses in causing injury is intuitive, as this response category comprises of dispersal strategies that intend to cause discomfort at the minimum, and harm at the most, such as tear gas, water cannons, K9 units and rubber bullets.
The deployment of such measures is often accompanied by disorganisation, increasing the likelihood of injury.
Previous research has shown that police responses are impacted by perceived threat, thus aggressive actions (such as employing stun grenades, tear gas and rubber bullets) are more likely to occur when authorities feel overwhelmed and outnumbered, 50 which explains the high levels of injury during the years when there was a reduction in POP units' manpower.

Limitations
First, the study is limited by the database that was analysed. IRIS has been subject to substantial critiques relating to a lack of clear protocols when capturing data, as well as a lack of verification and auditing processes. 51 As a result, a considerable portion of the dataset received had to be excluded from analyses due to missing data. Similarly, the database has been criticised for under-reporting peaceful protests, and under-reporting injuries among civilians. 52 Secondly, due to the nature of the database, we are unable to identify the sequence of events as they unfold at the protest events. The sequence of events has been previously identified to be important in understanding specific triggers and actions that precipitate protest violence. 53 Likewise, our analyses were limited by its focus on the specific contextual and situational variables present in the IRIS database. Other situational contributing factors, such as individual protestor characteristics, emotions experienced at protest events, the number of POP officials dispatched to each protest, and POP officials' personal appraisals of protestor threat and safety would be worth exploring in future research.
Thirdly, the small number of incidences marking injurious outcomes means that it is more challenging to achieve statistical significance. This may have affected predictive variables, which yielded no statistically significant results.
Despite these limitations, our study suggests that situational characteristics play an important role in identifying protests that result in injurious outcomes, from those which do not.

Conclusion
In South Africa, protest action remains a key issue in public debate and discourse. In this paper, we examined how situational context and situational dynamics contribute to, or inhibit, violence at protests and by implication, civilian protest-related injuries. Overall, these situational explanations should be regarded as one factor in a complex system of independent yet interlinked factors that contribute to the escalation of violence.
The complex nature of protest action and the personal consequences it bares for protestors and bystanders warrant further investigation. More nuanced investigations into how protest action results in injuries are needed. These may help pinpoint specific actions that lead to escalating violence, for which preventative measures could be implemented. In addition, future research should replicate this study to include data from other provinces.