Unpacking discontent Where and why protest happens in South Africa

A series of protests, demonstrations, strikes and political violence can, if conditions are right, gradually turn into social unrest. Social unrest can

participation (such as democratic processes) can, over time, transform into unconventional political participation like violent protest or political violence. What is considered conventional across the world not only depends on the period in time in which it takes place, but also on the geographic location, and that particular society's definition of what is socially acceptable. Perceptions of what is legitimate political action can change, either as satisfaction with the state changes or as a growing number of citizens believe that peaceful protest or voting in elections are ineffective at raising their issues and achieving their goals. 2 Recent research by Bohler-Muller et al. 3 into attitudes towards different forms of protest in South Africa suggests that disruptive and even violent protest may be becoming more acceptable, given that a growing number of South Africans believe these forms of protest yield more successful results than peaceful protest action. According to this research, perceptions are also linked to when and how authorities react to various forms of protest. 4 Existing research also illustrates that social unrest can build up over time. 5

Protest and Public Violence Monitoring project
In 2013, the ISS developed a database of crowd-based events that aimed to monitor the frequency, location, nature and extent of collective action activities taking place in South Africa. 8 The database also seeks to track all forms of peaceful protest, crowdbased disruptions and violence occurring in public places. The PPVM collects information on community protest events of various kinds: protests against municipalities and other public sector services, industrial strike action, protests against crime, protests against private sector practices, and party-political protests.
The system also monitors forms of violence associated with protest action as well as vigilantism, xenophobia, political conflict and other forms of crowd or inter-group activity.
Collecting better information on protest and related trends and on the nature and extent of this complex phenomena has the potential to contribute to an improved understanding of the nature, scale and patterns of collective action. This in turn can foster the development of more appropriate responses by state and non-state role players, most notably the police, municipalities, community and political leaders, and protest organisers, 9 because they are better able to take into account current political and class struggles.

Methodology
The PPVM's data is collected by scrutinising reports from more than 100 local, national and international online news sources as well as newsletters and notices by trade unions, political parties, and universities.
Incidents that are captured by the database include industrial strike action, vigilantism and political attacks -activities which normally fall outside the scope of protest analysis. However, the broad scope employed by the PPVM allows for an interesting analysis of the scale of collective action. This scope is aligned with the definition of social unrest by Schroeter et al., which includes events linked not only to civil protest but also to strikes and political and civil violence. 10 The data capturing instrument collects information on more than 30 grievance types. For the purpose of this analysis, these grievances have been categorised into 13 broad categories, which span the spectrum from private sector services, corruption, housing/ land and transport, to xenophobia. The working definitions for each category are available in Appendix 1.
Relying on media reports alone poses substantial limitations. Firstly, not all incidents of public violence are reported in the media. Analysis by the University of Johannesburg estimates that only one in four events are reported in the media. 11 Moreover, media reports tend to contain few facts about the exact nature of the event and its causes, and the extent to which there was violence or not. For instance, community-level protests are often reported as 'service delivery' protests but the exact grievances (that is, whether the protest is over water, electricity, housing, etc.) are not specified. Furthermore, reports may not focus on the primary grievances that led to a protest, but may highlight only the escalation into violence. For example, reports will cover attacks on businesses owned by foreign nationals during protest action, but not the original protest or, in some cases, the fact that local businesses were also targeted. 12 As media organisations are typically based in metros and large cities, the coverage of events in those areas will be greater than in smaller towns and rural areas. In addition, coverage may differ from province to province, for the same reason. 13 Lastly, media reports in various languages are not necessarily available online. 14 The PPVM treats events as violent when they are described as such in the media and if, based on the reported facts, an incident may contravene the provisions of the Regulation of Gatherings Act and involve a criminal act of violence. The definition of protest also allows for the inclusion of events where violence is initiated not only by protestors but also by other parties such as law enforcement agencies or other groups.
The PPVM provides an open source, virtually real-time, geo-referenced record of a comprehensive list of crowd or collective actions that have the potential to escalate into disruptions or violence. An interactive map of all events is displayed online on the ISS's Crime and Justice Information and Analysis Hub at https://issafrica.org/crimehub/.
The PPVM is by no means the only protest database available. The section below examines several other data collection efforts aimed at measuring and understanding protest, and discusses how these differ from the PPVM.
How do other measures of protest differ?
An increase in protests in recent years has sparked new research interest into the topic. Yet, the currently available sources of information vary greatly because of different data collection methodologies, counting rules, scope and variance in the definitions that they apply.
Arguably the most comprehensive source of protest or crowd gathering data in South Africa is that recorded by the South African Police Service's (SAPS) Incident Registration Information System (IRIS). This system is designed to capture crowd management activities and interventions by the Public Order Policing Units (POP units) during all types of crowd events.
A summary of the latest IRIS data is contained in the SAPS annual report. The 2016/17 annual report noted that the SAPS monitored 14 693 'crowd-related incidents' (including sports, recreational, religious and cultural events) between April 2016 and March 2017. 15 Threequarters of these events (n=10 978) were described as 'peaceful-related' incidents with the remaining 3 715 events (25%) termed as 'unrest-related' incidents. 16 This represents a 10% increase in the percentage of 'unrestrelated' incidents since 2013/14.
The Centre for Social Change (CSC) at the University of Johannesburg accessed the IRIS data for the period 1997 to 2013 through a Promotion of Access to Information Act (PAIA) application. 17 After an in-depth analysis of the IRIS data, it concluded that these 'unrestrelated' incidents are not necessarily all violent but are labelled as 'unrest-related' because they required 'interventions' by the POP units. These interventions included actions such as directing or dispersing a crowd through the use of various crowd control techniques, including (often controversial) non-lethal crowd management equipment such as tear gas, stun grenades, rubber bullets and water cannons, or making arrests. 18 Any crowd event not requiring intervention is regarded as peaceful. 19 The 10% increase in the number of 'unrestrelated' incidents over the four years since 2013/14 may well require a more nuanced analysis than merely equating this figure to an increase in the frequency of protest. This rise may rather be the result of the re-establishment of more POP units and the acquisition of more equipment in recent years. For example, by 2015/16, the SAPS had 28 POP units consisting of 4 617 operational members and support personnel. In 2016/17, the number increased by almost 9% in a single year to 5 025 members. 20 This represents a 94% increase from the 2 595 members and 23 units in 2006. 21 Following the closing of many POP units by 2006, the IRIS data showed significant decreases in recorded incidents, attributed to the decreased capacity of the units. 22 The increase in units since 2013 has led to a greater capacity for (and therefore probability of) intervention by the SAPS, which, coupled with improved IRIS record keeping, has increased the frequency of protests on the IRIS database. The increase is therefore not merely a reflection of an increase in the actual number of public assemblies or protests. 23 IRIS data entry practices remain fairly arbitrary and uneven due to the absence of enforced uniform protocols, definitions and categorisations. 24 Furthermore, transparency in data collection practices is limited because the SAPS only releases the details of the events in instances where PAIA applications are made.
Besides IRIS there are several research organisations that capture protest action through media reporting. For ease of comparison, Table 1 provides a summary of the scope of events covered by these data collection efforts. The table also shows the reported frequency of events estimated by each database over the last six years (2012 to 2017).
There are clear differences in the data. The CSC (in a separate study to its in-depth analysis of the SAPS's IRIS data) collected 3 526 media-reported community protests (MRCPs) between 2005 and 2017, including all types of community protests and not only service delivery protests aimed at municipalities. This database showed 375 incidents for 2017. 25 In contrast, Municipal IQ, a web-based data and intelligence service, focuses on municipallevel service delivery protests and publishes the Municipal Hotspots Monitor. Since 2004, the Monitor has collected data on 'major protests staged by community members (who can be identified as living in a particular ward) against a municipality, as recorded by the media (or other public domain sources such as SAPS press releases)'. 26 Because of the narrow scope, the frequency of events is lower than that captured by the CSC and PPVM. In 2014 the Monitor recorded 191 protests, which decreased to 164 and 137 protests in 2015 and 2016 respectively. 27 It recorded a slight increase in 2017, to 152 protests. 28 The Civic Protest Barometer has also monitored trends relating to protest action at a municipal level since 2007. 29 The Barometer is a project of the Dullah Omar Institute at the University of the Western Cape. 30  incidents. 31 However, overall, it records slightly fewer incidents than Municipal IQ.
With the exception of the IRIS database and the PPVM, none of the local data recording efforts covers a broad definition of protests or incidents of public violence. These organisations measure either community-level protests or those targeting municipalities. Jane Duncan notes that 'the inherent assumption that community protests are largely aimed at local government failure, ignores the service delivery complexities and varying service delivery mandates of different spheres of government'. 32 International data collection efforts also maintain a wide scope. One example is the Armed Conflict Location and Event Data Project (ACLED). ACLED describes its initiative as 'a disaggregated conflict collection, analysis and crisis mapping project. ACLED collects the  Similar to the PPVM, the ACLED database covers a wide range of protest action as well as forms of political mobilisation and violence (including police and other state action against civilians or battles for territory, political rallies and other political assemblies) since 1997. 35 However, ACLED differs from the PPVM and most local databases in the way it counts a single event. According to ACLED's counting rules, protracted events are counted multiple times in the database, as each day is counted as a single event for as long as each event continues. So, for example, a three-day #FeesMustFall protest action will be counted as three entries. For this reason, the number of events captured by ACLED is higher than that recorded in the PPVM. In 2017, it recorded 1 026 events of which 935 were described as riots or protests. 36 A growing number of total protest days may well be an indicator of growing social unrest. However, the main aim of the PPVM is to measure the number and nature of distinct geographic events, rather than to produce an estimate of the number of total protest days alone.
Comparatively, the publicly available international data and IRIS cover a wider range of incidents than other databases, but the counting rules vary greatly from other local efforts. The PPVM covers a wide spectrum of events and has similar counting rules to local databases. Except for ACLED and the PPVM, none of the other databases offers freely available data, has a sufficiently wide scope, or provides the nuanced detail required to undertake an in-depth analysis on the location of protests, the number of protests over time in similar locations, or the socio-economic grievances leading to protests. 37 Furthermore, there is also no agreement across the databases on what actions should be regarded as violent, disruptive or peaceful. Further analysis shows that these decreases at the national level were driven mainly by decreases in the levels of protest in Gauteng. This may be due to government interventions in place at the time that may have mitigated the impetus for protest, such as the formation of task teams to address community grievances. One such initiative, the Ntirhisano community outreach programme in Gauteng, aimed to foster trust and confidence in government by improving service delivery and accountability. 38 To date, the success of this programme is unclear and further research is required to understand the effectiveness of this programme in terms of its implementation and impact.

Key findings of the PPVM
Between 2016 and 2017 the frequency of protest events increased by 42%. This may be related to the change in government in two of the main metropolitan areas in Gauteng after the 2016 elections, which may have reversed the public's positive view of the initiatives described above, as the effectiveness of these initiatives became strained due to a change in dynamics between the various spheres of government.

Seasonality of protest
The seasonality of protest is clearly depicted in the monthly trends presented in Figure 2. The number of incidents recorded in the first six months of 2017 surpassed the total number of all incidents in 2015. Thereafter, the numbers decreased quite substantially in the second half of the year. Traditionally, events peak around May to September each year and start dropping after October. December typically has the fewest incidents.
There are a number of possible explanations for these patterns. May is widely regarded as the start of 'strike season', as new salary negotiation processes commence across the country, during which wage disputes are declared between employers and their workers. Some of these processes lead to protracted negotiations and sporadic strikes across various industries, which continue for much of the South African winter months. 39 The data shows that all types of protest peak during this period, not only labour-related industrial strike action. The winter months, which bring with them cold weather and increased vulnerability for many South Africans,  The marked drop in the December period coincides with the closure of industries and schools in South Africa between mid-December and mid-January for the so-called 'festive season', when many potential grievance makers and service delivery agents are on leave and travel for the holidays.

Types of protest
The PPVM classifies each incident of protest action according to the crowd's main grievance, as identified in the media or other online source. 40 A breakdown of these main grievance categories is contained in Figure 3.
Analysis of the data shows that most collective action between 2013 and 2017 has not been focused on highlighting the problems with municipal service delivery, but was more often related to industrial strike action (comprising 19% of all events.) The second most recorded grievance category (accounting for 16% of incidents) consisted of grievances about the police's inability to reduce crime levels or to solve particular crimes. 41 Protests around municipal services made up the third most recorded grievance category, contributing a further 16% of protest incidents.
Twelve per cent of the protests recorded by PPVM were education-related, which included grievances concerning basic as well as tertiary education (including #FeesMustFall).
A further 11% of the protests or crowd events related to politics or elections. Other prominent protest concerns related, among others, to housing and/or land, transport, other socioeconomic rights, xenophobia and corruption, and unfair business practices.
Further analysis of the data shows that protest incidents are frequently organised to express dissatisfaction with more than one grievance issue. For example, municipal service delivery protests may be focused on water and electricity delivery, and may raise problems regarding either terms of quality or quantity of services, or both.
The findings about overall grievance type are important because they illustrate clearly that most grievances do not relate to municipal service delivery specifically, but are rather an indictment of government services in general.
Of particular concern is the large proportion of protests against the police and/or crime, and against the departments of basic and higher education.
Beyond basic municipal service delivery, the majority of protests illustrate the daily struggles of ordinary people to access their constitutionally protected socio-economic rights, such as access to jobs, fair wages, safety, decent education and transport. This includes grievances aimed at the private sector, although these were not nearly as prominent as grievances aimed at the state.
The sheer scale of protest aimed at this wide range of issues provides justification for the need to widen the scope of study to include all types of protest away from a narrow focus on 'service delivery', 'municipal services' and 'community protest'.

Violent protests
More than half (55%) of the incidents captured on the PPVM were termed 'violent' or 'disruptive' in media reports. The proportion of violent or disruptive incidents per year has been increasing year on year from 43% in 2013 to 65% in 2016, as Figure 4 below illustrates.
It is interesting to note, however, that the percentage of violent or disruptive incidents recorded in the first six months of 2017 is lower than in 2015 and 2016.
It is often difficult to clearly establish from media reports whether incidents are violent or simply disruptive. It is also difficult to know when the status of an incident may shift from being peaceful or disruptive into one that is violent. Yet, the distinction between what is regarded as violent or disruptive is important. Alexander et al. introduced this distinction to ensure that the popular narrative moves away from viewing protestors as violent unprovoked agitators, in an effort to be 'more sympathetic to protesters and to the history of protest'. 42 International literature Peaceful Violent/disruptive suggests that the likelihood of an event leading to violence is influenced by several factors, including actions or a lack of action on the part of both protestors and other roleplayers (such as the police or the party the protest is aimed at). Examples include heavy-handed policing practices or a lack of responsiveness by the relevant government department. 43

Location of protests
The events in the PPVM database were analysed by type of location, in other words, whether they took place in metropolitan, urban or rural areas. 44 Two-thirds (67%) of the incidents took place in metro areas, while 17% and 16% took place in urban and rural areas respectively. Metro areas often experience challenges because of rapid urbanisation, higher population density and migration. 45 Exceptional growth of the metros gives rise to increased demand for basic municipal services and other socio-economic rights such as housing, land, healthcare, education, infrastructure, transport, employment and security.
Urbanised provinces are predominantly the location of most protest events. Figure 5 below shows that almost one-third of all events took place in Gauteng (31%), followed by the Western Cape (20%), KwaZulu-Natal (16%) and the Eastern Cape (14%). These provinces are the most highly populated and most urbanised, and therefore contain the largest metro areas. The Northern Cape experienced the lowest percentage of protest and strike action (1%), closely behind the Free State (2%), Mpumalanga (4%), North West (5%) and Limpopo (7%). These proportions remain largely unchanged even where strike action (which favours largely industrialised provinces) is excluded from the analysis to examine whether provincial patterns of non-labour related community protest differ.
Some locations experienced a higher frequency of protest action than others. One of the simplest yet most effective ways to analyse data spatially is through mapping. The frequency of protests at different locations can be measured over time to yield so-called protest 'hotspots'. 46 Hotspot analysis can be done to look at patterns of distribution in protest for a specific moment in time, and can also be used to determine any changes in hotspots over time.
For instance, hotspot maps can show changes year-on-year or month-on-month, or can visualise data for a particular election period. The formation of hotspots illustrates that certain areas experience persistent protest action. Hotspot analysis is therefore a useful planning and monitoring tool to facilitate targeted and sustained interventions in areas where protests occur frequently.  The map below shows the Gauteng protest hotspots that were calculated using PPVM data between January 2013 and March 2018. 47 With geographic information system (GIS) software, protests were clustered within a 2.5km radius to estimate where most protest hotspots are located. The intensity of the hotspots was based on the frequency of protests at locations, using statistical calculation methods such as kernel density estimates. 48

Concluding remarks
A comprehensive understanding of protest action necessitates the study of a wide scope of collective action including, but not limited to, protests aimed at municipalities and other narrow community interests. The PPVM is the only local, publicly accessible information portal that captures all forms of protest, disorder and public violence associated with social and economic discontent. Its findings therefore present an important contribution to what we know about the frequency and nature of protest and discontent.
The findings emerging from the PPVM data illustrate how wide-ranging protest grievances are, and how geographically widespread they are. Contrary to widely held opinion, the majority of grievances resulting in protest action are not about basic municipal-level service delivery issues, but include concerns about safety, education, employment and other broader socio-economic rights. While most protest is aimed at the state, the number of events mounted against the private sector in the form of strikes or business practices is significant.
Preliminary hotspot analysis suggests that the likelihood of escalation of protest action into violence is dependent on the frequency of protests in a place (highlighting the possible lack of resolution of issues) as well as the role played by parties such as the police, leaders or officials. Further research should focus on this.
The PPVM has its share of data challenges, and the project is adapting to address these concerns. For example, the PPVM data sources are being expanded to curb a dependency on media reporting. To this end, the ISS has developed a smartphone application to augment existing data sources. The Protest Reporter app allows members of the public to report incidents directly to the PPVM database, using a smartphone. It is available on both the Android and iPhone app stores, free of charge. Any protest or gathering aimed mainly at basic municipal services, including grievances against ward councillors for non-response, but excluding allegations of corruption

Socioeconomic
Any protest or gathering aimed at any socio-economic right not specifically mentioned in any other category, focused often at marginalised groups or human rights abuses such as equal rights, or ending abuse, or discrimination campaigns, or jobs campaigns, or healthcare services Transport Any protest or gathering aimed mainly at transport-related services, including public transport, tolling of roads but excluding transport issues falling under the Department of Basic Education Xenophobia Any protest or gathering, attempt or attack based on any form of discrimination or violence against people due to their ethnic, linguistic or national background. It may be against immigrants or refugees, but it may also be against South Africans from other villages, ethnic groups, religions or languages. Even if such incidents intersect with other forms of crimeviolence, looting, threats, or attacks -they are recorded. However, only if this category is the main motivation of the collective action will it be recorded as a main grievance type. 51 The process and findings will be documented in a future paper.