Definitions of violence vary according to factors associated to culture, history and politics. Yet, all cultures recognize, pinpoint and, to some extent, ban and ostracize forms of physical and psychological violence. In today’s Nigeria, rumors and claims regarding violence ripple through the media and the public worldwide. While some of them may reflect rather accurately the reality of circumscribed areas of present-day Nigeria, a number are also close to sheer misconceptions that misinform the audience. This recurrent issue has several origins and results from the specificities of violence as a piece of news or research topic. First, the contrasting temporalities and localities of violent phenomena make them difficult to observe and analyse. Ethical and methodological issues surrounding the contact with victims of violence and their relatives are also serious constraints. Finally, sources informing deadly violence are often patchy, missing or unreliable, a pattern that obstructs systematic reporting and studies.
Starting in June 2006, the Nigeria Watch project aims to offer a rigorous outlook on deadly violent phenomena at the Nigerian national scale (http://www.nigeriawatch.org/). The project consists in constituting a database listing violent deaths in Nigeria based on their coverage in the media. Categories expected to represent accurately the variety of violent deaths have been set forth and analytic tools incorporated to enable users to browse the collected data, establish their own datasets and build upon them for their own analyses. The database is open to personal as well as institutional users. It can be of help to academic and independent researchers and students, but also to corporate entities operating in Nigeria, Nigerian or foreign public administrations, international governmental organisations, and local NGOs, among others.
The Nigeria Watch project has been critical to set out a general overview of violent trends in Nigeria. It has also helped locate spatially and temporally violent phenomena. Additionally, the toolkit developed by the Nigeria Watch team enables users to identify quantitative trends and build sophisticated representations. Yet, an optimal use of the database should be predicated upon the recognition of methodological biases, especially as induced by the nature of the sources mobilised in the research process. Arguably, violence goes well beyond death and it can be surmised that most violent incidents do not lead to death. Yet, non-deadly incidents are under-reported in the news, which led the Nigeria Watch team to make methodological choices and focus strictly on deadly violence.
This paper aims to initiate a reflection on the issue of sourcing violence in Nigeria and beyond, building on the Nigeria Watch project case. It ponders the limits of the database, especially as resulting from the specificities of the type of sources mobilised in the study of violence and the difficulties it presents. It finally sets out some possibilities to improve the production and the use of the database.
First, this paper presents the Nigeria Watch project and the construction of the database. It gives insights into its accessibility by users and the usages they can make of the project overall. Second, the paper turns to the hurdles faced by the project. Indeed, conducting research on the subject of violence is not exempt from obstacles. Therefore, the paper examines examples of violent events and how they underline the shortcomings of the database. The paper concludes by noting that the Nigeria Watch database can be a markedly useful tool in academia and beyond as long as users—and the Nigeria Watch project team—are aware of its limitations. More generally, the project demonstrates the uniqueness of violence as a research topic.
The Nigeria Watch project started in July 2006. It was first supported by the UK’s Department for International Development (DFID) through the Nigeria Stability and Reconciliation Programme (NSRP). It is currently funded by the Japan International Cooperation Agency (JICA), the British Council’s Managing Conflict in North East Nigeria (MCN) Programme, the Foundation for Partnership Initiatives in the Niger Delta, the Institut de recherche pour le développement (IRD) and others.1 Since its inception, the project has been coordinated by Marc-Antoine Pérouse de Montclos, senior researcher at IRD and, since its relocation to Nigeria in 2013, the director of IFRA is supervising the team of analysts in Ibadan.
The database is accessible online2 and features advanced options for password holders. Individual researchers can obtain a password without charges upon request, through the contact form on the website—explaining briefly how the access to the data base will advance their work. Institutions can also obtain access and use the database through an agreement with IFRA. This process helps maintaining the economic model of the projects that relies on partners to cover its running costs. Moreover, it prevents an extensive commercial use of the database without compensation, in the context of growing interest for sustained quantitative data on violence, particularly on the African continent.
The Nigeria Watch database relies on the reporting of fatal incidents in the following national daily newspapers: The Sun Nigeria, The Guardian, Leadership, The Punch, This Day, The Vanguard, The Independent, Nigerian Tribune, Daily Trust and The Nation. These figures are supplemented by reports from human rights organisations such as Human Rights Watch and Amnesty International.
Launched in 2006, the project started to compile violent deaths in Nigeria from June 2006 onwards. Therefore, data for 2006 include only violent deaths reported from June to December. Until 2007, additional sources were also used including three regional daily and two national weekly newspapers as well as alternative institutional sources (reports by the police, hospitals and the judiciary and private security firms, companies and embassies publications). Nevertheless, it was later decided to set these sources aside. Several reasons may have warranted this decision: unlike newspapers, institutional reports are not issued on a regular basis. They may contain heterogeneous materials obtained from a wide-range of sources. Their biases are often of a different nature from those of newspapers. Although institutions like police forces deal directly with violence, they may also pursue policy agendas that necessarily affect the quantitative or qualitative data they produce and present. Every day, each project team member goes through three newspapers to extract relevant information. Pertinent data are then keyed into the database and processed as events with identification numbers. Crucially, the project team favors print newspapers to avoid inconveniences triggered by internet network failures. While deadly violence figures prominently in Nigerian newspapers, it is necessary to filter the sources of information. Therefore, the set of newspapers has been selected based on criteria that include their regular and sound reporting of violent deaths, the coverage of the most significant regions of Nigeria through the main national outlets, and the overall reliability of the publications.
Once an incident has been identified in an article, key elements are colour-coded to facilitate data entry at a later stage. The number of deaths is coded in red—this colour will also be used for victims’ names if they appear—; the location of the incident (State, local government area, up to the village if known) is coded in blue; the date of the incident is coded in green; and the protagonists (that is the actors and all those who are the parties to the incident) are coded in black if they are mentioned. The team also scans the sources to produce PDF versions of the articles. Once all incidents have been pinpointed and coded, the Assistant Coordinator verifies them against existing entries in the database in order to avoid data duplication and maximise the data accuracy. Indeed, journalists may report a past event which has already been recorded and avoiding double reporting of death numbers is crucial. Therefore, the Assistant Coordinator flags each article reporting events already identified.
Figure 2 shows the reporting of one particular deadly event.
Different articles often report varying death tolls of a same event. For instance, Figure 2 above represents a given coded event (a Boko Haram attack), for which sources reported contradicting death tolls: 100 people killed according to the Nigerian Tribune, 58 according to the Daily Trust, 50 according to The Punch… In such case, the team calculates the mean of death tolls as reported in all sources.
When an event includes different counts, the number used to generate trends and maps is an average of the numbers available, so that each event can be used in the calculation as a single number. This number is also published on the public database—only password access allows a user to see the different numbers reported for the event. When conflicting numbers are reported for a single event, it proves impossible to determine which one is accurate: ranges rather than averages is thus used during the data extraction to avoid this problem. Ranges allow for better accuracy while averages could be misleading and unrepresentative of the situation when you have widely varying reports—e.g. one quoting 3 deaths and another 10. In the case of Nigeria Watch the choice was made to use averages because of the geographical information system (GIS) to which the data was integrated to return mappings of violence.
Protagonists of fatal events referenced in the Nigeria Watch database include security forces, criminal groups, religious, political and ethnic groups and oil companies. Causes of violence referenced in the database are land issue, oil production, oil distribution, political issue, religious issue, cattle grazing, crime, road accidents, other accidents, fire/explosion, natural disasters and sorcery. Categorising deadly events is a critical issue that will be examined later in the second part of this paper.
After it is completed, the database is saved everyday on the server of the Nigeria Watch team located in the project’s office in Ibadan. A weekly backup is also made on an external server in Paris to guarantee the data’s safety and continuous availability.
Password holders may access online version of the newspaper articles used by the research team and review the research process. They are also entitled to full access PDF versions of the sources. This notably enables advanced users to cross-check the sources used in the database. The database also includes a built-in mechanism to process data into graphs and maps according to a variety of factors. All the charts displayed below were created with the tools provided by the database with advanced options.
The database can be used to map general trends and patterns of violence in Nigeria. Users can streamline their query to location of violence (States, local governments and communities), date, cause of violence, protagonists involved, and sources of information with an attached PDF version and a brief description of the violent incident. Users can also trend fatalities by year. As an example, the chart below produced through the Nigeria Watch database highlights the total number of violent deaths in Nigeria between June 2006 and December 2019:
Yearly trends can be analysed through the production of relevant charts. Figure 4 shows that year 2014 was the deadliest one in Nigeria since 20063 with 22,502 violent deaths. The insurgency in some parts of North Eastern states (Adamawa, Borno, Yobe) by Boko Haram contributed immensely to the great fatalities recorded in 2014 and 2015 when a state of emergency was declared in those areas. Figure 5 shows that crime caused most of the deaths with 51,876 deaths between June 2006 and December 2019, followed by political and religious clashes.
Results in the figures show that the top fatality cause in Nigeria between June 2006 and December 2019 emanated from crimes. Yet, figure 5 alone cannot suffice to identify the actual nature of these “crimes” and their potential causes. A number of hypotheses may be proposed to explain the large number of crimes in comparison with other death causes throughout the period. For instance, the 2009-2015 period corresponds to a period of surge of Boko Haram activities in Northern Nigeria. In fact, 2009 is often identified as a “tipping point” (Chouin, Reinert, and Apard 2014, 216) in the Boko Haram crisis. Chouin, Reinert and Apard have browsed the Nigeria Watch database searching the words “Boko” and “Haram” to list a total of 360 Boko Haram-related deadly events between July 2009 and December 2012 (Ibid, 221). Arguably, methodological challenges rapidly arise as to the browsing method through searching key words in the database but here is not the place to discuss them.
While the graph provided in Figure 4 shows a general trend of fatalities in Nigeria between June 2006 and December 2019, the security map (Figure 7) highlights the intensity of fatalities by state. Borno state recorded the highest fatalities since June 2006 with year 2014 accounting for most of the deaths. It can be speculated that this trend is related to Boko Haram, as the group’s attacks mostly occurred in Adamawa, Borno and Yobe. The data also highlighted how the engagement between two or more protagonists results in fatal death. Indeed, the results show that many people died as a result of clashes between security personnel and Islamic insurgent groups, especially Boko Haram.
Nigeria Watch also enables users, and more specifically password holders, to generate a wide array of maps and graphs. For example, users may rely on varying timescales to map out fatalities across sets of periods they pick. They may also produce thematic maps displaying regional intensities of particular death causes. The user-friendly interface displays death counts when users hover their cursors over given states or specific areas. Furthermore, researchers are granted the opportunity of pooling Nigeria Watch data into other formats for further manipulations. Such export grants researchers the opportunities of engaging in more complex queries with Microsoft Excel.
This table shows an information extraction on fatalities in Borno as a result of the Boko Haram insurgency in 2015. About 187 results were found in 4 pages. When exported on an Excel sheet, events can be tracked by identifier number, title, brief description, duration, states, council area and location of fatal occurrence.
As an example, Nigeria Watch has provided material for studies focusing on violence spread at the scale of the Nigerian territory. The most topical and influential observed phenomenon is the Boko Haram insurgency (Chouin, Reinert, and Apard, 2014). The authors have attempted to identify the status and religion of the victims of the Boko Haram insurgency since its outbreak. Noting that flaws in the breakdown are inevitable, they insist that the methodology used is the result of a “transparent approach” (2014, 234).
Researching violence poses epistemological, methodological and ethical issues. The first empirical problem that usually appears is feasibility. Areas of generalised or institutionalised violence such as (post)-conflict zones are difficult to access for single researchers or teams, as it has been experienced in Rwanda (Thomson 2010) and the Democratic Republic of the Congo (Van Damme 2013) among other settings. Peaceful countries also comprise sets of feasibility difficulties. The psychological fragility of victims undermine the state of research; the lack of reliable information contributes to spreading rumours and hearsay; if not necessarily untrue, rumours as evanescent social facts are often hard to trace historically and spatially (Bonhomme 2016). They may complexify the quantitative or factual reporting of death occurrences if the media do not cite their sources precisely enough.
The Nigeria Watch team was confronted with these difficulties early in the project. Relying on journalistic sources was aimed at mitigating the feasibility issue by avoiding sending researchers in direct contact with violence. Admittedly, biases may derive from this decision as the reporting of violence by newspapers is unfortunately conditioned by the “spectacularity” and attractiveness of its object. In spite of international regulations and local customs which condemn graphic depictions of violence in the media, Nigerian newspapers are riddled with ostentatious images of dead bodies and blood accompanied by dramatic titles. The sensationalist approach adopted by many Nigerian journalists, including those working for the most prestigious titles of the national press, sometimes bring about an overemphasis on sordid stories on the premise that they appeal to the readers’ morbid curiosity. If a general investigation of the salience of catchy, shady accounts in the Nigerian media remains to be done, clues indicate that they are largely over-represented. Conversely, one may suspect that less spectacular forms of violence capture a smaller amount of coverage, even when they result into death or permanent injuries. In addition, reporting psychological forms of violence like sexual harassment, insults or death threats is also difficult, although they are a critical indicator of the spread of violence in a society.
Finally, the fact that reported fatalities sometimes strikingly differ from one press title to another indicates notable dysfunctions in journalistic practices that sometimes account as a lack of rigour in the investigation process. Therefore, it cannot be excluded that fatalities may sometimes be overestimated for the sake of sensationalism. Yet, only rigorous qualitative research on Nigerian journalistic investigation practices could substantiate this hypothesis.
The project team has defined categories for the listing of deadly events following empirical considerations. The goal was to find the salient causes of violent deaths as reported by the media in Nigeria. This raises a series of crucial questions: should Nigeria Watch researchers rely on the categories used by the media themselves, or create their own? How can a category be considered “relevant” and another one “biased”? Are the actual causes of death always possible to identify in newspapers reporting? Additionally, and clearly pointing at the limitations of taxonomizing, the project attempts to integrate category imperfections by creating new ones that suit better certain types of events. For example, the category “fire/explosion” was subsequently created to identify events that were not previously identified as such.
The project team tries to bear in mind such concerns while working on the database. Deadly events may lie “in between” various categories utilised in the database. They may also be described differently or inconsistently in different sources or, even with no apparent Nigeria Watch categories that may be obviously suitable to describe them. Yet, it is impossible to neutralise fully the effects of subjectivity in the classification of violence and the team often has to take difficult decisions and use categories they subjectively think as most appropriate to count an event. While some categories do not usually pose major problems in the media reporting, such as car crashes or other vehicle accidents, boundaries between several categories are blurrier. For example, the same group can be described as “political” or “criminal.” Differences between “religious issue” and “political issue” (both Nigeria Watch categories) related fatalities may also sometimes arise as contentious, questionable, blurry or sheerly indistinguishable. Moreover, a political element may pile up to a context of land contention, blurring the distinction between “political issue” and “land issue.” Additionally, time constraints and lack of resources sometimes bind the research team to trust newspaper articles.
A particularly critical issue relates to the use of the “sorcery/witchcraft” category by the Nigeria Watch team. Witchcraft is a much debated phenomenon since Evans-Pritchard seminal anthropological work (Evans-Pritchard 2017 ). The protocol implemented by the team4 specifies that deadly events related to “cult societies or human sacrifices” are included in the category. Here, the classification in use is entirely determined by the conditions of media reporting: deadly violence is considered as resulting from “sorcery/witchcraft” if the newspaper articles in question mention the word and treat the event as such. For example, the case of the execution of human beings as part of a magical ritual by politicians would fit into this classification. Arguably, this does not suppose any other understanding of witchcraft than the one of a “normal social fact” (Durkheim 1982 ), pervasive in this context, neither does it require any characterisation of sorcery as “true” or “wrong.” In that sense, this classification complies with a scientific, epistemological stand that compels the research team to set aside any preconceptions about sorcery. However, it also binds the research team to fully trust the media accounts of the sorcery/witchcraft events, in absence of any other quantitative or qualitative sources about the cases. But what is it that media consider as “sorcery” or “witchcraft”? The protocol specifies that deadly events related to “cult societies or human sacrifices” are witchcraft. But the magical aspects might also be entangled with financial, political and social considerations. In many other cases, articles mention witchcraft without evidencing their claims, not mentioning a single convincing source. For all these reasons, if this category proves useful to understand which deadly events are classified as pertaining to sorcery/witchcraft, database users and the Nigeria Watch research team should treat the articles reporting the alleged cases with care and critical distance. Moreover, academics and policy-makers treating the data must acknowledge that deadly violence events may overlap several categories and that any standpoint concerning violence classification is necessarily questionable.
Spatial and temporal biases are a recurring issue in the reporting. For example, the available set of newspapers tends to offer a deeper coverage of south-western Nigeria while information about the South-South geopolitical zone (Akwa-Ibom, Bayelsa, Cross River, Delta, Edo and Rivers States) is scarcer. The anchorage of Nigerian newspapers used by NW is provided below.
|Lagos State||The Sun Nigeria; The Guardian; The Punch; This Day; The Vanguard; The Independent; The Nation|
The table shows that newspapers used by Nigeria Watch are based in Lagos with only two exceptions. The Daily Trust tends to have a more “federal” overview of the news, being based in Abuja, while the Nigeria Tribune, which dates back to the colonial era and remains the oldest active private newspaper in the country, is headquartered in Ibadan. Although all these newspapers officially adopt a national coverage, their various origins and locations suggest that they may develop regional perspectives. For examples, the significant amount of news on Lagos may be associated with the accessibility of the area for reporters whose employers are based in Lagos State. The reliance of the data on paper and internet media sources may raise issues on the accuracy of the data and may constitute a troublesome hurdle for data users. For example, Lagos has, of course, more coverage as a megacity than Taraba. The comprehensiveness of this newspaper reporting is thus questionable and the number of unreported violent deaths in Nigerian media is hard to estimate - which makes them underestimated in the database.
Another common bias is gender. In a paper presented at the University of Birmingham in 2016, Oladunjoye looks more particularly at travelling processes, deadly accidents and their media reporting in respect to gender. According to the Nigeria Watch database, there have been 5,872 deadly events related to travelling between June 2006 and December 2019, leading to a total of 28,886 deaths (3,642 events and 19,9859 deaths June 2006 and June 2015, the period considered by the paper). The paper lists a number of stories about travelling and road accidents in Nigeria, and particularly whether they portray women’s travels, accidents and deaths in a different way from those of men. Based on the Nigeria Watch database, the paper demonstrates the existence of serious gender disparities in the reporting of the deaths caused by accident and their locations. Indeed, some media reports discard gender as an irrelevant feature and do not detail the men/ women ratio involved.
More broadly and regardless of death causes, gender do factor markedly in deadly events, although gender-based violence is too often glossed over by journalists. For instance, gender-based violence within Nigerian households are seldom accounted for. Newspaper accounts may insist on other characteristics of the victims than their gender, and treat the event as a consequence of other factors. It is highly problematic inasmuch as extensive research in various fields of the social sciences has insisted on gender as a crucial element in violence.
The example of the underreporting of gender-based violence highlights that researchers should pay crucial attention to the origin of the data they employ. This necessary critical approach to data is a continuous process: as pointed out earlier, Nigeria Watch sources have evolved overtime. Some initial sources were abandoned following observations that their treatment provided inconsistent data, or no data whatsoever. Others were hardly accessible and would bias significantly the database by contradicting heavily figures provided by the newspapers. It was the case for example of embassy and security firm releases, as well as police, judicial and hospital reports. The team took the decision to focus on newspapers following difficulties experienced in the treatment of such materials. Finally, the objective of long-term trend analysis is also the reason why the categories originally created to sort the data have remained the same. Any change in categories would lead to the skewing of trends and hinder consistent monitoring of the evolution of violence over time.
Providing reliable data about violence in Nigeria and beyond is critical for academia and the general audience, even if trustworthy sources will remain difficult to locate. As such, the Nigeria Watch database merits to be known and used by a wider range of actors in Nigeria and abroad, be they academics, journalists, corporations or private individuals eager to locate information about deadly violence. The project team is building an extensive body of data that can have multiple usages and generate sophisticated qualitative and quantitative representations. Nevertheless, the database also reflects the difficulties and specificities of violence as a research topic. Relying on secondary sources, researchers need to adopt a critical lens on elements such as contentious categorising by journalists.
As of 2020, the database has become more popular among policy-makers, journalists and activists, notably thanks to knowledge-sharing initiatives launched by IFRA-Nigeria and the French Embassy in Nigeria. Yet, not enough academics are aware of the existence of the database. Therefore, raising awareness in academic circles about Nigeria Watch is one of the tasks to improve its impact. A regular group of users has developed gradually since the outset of the database, based notably on the professional network of the founders. Numerous individuals and institutional bodies in Nigeria and abroad seemingly still ignore the existence of the database, although its functionalities may undoubtedly interest them. This lack of awareness hampers significantly the database development, especially in terms of attracting subsequent funding to support its running costs.
Paradoxically, the Nigeria Watch project seems more cited and known by foreign researchers as demonstrated by international publications quotations than across Nigeria. From the user’s viewpoint, practical access in Nigeria is also hobbled by the limited and costly availability of the internet in spite of major improvements in the recent years. Beyond the sole access to the website, advanced functionalities provided to password holders are more internet-data consuming, hence more difficult to produce with an erratic internet access. Improving internet access for academic and professional usages is therefore critical to enhance this type of initiative.