Introduction

As a low income, turmoiled transitional country, Burundi has faced and still faces the challenge of mastering its security sector after a protracted civil war, characterized by large-scale massacres. This article sheds light on several internal security arrangements in Burundi, involving ‘beyond the state’ security, whether they are corporate-driven or informal mechanisms of social control. Alongside a corporate security sector, which presents the peculiarity of being almost completely local, Burundi also experiences occurrences of grassroots arrangements in a multi-choice policing environment [1]. These arrangements go from the use of informal sentinels in Bujumbura, to the employment of former combatants to guard palm oil plots in the countryside.

Drawing on an analysis of the above-mentioned grassroots arrangements, this article develops a back-to-back approach to the notions of ‘security governance’ and ‘security sector reform’ (SSR). The latter has become, in the late 2000s, a touchstone policy reference, emphasizing holistic approaches of security reforms at least on a conceptual level. The literature presented in the article accounts for years of debate over the progressive incorporation of non-state actors to SSR concepts and practices [2]. On the other hand, the notion of security governance remains a shallow tool of multiple social science subfields, summoning issues of coordinated management and regulations as means for governing security through heterarchical relations between various actors [3]. More importantly, the security governance literature varies widely in its approaches. For example, recent works on nodal security governance have pushed an ambitious research agenda for analytical accountings of sites of security [4]. On the contrary, concerns about the rule of law, due process and general accountability, in an environment of overlapping private and public orders, have led authors to assume various normative stances over specific nodes’ role in steering security [5].

Burundi is a sort of Petri dish of security arrangements and SSR practices. On the one hand, grassroots arrangements arguably reflect the unwillingness or inability of homeowners or entrepreneurs to resort to state or corporate security. On the other hand, Burundi has been a site of experimentation for some of the latest conceptualizations of SSR processes.

After contextualizing SSR processes in the country and some of their realizations, the article moves to the literature on security governance and underlines its moments of dialogue with the literature on SSR. In a more empirical section, insights are offered on corporate, then grassroots arrangements of security in Burundi. The use of self-policing mechanisms not only adds to the case for considering the persistence of flaws in SSR conceptualizations or implementations. It also seemingly shows how sites of security in areas of limited statehood, where different agents and normativities are intertwined in the provision of security, may remain impervious to external interventions.

The main aim of the article is twofold. First, it intends to strengthen empirical knowledge of the multi-choice policing environment in Burundi. Second and against this background, it questions the persisting connections and disconnections between the literature on security governance and the literature on SSR, both in their analytical and normative perspectives.

Burundi: a tale of needed security sector reforms

Ever since 1962, Burundi’s national history as an independent country has been scourged with conflicts. Whether interpreted as ethnic or political, outbursts of violence culminated on several occasions [6]: in 1972, massacres of unprecedented scales resulted in the death of approximately 200,000 Hutus between April and November; in 1988, around 20,000 people were killed in a new cycle of violence barely a year after Major Pierre Buyoya bloodless coup; from 1993, following the murder of the first democratically-elected president of Burundi, Melchior Ndadaye, violence settled in the country for thirteen to sixteen years [7]. Each time, the role of the Burundian security forces had been central, and recognized as such on several occasions by national elites [8]. In fact, numerous attempts at turning security forces into peace vectors have failed in the past [9].

Consequently, during the process leading to the 2000 Arusha agreement for peace and reconciliation, an exclusive committee was put in motion to negotiate the future of the Burundian security landscape as part of the broader settlement [10]. Negotiators at Arusha acknowledged that the control of security forces by a fraction of the population couldn’t guarantee protection for the whole population. The pre-Arusha situation was indeed very much in line with what Hills [11] has shown for other newly independent countries: as new rulers needed last resort authority to draw on if their position was to be challenged, police could provide a coercive force to back this authority. As an institution playing such an important role in power preservation, police in pre-Arusha Burundi had been able to resist transformation, precisely thanks to its entanglement with political authority.

For these very reasons, the Arusha agreement, and particularly its third protocol, explicitly identified the security institutions as the main forces behind the violence [12]. The agreement acknowledged that the instruments of state power in charge of public safety were directly or indirectly responsible for the massacres, precisely because they never became inclusive instruments in the service of the population, instead of a tool for authority preservation [13]. Therefore, the Arusha agreement pursued two important goals related to national peace and security. First, the agreement addressed the delicate psychological and political equation by introducing ethnic balance through strict quotas in, among other public institutions, the defense and security forces. Second, the agreement required the separation of the ‘gendarmerie’ from the army through the creation of a new stand-alone national police force. Even though the project had already been proposed by President Ndadaye’s administration in 1993 [14], resentment in military circles and the fear that a national police would counterbalance the army regularly overturned the project, especially after the murder of Ndadaye [15]. While the reform proposed in the Arusha agreement concerned all security actors, the agreement remarkably put forward the ambition to create three separate bodies, with specific missions. A military body: the Force de défense nationale (FDN); a public security body: the Police nationale du Burundi (PNB); and a national intelligence service: the Service national des renseignements (SNR). In terms of political management of these defense and security bodies, the agreement paved the way for democratic civilian oversight and clear demarcation between the political and military spheres. It mentioned the principle of accountability for these new institutions and political neutrality, along with parliamentary control of their activities.

On an implementation level, while the Arusha agreement was massively ambitious, it faced one arguably last challenge: the need to end the hostilities with the remaining main rebel movements. If the agreement was signed in 2000, it was in the notable absence of the two largest Hutu rebel groups: the Palipehutu-Forces nationales de libération (Palipehutu-FNL) and the Conseil National pour la Défense de la Démocratie-Forces de Défense de la Démocratie (CNDD-FDD). Following a change of direction in 2003, the CNDD-FDD signed a special cease-fire agreement with the transitional government. And while it didn’t become party to the Arusha agreement, leaders of the CNDD-FDD actually started to act in its spirit [16]. It was particularly the case from 2005, when Pierre Nkurunziza, a leader of the CNDD-FDD, was elected president and formed an inclusive government in formal compliance with the Arusha agreement. On a political level though, many authors have suggested that new commitments related to the agreement were yet to be fully implemented, which the CNDD-FDD never actually did [17]. At that time, the reorganization of the PNB had already started, with a series of laws promulgated in December 2004 to shape the institution as prescribed by the Arusha agreement. These laws formed the first backbone of the mandates, missions and bodies of the PNB [18]. For these reasons, the new ruling party had to quickly engage with the continuation of this PNB overhaul. In a relatively short period of time, the PNB thus replaced the gendarmerie to become a significantly restructured body of civilian security, with its members increasing from about 2000 officers to more than 18,000 – coming from the large pool of former rebel combatants and military – in 2008 [19].

It can be acknowledged on the one hand that, from 2005, Burundi has been undergoing ambitious SSR processes, along with the continuation of Disarmament, Demobilization and Reintegration (DDR) processes. Initially solely focused on ethnical and political balance in the institutions, DDR and SSR processes have permitted to achieve a first wave of reorganization of the country’s defense and security forces.

But on the other hand, the ambitions expressed in the Arusha agreement have met several pitfalls. For instance, the massive and heterogeneous recruitments quickly posed challenges to the professionalization of the new PNB, especially regarding its interaction with civilian population [20].

Several donor states and development partners have been dedicating a significant portion of their aid budgets – through both bilateral and multilateral programs [21] – to support the police. As a result, progress has undoubtedly been made as police reform is concerned. While the formulation of the 2004 round of laws and decrees related to the PNB organization was clearly a very much prospective one – with lawmakers essentially focusing on the hypothetical future of the post-Arusha police – the following organizational vacuum has been the target of several strategic and ethical plans. For instance, the decree organizing the missions and organization of the General Directorate of the PNB – the structure in charge for the daily management of the police inside the Ministry of Public Security (MSP) – emerged in 2007 [22]. Police staff status and ethical codes were adopted in 2010 [23]. In the same context, new material and train-and-equip programs have also been developed to strengthen professionalism within police forces. However, and against this background, many indicators show that SSR processes regarding the PNB have produced mixed results: from 2004 onwards, corruption [24], the question of resources [25] and the question of accountability [26] haven’t been sufficiently tackled with, demonstrating insufficient progress in the development of a transparent, professional and accountable institution of public security in Burundi.

This broad context sheds some light on the complexity of security governance in post-Arusha Burundi.

Notional boundaries of security governance and analytical value

Sometimes highlighted as a key to understand the world, the notion of governance remains elusive. The history of the concept brings no comfort here: from its birth in the business world, to the prescriptive or ideological meanings that it has been and is still carrying, particularly in the development community, governance as an operational concept may have been rejected, particularly in French academia [27].

Still, security governance has risen to prominence these last years, including as a seemingly stand-alone concept. It has indeed been used by scholars to account for the contemporary replacement of vertical, hierarchical structures of governing security affairs by horizontal networks, coordinated management and regulations between multiple and diverse actors [28]. However, the analytical potential of the concept of security governance varies from one author to the other, as does the very value of the concept as an independent variable to address empirical observations. This seems to derive from the widely heterogeneous ways in which the concept of security governance has been applied in social science literatures such as International Relations (IR) and criminology.

In IR, security governance has mainly been looked at through a global lens, stressing the emergence of new transnational security architectures since the end of the Cold War [29]. In this fashion, the interest of the concept of global security governance lies in its ability to apprehend some of the key questions of post-Cold War security management, notably the competing approaches of multilateralism [30] and the rise of international regimes bolstered by civil society [31]. Also, IR attention has recently focused on private security governance, following the growing use of military companies operating outside their home state with the potential to resort to lethal force [32]. In this context, many authors have questioned the opportunity to create new transnational normative frameworks, whether state-driven [33] or market-driven [34].

In criminology, security governance has been a way to describe a ‘quiet revolution’ [35] in security provision. Throughout history, the centralization of social control in one single public police – a state body consisting of specialized officers of public security – lasted only a short time [36]. In this respect, the pluralism that characterizes the development of security arrangements was not a novelty in the 1970s. This pluralism has however accelerated even further recently. Accordingly, the criminological perspective on the post-Keynesian police gradually crystallized around the notion of policing [37]. In this sense, security governance in criminology amounts de facto to governance of policing, policing being.

‘(…) those organised forms of order-maintenance, peacekeeping, rule or law enforcement, crime investigation and prevention and other forms of investigation and associated information-brokering – which may involve a conscious exercise of coercive power – undertaken by individuals or organisations’ [38].

In this last sense, security governance has much more to do with the management and enforcement of social control and community conflict resolution than it has with international peace and security.

Despite the utility of the many iterations it can be apprehended with, this short literature review allows to argue that the analytical concept of security governance draws from similar premises. This can be deducted not only from the common semantics of both IR and criminology but also from a close examination of their common theoretical foundations [39]. Cited works often comprise references to Beck – on risk and the management of a ‘bottomless barrel of demands’ [40] for security – and to Foucault’s views on a shared and widely spread conception of power [41].

As a result, four overlapping characteristics of the concept of security governance can be identified, arguably forming the notional boundaries of this concept.

First, the concept refers to the reconfiguration of the state position in a changing security environment, for instance in matters of international high politics or regarding social control.

Second and consequently, security management and production do not fall under the exclusive purview of the state, but rather is the result of a proliferation of public, private or hybrid actors acting towards each other in various fashions.

Third, the state may nevertheless retain, to a certain extent, a role in the steering of interactions between security actors, through rules and regulations for instance. But the state loses its position as highest item of a hierarchical, vertical decision and implementation structure.

Fourth, there is a new orchestration of security production and management, made from multi-level, heterarchical relations in multiple security arrangements.

Within these notional boundaries, security governance can be understood as a broad analytical framework of interpretation. In this sense, security governance would be considered a minima as a ‘general trend of research that requires the analytical look to also cover non-institutional aspects of government processes’ [42] in the security environment. This understanding of security governance implies that all modes of coordination between every stakeholders of security, their motivations and resources should be looked upon, from individuals to social groups, corporate security businesses, local communities and (international) institutions that collectively lead or seek to provide security services. This is very much in line with Wood and Shearing’s methodological considerations on nodal security governance [43]. The authors advocate for ‘a comprehensive empirical mapping of existing governance nodes and networks within specific sites’ [44], as a key phase for engaging in the design, implementation and diffusion of ideas and practices. This epistemological stance is ambitious, and the empirical work it suggests to undertake seems limitless. However, this research agenda has been repeatedly promoted, as it was the case in Hufty’s early work on a ‘governance analysis framework’ [45]. This framework aims to be a similar practical methodology and also promotes mappings of governance sites, showing networks of interacting nodes, standards and norms, actors, and processes.

Finally, it should be mentioned that this agenda has started to produce innovative and comprehensive research [46]. Drawing on Bourdieu’s theory of capitals, Abrahamsen and Williams precisely manage to overcome Hufty’s fear of the impossibility to effectively order such an amount of data. They report and analyze security assemblages in various sites of Sub-Saharan Africa to account for.

‘new practices and forms of power that cannot be neatly contained within the geographical boundaries of the nation-state (where) various security actors interact in a field of tension, structured by the opposition between the public and the private and their different forms of material and symbolic’ [47].

Berndtsson and Stern also provide the mapping of ‘complex struggles and negotiations involved in the redrawing of public–private divides and sites of authority’ [48] at another site of security: the Stockholm-Arlanda airport.

Wood and Shearing’s proposal, as Hufty’s, invite to acknowledge and operationalize the common premises of the security governance literature. Both seem indeed particularly suited to produce analysis of security governance in given sites of security. Most of all, both projects intend to cope with the often normative persistence of the Weberian model of statehood in security environment analysis. As Wood points out:

‘For researchers, this means that the explanatory work they carry out should not be driven by the assumption that particular nodes are more effective or, for that matter, more democratic than others’ [49].

The following section presents how both first generation critiques of SSR and contemporary SSR policy concepts seem precisely to have been shaped, to a significant extent, by the idea that a particular node should be put at the center of a security network.

SSR as a node-centered view of security governance

Somehow departing from nodal security governance as sketched out above, some authors in IR and criminology have made use of the concept of security governance more as a normative tool. Those authors explicitly or implicitly emphasize how specific nodes ought to play a role or how specific nodes ought to be taken into account in the security network.

The first common bias to be considered concerns the persistence of the Weberian model of statehood. Deriving from considerations on the rule of law or democratic oversight of the security sector, strong emphasis may be put on the state, framed as a necessary steering node of security. The idea that the state should remain the central locus of security governance offers ways to cope with ‘questions related to accountability and governance within this new landscape that could have profound implications for human rights, privacy and civic society as a whole’ [5]. This normative viewpoint has led to strong debates precisely over nodal security governance, its use as an analytical framework and its possible – if not desirable – development into a normative one [50]. On a similar note, authors in IR have suggested that a regulatory framework for the private security industry ought to be state-led or at led along state parameters [51]. Some authors have, since, changed their minds [52].

The connected, second, more recent bias to be found refers to considerations on universal provision of security or the value of market-driven normative frameworks of security. Following years of prominent malpractices in the private security industry, authors may have had a tendency to emphasize the role of corporate security providers in security governance. While claiming to engage in hybrid approaches of the security environment, this may ultimately have led to the neglect of informal or multi-choice policing [53]. For instance, Abrahamsen and Williams’ early work on ‘bringing the private in (SSR)’ [54] was clearly fundamental in 2006. This was especially the case at that time, given the epistemological stands and the seemingly impossible look beyond statehood in SSR early concepts [2]. However, such a focus on a single node – here the corporate security sector – is precisely one of the flaws of contemporary SSR concepts, as it will be discussed below.

Finally, issues of legitimacy in the public-private debates are still sensitive, particularly in areas of limited statehood. As Börzel and Risse argue:

‘it goes without saying that the “horizon of legitimacy” provided by the state is equally weak. Moreover, the weak state of developing countries is usually not very democratic and the rule of law is equally deficient. As a result, we need to have a close look at the various governance processes and institutions to ascertain their legitimacy’ [55].

Focusing on the state or the corporate sector as primary nodes in the security network do not give full merits to Wood and Shearing’s approach of nodal security governance. Moreover, this state or corporate-bias also seem to be at work in even recent, more inclusive SSR policies. In this respect, the literature on nodal security governance and former or recent SSR concepts arguably disconnect from one another on the issue of node-centered views.

Framed as a ‘new aid paradigm’ [56], SSR models have been conceived as ‘good scenarios’ of development for state donors’ action in the realm of security. This openly normative agenda [57], tracing back to the end of the Cold War [58], is best found in the Organisation for Economic Cooperation and Development (OECD) policy documents. Inside the OECD, the Development Assistance Committee (DAC) has emerged as a major organization in the elaboration of guidance for development aid. Moreover, the DAC has been, since 1997, one of the main sources for SSR guidelines. Bryden [59] traces the conceptualization of SSR in DAC forums. He stresses how, from the creation of a dedicated network of discussion in 1995, to the elaboration of formal guidelines by state members, the OECD/DAC have been instrumental in the affirmation of the security-development nexus, offering ready-to-go conceptual framework for donors’ work on development cooperation in conflict and post-conflict situations [60]. SSR’s approach is problem-solving and aims to realize convergence between security and development programs through new synergies. Although the DAC guidelines and policy documents have often been criticized by scholars as technical and managerial methods promoting western liberal values [61], many donor states have worked in line with and have implemented DAC recommendations. Therefore, many states rely heavily on DAC policy documents to design their approach to SSR. It is particularly the case for the Netherlands, which growingly aligns on the OECD/DAC Handbook on SSR [62]. The Handbook goes a step further than former DAC guidelines and constitutes one the most important document dealing with SSR concepts and implementation.

Notably, the Handbook allegedly promotes a genuine paradigm shift in the way security governance is conceptualized and SSR is to be implemented. As a matter of fact, recommendations enclosed in the Handbook explicitly stress the necessity of a holistic, hybrid, so-called ‘wide’ approach in SSR, nominally extending fields of intervention far beyond ‘the usual suspects’ of defense and social control. As it reads, the Handbook even seems to reconnect with contemporary analytical approaches of security governance:

‘Understanding who provides security and justice is central to SSR. The reality in most countries is that these services are delivered by a large number of actors. Some are state agencies and services, but some are likely to be non-state organisations and systems. Although the state has an irreducible role in justice and security provision, effective reform across the system requires working with a broad spectrum of actors’ [63].

For instance, the Handbook largely emphasize the question of non-state corporate security actors, raising concerns over universal provision of security and accountability. These concerns have also been raised in other development policy documents, such as in the World Bank’s [64].

However, this nominal account of broadly defined actors reportedly produces mixed results. The more recent and prominent policy documents do formally include references to various actors and their interactions, and have extended the definition of the security sector to include more than the police and the armed forces. But grassroots policing mechanisms and actual encompassing mechanics of security are still on the sideline of SSR, or at least do not overcome implementations hurdles. As a matter of fact, Brzoska [65] points out the risk that SSR, as it is currently conceived, could be met with reactive behavior on the ground and produce unintended outcomes. On a same note, numerous authors have recently accounted for situations of disconnection between SSR and local dynamics: Hills [66], Baker and Scheye [67], Menkhaus [68], Lemay-Hebert [69], etc. The ‘new perspectives’ on SSR promoted by Schroeder and Chappuis in this context are enlightening:

‘Yet what is lacking in understanding the exigencies of context are generalizable insights into the nature of security governance and reform at the local, inter-personal levels of interaction. Fine-grained, micro-focused empirical data, describing security from the bottom-up, has been largely absent from studies of SSR’ [70].

This is precisely where the literature on SSR – or rather this new critical perspective on SSR – and the literature on nodal security governance connect again. Read together, they invite to investigate local security dynamics in association with SSR concepts. In Burundi for instance, examples of grassroots policing arrangements not only tend to show that ‘the future is non-state’ [71] but also that the future is non-corporate.

Methodology and data collection

Given Burundi’s peculiar state of bureaucracy and the consequent difficulties to effectively rely on its administration [72], the method of data collection used for this research was open-ended. The research questions revolve around security governance, internationally backed SSR, corporate security and grassroots policing. Consequently the study mainly makes use of various publicly available documents: Burundi’s laws and decrees regarding security, international guidelines issued by the OECD/DAC, evaluations conducted by donor states on security needs and on the private security sector in the country, a comprehensive evaluation of all cooperation and development policies of six European countries and the European Commission in Burundi, several reports from local and international research centers on local security initiatives, etc. While publicly available, certain documents such as Burundi’s laws and decrees may not be available online. Publications of the government gazette, the Bulletin Officiel du Burundi are, at best, scarce and generally not available online. Numerous background interviews were also carried out. The interviews were conducted between 2011 and 2014 with various stakeholders, including officials from the MSP, police officers, representatives of the Belgian and Dutch development programs, executives of all major gardiennage companies [73] in Burundi except for one, local journalists and scholars. In total, 33 interviews were conducted in-depth but often in an informal fashion. Indeed, the general culture of secrecy surrounding domestic security as stressed by Westley [74] remains very much at work in Burundi, and several officials and executives politely declined to be taped, although accepting ‘to go on record’, as long as the record was cautiously written and, on some occasions, revised.

Therefore, the methodological walkthrough for this paper is pretty straightforward. First, it resumes the narrative initiated in the first part of the article, then presents the realizations of the most ambitious SSR program in Burundi, to quickly assess these realizations in the light of OECD’s conceptual ambitions. Finally, the article moves to the analysis of grassroots, policing arrangements, as examples of security mechanisms that remain either untouched or arguably fostered by SSR processes.

Coping with the blind spots with blind spots: examples of grassroots policing mechanisms in Burundi

Quickly following the signing of the Arusha agreement for Burundi, DDR and SSR processes began with ambitions of a complete security landscape overhaul. Over time, this broad transformation process has been supported by many multilateral and bilateral donors: main donors are the World Bank through the International Development Association, the United Nations (UN) office in Burundi, the United States, the Global Fund, the United Kingdom until 2012, the European Union, Belgium, Germany, the Netherlands and France.

As mentioned above, SSR programs in Burundi have achieved several technical and governance-oriented objectives toward better functioning defense and security forces [75]. However, train-and-equip programs have not been fully completed across the country [76] and political will has been reportedly weak as governance-oriented objectives are concerned [77].

Along with state-centric measures, donors in Burundi and Burundian authorities have nevertheless seemingly begun to embrace a holistic approach of security. In particular, in 2009, the Netherlands and Burundi launched a comprehensive program of ‘security sector development’ (DSS) [78].

Initially designed as a three-year plan, the partnership was still operational in late 2015 and has shown progressive ambitions to the transformation of Burundi security sector [79]. Not only has the DSS program targeted the reinforcement of the PNB, it has also specifically targeted non-state actors and local dynamics in an incremental fashion. In terms of realization, the DSS has been instrumental to the enhancement of the capacities of the MSP to supervise public security as a whole [80]. Several Burundian laws and decrees clearly derive from this cooperation, such as the 2009 decree on the Ministry reorganization [81], the 2011 decree on the creation of a division of the MSP designed to oversee and control gardiennage companies [82] and the 2013 decree defining the missions of this division [83]. The creation of the division was pure pragmatism: a recent study commissioned by the DSS established the number of corporate guardians at 8878, working for 23 companies operating mainly in the capital-city, Bujumbura [84]. In contrast, there are about 18,000 members of the PNB across the whole country [85]. Yet, the creation of the division summons the specter of an aforementioned common problem in Burundi’s SSR: as it was the case for the MSP, several years separated the creation of the division and the legal definition of its missions. Furthermore, while the 2011 and 2013 decrees are a specific national legal instrument allegedly framing democratic oversight and the activities of the corporate security sector, the division remains an empty shell, with no historical practices and background [86]. Before 2013, a few gardiennage companies were actually regulated under unspecific provisions: the code of private societies, enforced by the Ministry of Trade and the Commercial Court. Other companies operated as non-profit associations [84]. Moreover, the newly created division is currently dramatically understaffed and underfinanced [87], and the 2013 decree lacks clarity on the methodology of concrete monitoring and control of gardiennage companies [88]. Finally, Burundi’s corporate security landscape is characterized by small, local companies. To date, one single international private security company, KK Security, is operating in the country. Other companies came mainly from initiatives of former police or military officers, or former rebels [89], with strong connections to national authorities. Therefore, recent efforts to clarify the normative framework of corporate security across Burundi have yet to produce tangible effects, as only 9 out of the 23 companies have complied with the homologation process as of June 2014 [84]. In Burundi, the local corporate security sector still resists global ethos, despite international support and the creation of institutional tools.

As Baker argues, the burden of survival and protection often transfers to individuals and their household in post-conflict settings. In such circumstances, self-policing becomes the norm [90], whether in the form of resorting to corporate security or grassroots arrangements. In Bujumbura, if the first form is quite visible, the latter seems favored by the population [91]. Actually, a vast majority of home and business owners don’t rely on companies focused on security. Instead, hotels, expatriates, bakeries, banks, phone operator agencies, grocery stores and home owners trust single individuals, whom they employ directly. Those individuals are usually called zamus – mostly by foreign nationals – or sentinelles. The sentinelle is part of the day to day life of the Burundians, and while the phenomenon is hardly quantifiable, sentinels dressed with the famous blue, sometimes brown, uniform are to be seen everywhere in the street. On Singer’s metaphoric spear for a private security typology [92], sentinels would likely add a bottom ladder. They indeed represent providers of security excluding almost any use of force. Sentinels are rather elements of dissuasion focused on alarm-triggering and door-opening, when they don’t act as chauffeur. They, in fact, could sometimes qualify as hotel concierge or handyman [76]. Informality is the norm here, as it is the case in other documented occurrences of informal security in urban settings [93], and in contrast with partially formalized instances elsewhere in the world [94].

The sentinel phenomenon may be looked upon as untouched by both the state and SSR processes. In post-colonial Burundi, the burden of crime control has remained for large parts on communities, even in rapidly transforming Bujumbura [95]. Moreover, the phenomenon doesn’t derive from an explicit governmental policy or strategy, adopted to progressively focus on core security tasks and actively solicit corporate actors or the general public [96]. As a matter of fact, Burundian strategy documents on security tend to emphasize the exact opposite and conceptually promote a state-led security environment:

‘The security policy will be comprehensive and will refer to a wide notion of security. Police will expand all its services to the population. Besides its traditional tasks of internal security, it will further develop prevention and emergency first line. The security policy will integrate all security actors: organizations related to control and management of security, judicial institutions, risk prevention and disaster management agencies, private security services, civil society and the population'[80].

While the Arusha agreement may have started new dynamics between society and the state, reforms have so far produced mixed outcomes as empowerment of population-oriented public policing is concerned. From its independence until today, Burundi remains a country where the police is seen as an instrument of domination or preservation of power rather than an institution operating under public accountability and democratic control. Major remnants of corruption, politicization, human rights violation, or simply lack of professionalism are to be found on a daily basis [97]. Recent crackdowns on dissent in the country have come as a confirmation of this observation [98]. Therefore, while wide programs have been put in motion since 2005, SSR processes have yet to realize their ambitions of state-centered, population-oriented security governance.

Most home or business owners in Bujumbura don’t rely on corporate private security either. In a multi-choice policing environment, this new choice seems to be welcomed with skepticism by the population. First, the corporate security sector is fairly young, with the creation of the first gardiennage company dating back to 1992. Second, gardiennage companies are seeking high-capital customers, such as diplomatic representations, non-governmental organizations, agencies of the UN system and foreign nationals [99]. This is understandable, as those institutions and individuals often prefer corporate services to those of the state, reportedly for the reason that the latter is not civilly responsible and rarely pays damages for the defects of its services [84]. Accordingly, the market for security in Burundi is marginally oriented towards the general population. Third, and consequently, corporate security remains expensive [100]. In contrast, sentinels remain cheap and flexible, as they enjoy a small salary often completed by informal modes of remuneration, such as the possibility of being housed, fed or dressed.

For the above-mentioned reasons, the sentinel phenomenon has so far resisted international and state ambitions towards state-centered reorganizations of security governance. This grassroots mechanism has seemingly been untouched by changes in the Burundian formal security architecture. More importantly, it should be noted that, so far, the use of informal sentinels has been completely overlooked as rules and regulations are concerned, both by the state and internationally backed SSR programs. For these reasons, the use of informal sentinels in Bujumbura remains a blind spot, untouched by both state-centered and corporate-centered approaches on security governance.

About 70 km south of Bujumbura is the city of Rumonge, located in the province of the same name, formerly the Bururi Province. Along with the city of Nyanza-Lac, Rumonge concentrates more than 10,000 ha of oil palm groves, scattered in hundreds of small plantations [101]. The Rumonge region is affected by at least two sources of insecurity. First, the civil war has become root for numerous land ownership conflicts [102]. In return from exile, part of the local population found that many of the properties changed owners, new owners often being close to local circles of power [103]. The Clingendael Institute [104] reports that there are over 3000 cases of unresolved land disputes in Rumonge alone. While the redistribution of land was supposed to be carried out by the state through the National Land and Asset Commission, issuance of ownership titles have been controversial over the time. In fact, the Commission is more and more seen as a tool to impose an unequal partition of land between returnees and new owners. In this fertile region, the problem of land ownership is felt acutely and occasionally leads to violent dispute [105]. Second source of insecurity is the general poverty and unemployment in a wealthy area. In Rumonge, the theft of palm oil crop – the source of wealth – was a recurring problem for the cooperatives stocking them [79]. Moreover, plot owners were not used to go to the police to signal the theft of their product. Testimonies gathered by Scheye are, as this matter is concerned, self-explanatory:

‘It is not the role of the police to patrol in the fields. Here, the police responsibility is to protect us in our homes’.

‘people have to organize themselves to protect themselves’

(…) Another said that the police stay on the roads and that what happens off the roads is the concern of the citizenry’ [104].

For these reasons, 2000 owners and tenants of palm groves, joining in the Palm Oil Guard Association (POGA), turned to another source of security than the police: ‘jobless, thieves, street kids and ex-combatants’ [106].

Directly recruiting from the pool of those who otherwise would have probably been the one stealing from them, owners of palm oil plots came in contract with former combatants, willing to join the POGA as guards. The POGA guards receive palm oil rations as remuneration. The association also serves as facilitator on the market [107]. Little to not equipped, the responsibility of the POGA guards is mainly to dissuade thieves to act at night, or to catch them. The guards’ behaviors have sometimes been questioned and, formations have been dispensed these last years by local partners, backed by the DSS to improve collaboration between the guards and local authorities [108]. As a consequence, cases of torture and beatings are being progressively replaced with surrendering to proper authorities. Furthermore, the POGA now promotes a peculiar reconciliation process between the owner of the stolen palm oil plot and the thief [104]. It should also be noted that the POGA provides its guards with insurances, in case of injury in the line of duty.

The POGA constitutes another example of grassroots policing. The first rationales behind this mechanism are similar than those accounting for the sentinel phenomenon. First, the PNB presence is scarcer than in Bujumbura and more prone to abuse [109]. Second, corporate security sector presence is limited, as most foreign institutions and nationals are concentrated in the capital-city.

The POGA case calls for two interesting observations. First, the POGA is a local grassroots policing arrangements occasionally backed by a state donor’s SSR program, namely the DSS. Second, the case nevertheless raises important questions regarding early SSR in Burundi. DDR processes may have appeared successful in terms of quantitative goals towards the reduction of state armed forces. However, three elements should be noted. First, although some stability has been found in Burundi after the implementation period of DDR processes, socio-economic development is still very low. The country remains one of the poorest in the world, leaving few job opportunities for ex-combatants: even under favorable conditions, the DDR process reinstated in poverty in major part of the countryside [110]. Furthermore, the poverty has also been aggravated by flaws in the reintegration phase. This phase has been achieved with little to no regards for the acquisition of new skills, vocational training and general guidance [111]. Second, the reintegration into new public security institutions dismissed many: disabled or invalid, older, undisciplined candidates, and those whose education level was too low to follow a professionalizing military or police training. Third and finally, the lack of implementation of transitional justice mechanisms in Burundi, the provisional immunity of ex-combatants under the various cease-fire agreements, and the little emotional support during DDR processes raise debates over the relationship ex-combatants have to violence. By hiring background-checked former combatants, the POGA thus not only provides security, it also embraces social considerations.

In this respect, the POGA case underlines two elements: occasional consideration from certain state donors for local dynamics of security on the one hand; major flaws in international and national efforts towards the reintegration of weakened population on the other hand.

Concluding remarks: on the nodal security governance research agenda in multi-choice policing environment

Both examples of the sentinels and the POGA guards provide interesting insights on grassroots policing in Burundi. Three short sets of conclusions can be drawn from these examples. First, it is clear that these arrangements arise from a certain degree of distrust in public security institutions. It is also clear that the persistence of the sentinel model has not been mitigated by the growth of a corporate security sector. Second, both arrangements have been almost completely overlooked by Burundian authorities so far. Yet, as highlighted in the paper, the sentinels and the POGA guards actively engage in order-maintenance and prevention, with various degree of coercive power. On the one hand, they could be considered as policing actors. On the other hand, it remains doubtful that the sentinels and the POGA guards’ employers themselves – home or business owners – consider provision of security as their key purpose, given their primary social role. However, the outcome remains the same: local businesses and individuals have engaged, for various reasons sketched out in this article, in security-related functions at their own benefits, at the benefits of their customer or, incidentally, at the benefits of the general public. In this respect, these security actors to contribute to multi-choice policing. Third, the most recent SSR policy documents and implementations fail to grasp the full extent of security governance on a given site. More importantly, the POGA guards case shows that grassroots policing arrangements may appear as a consequence of shortcomings and pitfalls of international interventions. The article shows how the notable failure of DDR to reintegrate weakened former combatants have contributed to situations of insecurity that have been addressed by the creation of the POGA. While some grassroots policing may have been around for some time, the formalization of this specific mechanism adds new layers of complexity to security governance in Burundi.

This last insight raises many questions regarding security governance and the way SSR research operationalizes it. At first sight, security governance could be framed as fragmented into separated, non-collaborative agents. On one side, state-centered security governance is backboned by donors’ discourses on local empowerment and caution vis-à-vis the corporate security sector. On the other side, grassroots security mechanisms, as observed here, could appear or could be framed as untouched by global discourses and norms. The empirical findings regarding the POGA guards show the exact opposite: palm oil plot sites of security involve donor states, local authorities, plot owners, criminals, former rebel groups etc. in a complex security architecture.

It seems clear that SSR researchers and policy-makers could benefit from the nodal security governance perspective. For example, the sentinel case continues to raise many questions, to be addressed with a full mapping of the security network. In this paper for instance, collected data is insufficient to assess mutual relations between the sentinels, gardiennage companies, the PNB, the MSP and external interventions. However, many hypotheses should be addressed as this site is concerned: are sentinels symbolically influenced by the new corporate sector regulations? Would this alleged influence derive from the fact that these regulations are internationally-sponsored? Has the sentinel phenomenon been fostered by a general sense of security/insecurity in Bujumbura? Is this sense of security/insecurity somehow related to the growth of the corporate security sector? Does any of these propositions make sense?

In his review of Democracy, Society and the Governance of Security, Haggerty discusses Wood’s methodological considerations on nodal security governance, stressing that:

‘The sheer scope of this epistemological project suggests that nodal governance seeks to be omniscient. Unfortunately, even with a small army of highly coordinated researchers such ‘preliminary’ research would likely take years to accomplish, and would culminate in a voluminous but unprioratized archive’ [112].

Researchers that engage in this type of work in sites of international interventions or more generally in sites of limited statehood must be prepared to face unparalleled uncertainty when gathering empirical evidence. Put in other words, they should be prepared to face a bottomless barrel of data.