Ethnic Quotas and Foreign NGOs in Burundi: Shrinking Civic Space Framed as Affirmative Action

Abstract

Since January 2017, foreign non-governmental organisations (ONGEs) active in Burundi are required to respect ethnic quotas (60 per cent Hutu, 40 per cent Tutsi) when employing local staff. The ethnic quota requirement was adopted amidst fears of re-ethnicisation of politics and society, enhanced control on civil society and tense relations between the Burundi government and its aid partners. While authorities justify the measure as a remedy for decades of discrimination along ethnic lines, an analysis of the legal reform shows that a variety of other motivations and dominant party interests account for its adoption and enforcement. While the reform mirrors a wider international trend of shrinking civic space, the Burundi case study also shows how a clever discursive strategy may skillfully divide ONGEs and their funding agencies. Furthermore, the case study reveals the instrumental use of obscurity and ambiguity in terms of the legal wording and enforcement of the ethnic quota requirement.

Introduction

In January 2017, Burundi enacted new legislation governing the admission and functioning of foreign non-governmental organisations (ONGEs) on its territory. An internationally mediatised row around its implementation peaked when, in September 2018, the National Security Council suspended all of the approximately 130 ONGEs then registered in Burundi. At a meeting of the UN Security Council on 13 November 2018, several members expressed concern (United Nations, 2018a). Some six months after their suspension, 93 ONGEs were re-admitted. During an interview on TV5 Monde, multi-award winner Marguerite Barankitse called them collaborators of a fascist regime. ¹

One of the most controversial aspects of the legal reform is the requirement for ONGEs to respect ethnic quotas in the recruitment of local staff. This paper sheds light on this aspect of Burundi’s new ONGE law, which was shrouded in obscurity and the subject of major controversy and polarisation. For ONGEs opposed to the reform, it seeks – at the very least – to extend government control over non-state actors and aid flows while – at worst – it risks sowing the seeds of renewed ethnic conflict in Burundi. Government officials, however, framed and justified the measure as a matter of affirmative action necessary to counter a longstanding practice of ethnic discrimination in ONGE local staff employment.

I analyse the ethnic quota provision in the 2017 ONGE legislation using the typology of effects of law (direct, indirect, independent, and unintended effects) developed by legal sociologist John Griffiths. The analysis shows how the quota requirement is highly (and probably deliberately) vague in terms of its direct effects, faces a credibility deficit in terms of its stated indirect effects, but produces a number of independent effects that serve various interests of the government and, in particular, the dominant party CNDD-FDD. The paper analyses the ONGE ethnic quota requirement against the background of a wider range of contemporary governance challenges in Burundi: ethnic diversity, state-party relations, and development partnerships. Methodologically, the paper is based on documentary analysis (legislation, parliamentary documents, official statements, press releases, tweets, diplomatic cables, etc.) and on interviews and confidential conversations between January 2017 and June 2019 with forty-four informants (government officials, diplomats, ONGE staff, etc.), none of whom wish to be identified, for obvious reasons. ONGEs are named only if they released a public statement (which surprisingly few of them did).

Before introducing the analytical framework, I first sketch the context in which the ONGE ethnic quota requirement came about: a history of ethnic conflict followed by ethnic power-sharing, the return to a tradition of dominant party control over state and society, and a crisis between the Burundi government and its aid partners.

Context

Burundi’s post-colonial history after 1962 was marked by approximately four decades of violent political conflict along ethnic lines, followed by a negotiated and initially widely applauded (Curtis, 2019; Grauvogel, 2016) institutional re-engineering of the state on the basis of ethnic power-sharing. Under UPRONA single party rule (1966–1993), Burundi was characterised by severe horizontal inequalities along ethnic but also regional lines (Nkurunziza, 2012: 215). Power was largely concentrated in the hands of members of the Tutsi (demographic minority) group from southern Bururi province. Four months after the 1993 democratic elections won by the FRODEBU party and its presidential candidate Melchior Ndadaye (of the Hutu demographic majority group and from central Gitega province), Tutsi military pre-empted a “hostile” takeover of the state by assassinating the newly elected president. This unleashed a decade of ethnic civil war between Tutsi dominated government forces and Hutu dominated armed rebel movements. Following Tanzanian (Nyerere) and South African (Mandela) mediation, a first peace agreement was signed in Arusha on 28 August 2000. The Arusha Peace and Reconciliation Agreement (APRA) included both interim and post-conflict power-sharing institutions, turning Burundi into the most consociational polity on the African continent (Lemarchand, 2007). An essential feature of the APRA was the use of ethnic quotas to allocate positions – either on a 50/50 per cent or a 60 per cent Hutu /40 per cent Tutsi basis ² – in the two chambers of the legislature, in government, the security sector and in state-owned companies. While the APRA paved the way for elections in 2005 and for a de-ethnicisation of politics (Reyntjens, 2016), it could not prevent the gradual return to an increasingly authoritarian regime controlled by a group of (exclusively Hutu) generals within the dominant CNDD-FDD party, the former Hutu rebel movement led by incumbent President Pierre Nkurunziza. In 2015, Nkurunziza’s contested third term candidacy provoked popular protests and a serious split within CNDD-FDD. This culminated in a failed coup attempt and a major humanitarian crisis. A 2018 constitutional reform formalised the gradual erosion of ethnic power sharing without, however, completely removing Burundi’s consociational institutions (McCulloch and Vandeginste, 2019). In short, the 2017 legal reform faced ONGEs with the “dilemma of recognition” of ethnic identity in a context where, seventeen years before, recognition and accommodation of ethnic segmentation were an essential part of the negotiated solution (King and Samii, 2018).

A second contextual element of relevance here is Burundi’s historical path dependency in terms of state-party relations and political governance. Both old and new political elites focused on controlling the State and “capturing the associated rents to sovereignty” (Nkurunziza, 2018: 4) and the new leadership – largely made up of victims of repression by past Tutsi-dominated regimes – soon “fell back onto the reproduction and extension of many pre-conflict governance practices” akin to those that made it initially take up arms (Burihabwa and Curtis, 2019: 566). Both under single party UPRONA rule and – after a period of institutional collapse during the civil war – under dominant party CNDD-FDD rule, political and military power entails a guaranteed access to education, employment, wealth, security, impunity, social prestige, and so on. After successive electoral victories in 2005, 2010, and 2015, CNDD-FDD has secured almost complete control over state and society, also by either disbanding or co-opting and silencing political opposition and local civil society, including media (International Crisis Group, 2016). The very idea that the ONGE sector, which employed an estimated 4,000 local staff (EurAc, 2018), remains immune from dominant party interference is therefore at odds with a long-standing political culture that currently prevails more than ever since the formal abolishment of one-party rule in 1992. At the very top within the party, power is in the hands of a nucleus of generals with “maquis legitimacy” operating outside the formal (officially still bi-ethnic) army commandment structures (Rufyikiri, 2017). The generals’ decision announced by the National Security Council (NSC) – officially a merely advisory body – to suspend all ONGEs was a rare moment of transparency about the real locus of power in Burundi. The announcement read out on television on 27 September 2018 ³ by NSC chair General Silas Ntigurirwa revealed how matters of vital interest for the regime are decided outside the institutional realm.

Thirdly, the ONGE reform came about in the context of increased tensions between Burundi and its traditional aid partners. More generally, while reaching out to new partners less vocal about human rights and democracy, aid-dependent Burundi has gradually become more and more isolated at the international level. The government accused both neighbouring Rwanda as well as Western countries, in particular Belgium, of sponsoring (failed) regime change in Burundi at the time of the 2015 third term crisis (République du Burundi, 2016: 12). In March 2016, the EU adopted “appropriate measures” under Article 96 of the EU-ACP Cotonou Agreement. These aid sanctions (which is indeed how the government perceives them) were called for by a number of ONGEs, present in Burundi. Among them were the member organisations of the European Network for Central Africa (EurAc) (EurAc, 2015). In addition, to attenuate the impact of the sanctions on the population, several donors announced they would henceforth provide aid directly to the population through NGOs. In other words, ONGEs – or those among them with an interest in political governance – were seen as actively taking part in and benefitting from an international conspiracy against the government, both by calling for international sanctions and by “diverting” donor money that would normally be part of bilateral aid. This explains a number of measures – other than the ethnic quota requirement – that were imposed by the 2017 ONGE law. ONGEs must deposit one third of their programme budget to a foreign currency account at the Central Bank; overhead costs should not exceed 35 per cent of the annual budget (Article 16) and a 35 per cent tax is levied on local staff salaries (Article 39 of the Law of 23 January 2017).

Analytical Framework

In the inaugural lecture he delivered on 19 September 1978, John Griffiths introduced a typology of the effects of legal rules, distinguishing direct, indirect, independent, and unintended effects (Griffiths, 1979). Direct effects refer to changes in the behaviour of the addressees of the legal rule. Primary direct effects concern the behaviour of primary addressees, often the general public, here ONGEs active in Burundi. Secondary direct effects refer to the behaviour of secondary addressees, namely state agents responsible for monitoring, enforcing, and sanctioning the (non-)compliant behaviour of primary addressees. For instance, the legal rule prohibiting alcohol while driving supposedly has primary direct effects on the behaviour of anyone driving a car and secondary direct effects on the actions undertaken by the police. Indirect effects are the intended social consequences (e.g. road safety) of the direct effects (e.g. drivers do not drink alcohol). They are the stated policy goals and societal changes a legislator uses to explain and justify the adoption of the legal rule. Independent effects are intended but occur independently of the rule-conform behaviour. For instance, prohibiting alcohol while driving may boost the electoral support for the ruling party; adopting anti-corruption legislation may attract more foreign aid. Finally, legal rules may generate unintended effects, which the legislator did not seek to attain. Within this category, I introduce the additional distinction De Zwart (2015) makes between unanticipated versus anticipated unintended effects. Prohibiting alcohol while driving may have unintended but anticipated negative effects on the turnover of pubs. As explained below, the 2017 ONGE ethnic quota requirement produced unintended effects, some of which were probably unanticipated by the legislator while others were anticipated and traded off against intended (indirect or independent) effects.

I use this typology as an analytical tool to unravel and structure what is at stake in the 2017 ONGE ethnic quota requirement. Who are the main actors involved in the legal reform? What were their motivations and perceptions? How was the legal measure implemented (or not)? What short-term dynamics did it produce? The use of Griffiths’ typology allows for integrating a variety of perspectives on the ethnic quota requirement and, thus, for capturing the diversity of meanings the reform also had in reality. Focused on the legal measure, the run-up to it and its immediate aftermath, the use of this analytical tool obviously does not allow a longitudinal impact evaluation of the reform on, for instance, horizontal inequalities between ethnic groups in Burundi. This would require a different methodology and temporal scope. Finally and to avoid confusion, Griffiths used this typology for an ultimately different purpose than mine. He developed it as a heuristic technique for his theory and critique on legal instrumentalism, which he defined as the (naïve) belief that legal rules lead to rule-conform behaviour of the addressees (direct effects) and, thus, produce the societal consequences desired by lawmakers (indirect effects). This was an important foundation and stepping stone for his later well-known work on legal pluralism, namely the presence of multiple (state and non-state) legal orders that determine people’s behaviour in one social field (Griffiths, 1986). In this paper, I use Griffiths’ typology of effects of law without, however, adopting a legal pluralist perspective.

Direct Effects: The Law’s Instrumental Obscurity in Terms of the Expected Behaviour of its Addressees

In line with the analytical framework, I distinguish between primary and secondary direct effects. A striking feature of the 2017 ONGE quota requirement is the extremely vague character of the legal provision both in terms of expected behaviour of the primary addressees and of the officials in charge of rule enforcement. Two important consequences are worth mentioning upfront. On the one hand, the obscurity created room for arbitrary enforcement. On the other hand, it suggests that an important (if not the main) motivation of the lawmaker were the independent effects – which, as noted above, arise independently of the rule-conform behaviour of the addressees – rather than the indirect effects, which do depend on the rule-conform behaviour and therefore require clarity on the expected behaviour of the primary and secondary addressees.

In terms of its primary direct effects, the 2017 ONGE law stipulates that “Hiring of local staff must be done in accordance with the ethnic and gender balances provided in the Constitution of the Republic of Burundi.” ⁴ On at least three accounts, this provision lacks clarity. First, the Constitution does not include any ethnic and gender balance for ONGEs or, more broadly, for private sector employment. The provision is therefore strictly speaking void and meaningless. Secondly, the provision refers to hiring of staff, suggesting it affects only new vacant positions, not existing staff. This reading is a contrario confirmed by the only other provision that refers to local staff and imposes a tax on the salary of all Burundian employees hired and employed by ONGEs (Article 38). Thirdly, the law does not explain how ONGEs are supposed to implement the provision. How to register ethnic identity without undue interference with privacy rights of staff? ⁵ If the existing situation is not in accordance with the required ethnic balance, should ONGEs dismiss staff? In response to these and other questions raised by ONGEs, journalists and diplomats, the meaning and scope of the measure was gradually clarified through a combination of public and informal statements by government and party officials. This in itself was, not surprisingly, perceived by many ONGEs as a confirmation of the arbitrary and intimidating nature of the decision. The measure was said to be of immediate application to all existing staff. Ethnic quotas to be applied were those used for political positions and state-owned companies (60 per cent Hutu, 40 per cent Tutsi), in combination with a 30 per cent quota for women. Furthermore, ONGEs who argued they did not know the ethnic identity of their staff were told: “If you pretend you don’t know, we will tell you because we know.” In terms of secondary direct effects, the law announces the establishment, by presidential decree, of an Inter-Ministerial Follow-Up Committee. This Committee shall report to the existing National Aid Coordination Committee (CNCA), which advises the Foreign Minister who decides on continuation or suspension of the ONGE (Article 37). Within six months, ONGEs must comply with the new rules (Article 38).

The reaction of ONGEs varied. While at least one ONGE conducted a census asking staff to fill out a form mentioning ethnic self-identification, ⁶ many other ONGEs – at least initially – refused to comply. None of the ONGEs, however, challenged the provision in court. According to an ONGE official, it was impossible to find a lawyer who dared taking the matter to court. Instead, Rassemblement, Echanges et Solutions entre ONG (RESO) – the network of most (but not all) ONGEs in Burundi which acts as an interlocutor with the authorities since 1999 – proposed a dialogue with the government to develop a charter, acceptable to all parties, outlining the implementation and enforcement modalities (i.e. the secondary direct effects) of the new law. Interestingly, Foreign Affairs Minister Nyamitwe went along with the idea of a negotiated charter, thus taking some distance vis-à-vis the more hardliner positions on government side. By December 2017, however, the dialogue came to an end without an agreed charter. For RESO, the main stumbling blocks were the proposed establishment of two types of recruitment committees – one internal committee within every ONGE and one governmental recruitment committee established in each of the 18 provinces, appointed by the Minister of the Interior and in charge of approving every new local staff recruitment – and the severe sanctions in case of false statements about ethnic identity. In April 2018, the Foreign Minister was replaced and his successor never reactivated the dialogue with RESO on an implementation charter. Around the time of the constitutional referendum in May 2018, the ONGE crisis seemingly calmed down. However, because no presidential decree had been adopted on the implementation of the 2017 law, ONGEs were left in legal limbo, acutely aware that they could be called to order at any time.

In September 2018, the crisis flared up. On 12 September 2018, the Senate announced it would visit all ONGEs to verify the composition of their local staff. Absent any legal basis in the 2017 ONGE law and making use of the obscurity in terms of its secondary direct effects, the Senate referred to its constitutional powers to verify the ethnic and gender balance in the public administration, stressing that ONGEs are subject to the same constitutional rules. This was, at the very least, a highly creative reading of both the Constitution and the 2017 ONGE law. Next, as mentioned above, on 27 September 2018, the National Security Council on television announced that, as of 1 October, all ONGEs were suspended for a three-month period because of their failure to respect the 2017 ONGE law. At a meeting on 2 October, Interior Minister Barandagiye – again without a ministerial ordinance or any other written text – clarified the decision, which, he argued, was taken by “a supreme organ situated above all ministries and chaired by the Head of State,” an empirically correct but legally nonsensical statement. ⁷ To be readmitted, ONGEs must produce four documents: (1) a cooperation agreement with the foreign ministry, (2) a technical protocol with the line ministry, (3) a commitment vis-à-vis the finance ministry to respect Burundi’s banking legislation, and (4) a plan to correct ethnic and gender imbalances within three years. The fourth document was the most controversial and, although this provided somewhat more clarity on the primary direct effects of the ONGE ethnic quota requirement, obscurity continued to prevail. While a template and a circular by Finance Minister Ndihokubwayo further clarified the third required document, no further instructions were given with regard to the fourth. At least three interpretations prevailed among ONGEs and their donors. According to the strictest reading, the fourth document should list all names of local staff with their gender and ethnic identity. For most ONGEs and European states funding them this was unacceptable. According to a second reading, the document should merely mention ethnic and gender statistics, without listing individual staff. For at least one European donor (but not so for several others nor for the United States), submitting ethnic statistics also crossed a red line with potential negative implications on future ONGE funding. Using a third interpretation, several ONGEs refused to provide details on the ethnic composition of existing or future staff and merely committed to developing a new recruitment procedure without explicit ⁸ references to ethnicity. RESO developed two versions of a standard template – one on the basis of the second, the other on the basis of the third reading – that however failed to obtain consensus among its members.

Did the ethnic quota requirement effectively produce the desired primary effects, inducing compliant behaviour by ONGEs? I have not been able to determine how many ONGEs submitted either lists detailing ethnic identity of their local staff or “anonymous” lists with ethnic statistics. In fact, none of the ONGEs that were readmitted – 84 before the 1 January 2019 deadline, 93 by the end of March 2019 – published the four documents or clarified what kind of “fourth document” they submitted to the Burundi authorities. Some ONGEs obtained their readmission allegedly without submitting an “ethnic fourth document.” For some, this was due to the fact that they did not have any local staff. Others benefitted from a remarkable and exceptional week of flexibility early November 2018 when several ONGEs were readmitted – according to some to their own surprise – without providing ethnic data on their local staff. Still others, like Doctors Without Borders (MSF), were allowed to continue operations, allegedly without submitting an “ethnic fourth document,” for reasons related to the law’s unintended effects explained below. Finally, unsubstantiated rumour had it that some ONGEs informally submitted “ethnic” annexes to their “non-ethnic” fourth document and even that some ONGEs added a financial incentive to support their file. The decisions taken by the Interior Minister to either accept or refuse re-admission were poorly motivated and do not allow to determine the criteria that were used to assess the applications. They consisted of no more than a short, two-paragraph letter to the ONGE, either noting that “all requested documents have been submitted” (in case of re-admission) or that “the documents submitted do not meet our expectations, because they neither reflect the current situation nor the steps that will be taken to correct possible imbalances in light of the constitutional quotas” (in case of refusal). ⁹ When closing their office in Burundi, some ONGEs – including Handicap International, Attorneys without Borders, 11.11.11, and Réseau des Citoyens – released a public statement stating their inability, for deontological reasons, to abide by the ethnic quota requirement.

In May 2019, more than two years after the adoption of the 2017 ONGE law, the cabinet of ministers discussed three presidential decrees specifying its secondary direct effects, namely on the establishment of an Inter-Ministerial Follow-Up Committee and a Single Window – as required by the ONGE law – and on the establishment of Recruitment Committees – as suggested in the Charter unsuccessfully negotiated between RESO and the Foreign Minister (see above). In June 2019, before the decrees were even signed, the Finance Minister requested all ONGEs to fill out a template for each of their projects, including the ethnic composition of their local staff. ¹⁰ Once again, the legal basis for this request was unclear. In short, as of June 2019, three different ministries (Interior, Foreign Affairs, and Finance) and the Senate were monitoring compliance with the ethnic quota requirement.

Few interlocutors I interviewed attributed the vagueness of the ONGE ethnic quota legislation to the government’s ill preparedness and the lack of time or technical expertise to draft and explain the measure. Most interlocutors pointed at the deliberate and instrumental nature of the obscurity in terms of both primary and secondary direct effects of the ONGE ethnic quota requirement. By omitting to provide more written guidance to ONGEs on how to comply with the law and by having a plethora of self-declared monitoring bodies, ¹¹ the government created room for arbitrariness to either readmit or expel ONGEs and effectively stirred uncertainty, division, and self-imposed restraint among ONGEs, thus producing one of the independent effects mentioned below.

Indirect Effects: The Contested Justification of the Ethnic Quota Requirement as Affirmation Action

What societal goal did the government seek to attain? During the parliamentary debate and in the draft Charter, it was explained that affirmative action was the policy aim that motivated the quota requirement: redressing a legacy of discrimination along ethnic lines and, thus, promoting social cohesion and reconciliation. In a joint statement, the United States, Canada, Switzerland, Japan, the EU and its member states and UN agencies recognised “Burundi’s aspirations to reach a labour force that reflects the country’s social diversity.” ¹² US ambassador Casper, considered quite supportive of the reform by her European peers, tweeted: “From the debate in Burundi over NGO hiring, I have come to see that what is at stake with #reconciliation is not just about healing memories but also about who accesses opportunities and how.” ¹³ Two obvious questions arise. Was unequal access to ONGE employment indeed a problem? If so, why did no ONGE publicly endorse the “affirmative action”-frame (as diplomats did in their statement) and why did many ONGEs reject the quota-based solution?

Burundi’s civil war was grievance-based and distributional. The first rebel movement, established some years after the 1972 “selective genocide” (Lemarchand and Martin, 1973) against Hutu, was not coincidentally named after “the need to liberate the Hutu” (PALIPEHUTU – Parti pour la libération du peuple hutu). Horizontal inequalities affected all societal domains (security sector, political institutions, judiciary, higher education, public sector employment, etc.) and, as noted above, were both regional and ethnic. Regarding employment, Nkurunziza (2018: 219) notes that, as of 2000, 89 per cent of all state-owned company managers were Tutsi (11 per cent Hutu) and 61 per cent were from southern Bururi province (39 per cent from the rest of the country). Although hard and updated quantitative data are lacking, extensive field research conducted in 2006 on thirty-three ONGEs (all of them RESO members) shows that local ONGE staff was overwhelmingly Tutsi (Munezero, 2008), not because of a deliberate policy of ethnic discrimination by ONGEs but because of a longstanding practice of clientelism – hiring of new staff from networks of existing staff – and because of the availability of generally better trained Tutsi applicants with the required medical, legal, or other expertise, in turn a result of decades of ethnic discrimination in higher education (Lemarchand, 1994: 138). Of critical and existential importance was – and remains ¹⁴ – the ethnic cleansing at the public Université du Burundi (UB) in 1995. Several among the current Hutu leadership – many of them also “orphans of 1972” as they are sometimes called – were students who narrowly escaped the 1995 UB massacres, including CNDD-FDD secretary-general Ndayishimiye, Interior Minister Barandagiye, Attorney General Bagorikunda and former Foreign Affairs Minister Nyamitwe, who published his own account of the events (Nyamitwe, 2006). Research conducted eight years after the first post-conflict elections won by CNDD-FDD, the former Hutu rebel movement, shows that 82 per cent of UB teaching staff – essentially made up of former UB students and assistants – were Tutsi (Nsaguye, 2013: 57).

While – admittedly few, probably due to the controversy surrounding the 2017 law – intellectuals like Batungwanayo (2018) and Cishahayo (Minani, 2019: 158) recalled the historical context and underscored the need to address the discrimination, many ONGEs – mainly European, less so American – did not “buy” the affirmative action frame of the 2017 quota requirement as its real motivation. Two main reasons stand out: the absence of a real policy and the risk of abuse.

First, the ethnic quota requirement was laid down in just one sentence in the 2017 ONGE law. It was an isolated measure, imposed twelve years after the CNDD-FDD came to power and not integrated in a broader policy on affirmative action and the use of ethnic quotas. Not involved in the APRA negotiations, CNDD-FDD always expressed reluctance against the use of quotas. In an interview in October 2004, ten months before taking the oath as President of Burundi, CNDD-FDD leader Pierre Nkurunziza stated that “CNDD-FDD is hostile to quotas,” arguing that conflict, politics and society in Burundi should not be reduced to ethnic divides and hoping the APRA-based quotas would soon be removed with the help of civil society. ¹⁵ Since the first post-conflict elections in 2005, no public debate was ever held on affirmative action in private sector employment, unlike in other countries like Malaysia (where quotas were used) or South Africa (where targets, rather than quotas, were used) that developed policies to redress a legacy of identity-based discrimination in employment and education (Lee, 2016). The 2014 National Policy Document on Employment – seventy pages long – identifies age-based, gender-based, and disability-based discrimination but remains completely silent on ethnicity-based discrimination (République du Burundi, 2014). The 1993 Labour Code – which has not been revised – prohibits distinction, exclusion or preference based on ethnic origin in the area of employment, promotion or remuneration (Article 6). The 2011 Law on Higher Education merely contains a general non-discrimination clause (Article 7), without any reference to measures aimed at compensating historical discrimination in terms of access to tertiary education. In other words, the ONGE ethnic quota requirement was not at all embedded in a policy based on prior dialogue, expert consultations, parliamentary hearings, and a consensus across political and ethnic lines. To defend the quotas, officials occasionally referred to the “spirit of Arusha” – which was convenient in terms of public discourse and for strategic political reasons, as explained below – and to the indeed rather systematic use of the 60/40 per cent quotas for the composition of senior bodies that, constitutionally speaking, are not subject to the ethnic quota requirement, such as the Electoral Commission, the Truth and Reconciliation Commission, and the National Observatory for the Prevention and Eradication of Genocide. It is important to note, however, that the ethnic quotas laid down in the APRA were, above all, a typically consociational mechanism seeking minority protection (in the Burundian case: benefitting the Tutsi demographic minority) rather than an affirmative action mechanism benefitting the historically disadvantaged group (in the Burundian case: the Hutu demographic majority) (McCulloch and Vandeginste, 2019: 1181). Given this background, what does the APRA tell on employment and the use of ethnic quotas? On public sector employment, the APRA explicitly states that recruitment must be based on objective criteria of aptitude and on “the need to correct imbalances and achieve broad representation” (Protocol II, Chapter 1, Article 10). This dual principle was incorporated in the 2005 Constitution, without the use of quotas for civil servants (Article 143) but – in line with the APRA – with powers for the Senate to monitor the implementation of the constitutional aim to achieve a more balanced ethnic and gender representation in all state institutions, including the public administration (Article 187). On private sector employment, the APRA contains no specific provision but implicitly discourages quotas. Its general principles and measures on the fight against exclusion seem to apply here, namely the “deliberate promotion of disadvantaged groups, particularly the Twa, to correct existing imbalances in all sectors. This exercise shall be conducted, while maintaining professionalism and avoiding the quota system […]” (Protocol I, Chapter 2, Article 7 – emphasis added).

Secondly, the ONGE law of January 2017 was adopted amidst internationally mediatised fears of escalation and re-ethnicisation of conflict in Burundi (Purdekova, 2019). In May 2016, International Crisis Group noted that “the Tutsi community stills feels persecuted and there is a widespread fear that the government plans a genocide” (International Crisis Group, 2016: 13). In September 2016, the UN Independent Investigation on Burundi noted “the general trend of ethnically divisive rhetoric by the Government, which may carry a serious potential of the situation spiralling out of control” (United Nations, 2016: 13). In November 2016, FIDH and Iteka published a 200-page report entitled “Repression and genocidal dynamics in Burundi,” as a result of which Iteka – Burundi’s oldest human rights league established in 1991 – was banned. Also in November 2016, the Senate launched an ethnic census of all civil servants, which again gave rise to the use of the term genocide in international media coverage. ¹⁶ Given this context and its use in the communication strategy by opposition actors (Vircoulon, 2018: 12), several ONGEs did not want to take the risk of collecting data that, so they feared, might at some point be used to target Tutsi among their local staff. One ONGE director I met referred to “Aiding Violence,” Peter Uvin’s account of the shortcomings of development actors in the run-up to the 1994 genocide in Rwanda (Uvin, 1998). It is worth noting that this ONGE risk aversion was clearly inspired by the climate of fear prevailing after the 2015 failed coup attempt. Five years before the adoption of the 2017 ONGE law, during Round 5 (2011–2013) of its national survey on democracy, governance and society, Afrobarometer (and its local NGO partner) asked 1,200 Burundian respondents for their ethnic identity. At that time, this did not cause any controversy. ¹⁷

In summary, while the legacy of pre-war ethnic discrimination surely warranted affirmative action and while some officials like former Foreign Affairs Minister Nyamitwe had personal reasons to award priority to it, the context made many ONGEs seriously doubt as to whether this stated indirect effect was indeed the government’s (read: the generals’) sole or main or even marginally relevant motivation.

Independent Effects: Dominant Party Interests Motivating the Ethnic Quota Requirement

Laws may produce effects that do not depend on the conforming behaviour of the primary addressees. Nevertheless, they may have important explanatory value for why the legislator enacted the law. I introduce three independent effects of the ONGE ethnic quota requirement: boosting electoral support for CNDD-FDD; weakening civil society; and protecting national security and sovereignty.

First, the ONGE ethnic quota requirement serves the agenda of CNDD-FDD ahead of the 2020 general elections. It enables the party to boost its image as the defender of the Hutu electorate’s interests. In February 2019, CNDD-FDD secretary-general Ndayishimiye declared that there is no longer any discrimination between Hutu, Tutsi, and Twa and that everybody is now entitled to an equal share in the public goods, an implicit but clear reference to the party’s achievement to end pre-war Hutu discrimination. ¹⁸ Furthermore, the adoption of the law sent the clear message that CNDD-FDD holds the key to accede to the financially interesting and socially prestigious jobs (Munezero, 2008: 16) within ONGEs, just like it does for jobs in the civil service. The proposed involvement of provincial committees, appointed by the minister of internal affairs, in the hiring of new ONGE staff reinforced this effect. Furthermore, to defend the ONGE quota requirement, some party officials and government supporters cleverly referred to the “Arusha spirit,” as noted above. This allowed them to reveal the inconsistency and undermine the credibility of the political opposition and civil society who, since 2015, constantly wave the APRA as the gold standard of peace in Burundi […] but who refuse to apply its spirit to the employment of local ONGE staff (Ruvyogo, 2019).

Secondly, for many international non-governmental (see e.g. EurAc, 2018) and intergovernmental (see e.g. European Parliament, 2017) organisations, the adoption of the 2017 law served the purpose of establishing tighter control on ONGEs and shrinking space for civil society more generally. The International Center for Not-for-Profit Law expressed concern at the “intrusive regulation” and “severe limitation on the independence” of ONGEs (ICNL, 2017). While part of a wider continental and global trend (Buyse, 2018; Civicus, 2018: 12), Burundi’s new legal ONGE framework “stands out for its unabashedly invasive framework” according to Freedom House (Musila, 2019: 11). Government officials themselves also occasionally referred to the need to reduce ONGE “disorder” and “chase ONGEs whose activities are contrary to Burundian culture,” ¹⁹ and to stop ONGEs from “replacing the State.” ²⁰ The adoption of the ONGE law was thus in line with earlier measures vis-à-vis domestic civil society, in particular the banning or suspension of ten local – mostly human rights – NGOs in October and November 2016. The most extreme form of control was obviously the non-readmission (de facto expulsion) of some ONGEs. Less extreme, but equally effective was the pre-emptive self-restraint adopted by the remaining ONGEs, in particular those with an interest in political governance. Faced with the repeated announcements that officials will monitor them and visit their premises, some ONGEs understandably feared being blamed for “undesirable” activities or being caught with “subversive” documents. Before leaving Burundi, at least one ONGE destroyed its own archives. In addition, the ethnic quota requirement strongly divided ONGEs thus reducing their collective bargaining power. Furthermore, the law pulled all ONGEs – including those without an explicit interest in political governance – into a political debate and heavily affected their legitimacy. ONGEs who were readmitted – none of them communicating transparently on their application – were suspected of collaborating with the regime and prioritising business turnover over staff security. On 4 January 2019, the popular anonymous account @iBurundi tweeted “We are looking for the names of the 84 foreign NGOs who accepted to ignore Humanitarian Principles in order to accommodate @BurundiGov. There will come a time when the NGOs might regret their decisions.” ²¹

Thirdly, at least for some officials, the ONGE law was also inspired by security considerations. It allowed the government, if necessary, to remove those ONGEs reluctant to abide by the ethnic quota requirement, also the prime suspects of employing mainly Tutsi and of acting as Trojan horses delivering information to international watchdogs (most notably the International Criminal Court, which opened a preliminary examination in April 2016, and UN investigators). The very fact that the National Security Council – not a body in charge of affirmative action – suspended ONGEs stands witness to this motivation. Furthermore, the suspension was decided immediately after the UN Commission of Inquiry, established by the UN Human Rights Council after the 2015 crisis, published a well-documented report concluding that serious human rights violations, including crimes against humanity, persisted in 2017 and 2018 (United Nations, 2018b). According to the then Burundian ambassador in France, ONGEs are not in Burundi to promote development but to protect the interests of their funding agencies. ²² Since the 2015 crisis, some ONGEs and their Burundian partner civil society organisations – in particular those perceived to be controlled by an urban Tutsi leadership suspected of loyalty to the old UPRONA regime (Bertelsmann, 2018: 29) – have been accused of taking part in a Western conspiracy stirring regime change in Burundi (Kavakure, 2016: 148). According to two government officials I interviewed, at the height of the popular demonstrations against President Nkurunziza’s third term in April to May 2015, at least two ONGE vehicles transported weapons. Security considerations did not only inspire government action vis-à-vis ONGEs. They may also explain why the local office of the UN High Commissioner for Human Rights was closed (Rufyikiri, 2019: 6). On 3 June 2019, the Interior Minister suspended governance watchdog PARCEM, arguing that the latter’s activities “tarnish the image of the country and its leaders, in order to disturb peace and public order.” ²³

Unintended Effects: A Reform Spiralling Out of Control?

Finally, laws often generate unintended effects. While not the purpose of the law-making activity, they are part of the new social reality shaped by the legislation. As De Zwart (2015: 292) rightly notes, a lack of intention is not necessarily identical to a lack of anticipation. Indeed, while some effects may catch the legislator by surprise, other unintended effects may be unwelcome but foreseen side effects. They are traded off against intended (indirect or independent) effects. Applied to the ONGE ethnic quota requirement, while it is not always clear which of the unintended effects were anticipated and which were not, some were clearly not due to errors or ignorance but rather to a rational cost-benefit analysis.

In October 2018, the UN Secretary-General attributed the suspension of a voluntary repatriation programme of Burundian refugees and the downscaling of humanitarian support to 70,000 Congolese refugees in four camps in Burundi to the suspension of ONGEs (United Nations, 2018c: 7). This illustrates a first unintended effect, namely the short-term impact of the ONGE suspension on the humanitarian situation. Furthermore, a number of longer-term humanitarian and developmental risks were associated with the possible departure of some ONGEs the Burundi government badly needed. Some of these ONGEs were also involved in the implementation of development programmes of intergovernmental actors. One critically important partner – even more so after the Ebola outbreak across the border in neighbouring DRC and the cholera outbreak in Rumonge town in December 2018 – was Doctors without Borders (MSF). Interestingly, allegedly without introducing an “ethnic” version of the fourth document required by the Interior Minister, ²⁴ MSF continued its activities pending its request for readmission (which it had not received by the 31 December 2018 deadline ²⁵ ) and after receiving its readmission in March 2019.

Although it is hard to tell how many actually used them, a number of ONGEs considered the use of evasion mechanisms, a second unintended effect. These included: replacing local staff by local consultants; registering as a foundation rather than as an ONGE; and relocating the ONGE office to a neighbouring country and implementing programmes through partnerships with local NGOs – not (or not yet?) subjected to the same ethnic quota requirement – rather than by Burundian staff. One other evasion strategy, namely the hiring of international staff to replace local staff, was blocked from the very start. Another provision of the 2017 ONGE law allows the hiring of international staff only when the required expertise is not locally available.

A third unintended effect was the negative impact on the government’s relationship with its bilateral aid partners. As noted above, donors funding ONGEs were forced to draw some red lines, which they did not want ONGEs to cross. So the ONGE crisis indirectly also amounted to some arm-wrestling between the government and its bilateral aid partners. More importantly and probably unanticipated by the government, the ONGE ethnic quota requirement weakened the position of those diplomats and civil servants who internally – that is, within their ministry or department – called for a gradual end to the “appropriate measures”-regime (aid sanctions) adopted under Article 96 of the EU-ACP Cotonou agreement in March 2016.

Finally, an outstanding question is what effect the measure may generate on social relations in Burundi. As noted in the introduction, the APRA adopted ethnic quotas to share positions in the senior political and security sphere. This institutional engineering significantly contributed to a de-politicisation of ethnicity, a de-ethnicisation of political mobilisation (Van Acker et al., 2018) and a de-ethnicisation of electoral violence (Colombo et al., 2019: 335). The non-consensual extension of the use of ethnic quotas to the ONGE sector, however, puts ethnicity back at the heart of political contestation and, at societal level, of competition for employment. Since the 2015 crisis, there has been a creeping re-ethnicisation of public discourse in various spheres of society (see i.a. Paviotti, 2019; Van Acker et al., 2018: 87) and ethnicity continues to have major mobilising power at societal level given Burundi’s history of violent conflict along ethnic lines (Gatugu, 2018: 60). During the National Assembly debate in December 2016 on the ONGE bill, some members of parliament expressed a concern about the risk of ethnicisation of society, to which the Interior Minister replied that ethnic balancing has shown to be a salvation for the evil of ethnic exclusion. ²⁶ However, one ONGE with some 60 local staff experienced internal tensions along ethnic lines when complying with the 2017 ONGE law. While it is too early to draw firm conclusions, it is clear that if the ONGE ethnic quota requirement contributes to a re-ethnicisation of society, this will – at best – have been an anticipated unintended effect.

Conclusion

Understanding the multiple meanings of the ethnic quota requirement imposed on ONGEs active in Burundi is complex. Despite the undeniable legacy of ethnic discrimination and historical injustice against the demographic Hutu majority, it can clearly not be reduced to one single “affirmative action”-motivation on behalf of the lawmaker. For a variety of reasons, including the tense political climate around the time of its adoption and early implementation, it has been perceived differently and with great suspicion by ONGEs, the very large majority of which have abstained from communicating publicly about their response to the requirement, thus contributing to the controversy around the measure.

A striking observation is the obscurity of the wording of the law, in particular of the desired behaviour of its primary and secondary addressees. There are various reasons to believe that this obscurity was not accidental but instrumental. First, two and a half years after its initial adoption, the exact legal requirement and the consequences in case of non-compliance have not been clarified in writing. Several ministries, the Senate and the National Security Council have seized the absence of a clear legal framework to grant themselves monitoring and enforcement powers, thus confronting the remaining ONGEs without a continuous threat of arbitrary and intrusive controls. Secondly, the obscurity enabled government actors to conceal internally divergent perspectives while, at the same time, combining and reconciling their different agendas (as reflected in the independent effects). Furthermore, learning from previous experiences, most notably the 2015 third term crisis (Vandeginste, 2016), the Burundi authorities may well have professionalised themselves in organising and exploiting legal loopholes in politically sensitive matters.

The analysis also shows that context matters enormously to understand the use of ethnic quotas. During the APRA peace talks, ethnic quotas were agreed upon for Burundi’s political and security institutions as part of a minority protection mechanism. However, the same quotas, extended to ONGE employment, are currently perceived by ONGEs with an interest and expertise in political governance as a potential threat for members of that same demographic Tutsi minority. Without wanting to analyse the counterfactual scenario, it is likely that the inclusion of such quotas in the APRA or their adoption as a result of a wider parliamentary debate on affirmative action after the first post-conflict elections would have led to very different reactions on behalf of ONGEs and donors. In the current context, however, both the timing of the legal reform and the absence of a broader policy on affirmative action suggest that the government’s reference to a legacy of discrimination amounted above all to the skilful use of a discursive frame which it drew from an internationally mediated peace agreement and which served the purpose of justifying a reform that was primarily intended to further reduce civic space and enhance authoritarian control over society, thus mirroring a wider international trend. Intrusive laws and policies imposed on domestic and on international civil society groups are indeed by no means unique to the case of Burundi. Restrictions on the funding, the registration, the permission to conduct activities and the independence of NGOs and watchdogs in the region (see, e.g. Smidt (2018: 5) on Kenya, a country also under investigation by the International Criminal Court) as well as the international donors’ response to shrinking civic space may have sent the signal to the Burundi government that, despite its aid dependency, it could safely adopt the 2017 ONGE legislation and get away with it.