I. Introduction
April 2020 marked the 26th anniversary of the worst mass killings in postcolonial Africa. In a televised address to the Rwandan people, UN Secretary-General António Guterres paid tribute to the ‘more than 1 million people [ … ] murdered in just 100 days during the genocide against the Tutsi’ (Guterres 2020). His words echoed the preamble to the Rwandan constitution which recalls ‘the genocide committed against Tutsi that decimated more than a million sons and daughters of Rwanda’.
Presumably, ‘more than a million’ refers to a census from 2000 that counted 1,074,017 genocide victims. However, earlier and later government-sponsored counts produced vastly different numbers. Hence certain questions arise. How were the counts conducted? Is the 2000 count more reliable than the others? How do Rwandan government number(s) compare with scholarly estimates? Can the official victim number(s) be reconciled with survivor numbers and demographic data? Besides answering these questions, I contrast Rwanda’s accounting for the death with Bosnia’s and propose an alternative method for estimating the number of Tutsi killed in 1994.
Some may question the relevance of revisiting the death toll after 25 years. A genocide took place against the Tutsi, and the exact death toll is a purely academic question. Of course it is, but it is also more. The Rwandan government has made the death of ‘more than a million’ Tutsi the foundation of the new Rwanda, where power is concentrated in the hands of President Paul Kagame and an inner circle of former exiles. At the same time, the accounting processes have been shrouded in secrecy; data files are kept under lock and key; and challenging the official death toll may lead to criminal prosecution.
Moreover, scholars revisit death tolls of wars and other humanitarian catastrophes all the time because politicians, militaries, rebels, journalists, activists, and, yes, academics inflate or deflate death tolls all the time (Best 2001; Andreas and Greenhill 2010; Pérouse de Montclos, Minor and Sinha 2016; Krause 2017). In sum, there are sound reasons to critically examine the official counts of genocide victims and survivors.1 A systematic review also provides insight into how history is written in the new Rwanda.
The article contributes to the literature on the Rwandan genocide (Prunier 1996; Reyntjens 1997, 2017; Des Forges 1999; Mamdani 2002; Dallaire 2003; Melvern 2004; Straus 2008; Fujii 2009; Guichaoua 2010); the consolidation of political power in post-genocide Rwanda (Reyntjens 2013; Thomson 2018); and the emerging field of demography of conflict and violence (Brunborg and Urdal 2005; Brunborg and Tabeau 2005).
Its contributions are threefold: instead of dismissing the official tallies as not credible, it exposes them; second, the article proposes an alternative to the counting methods so far used by Rwanda scholars; and third, it offers an explanation for why, demographics notwithstanding, the government insists that ‘more than a million’ died in the genocide against the Tutsi.
Importantly, the article considers only excess Tutsi mortality in 1994. It does not discuss excess mortality among Hutu and Twa. Some put the Hutu death toll in the hundreds of thousands (e.g. Reyntjens 1997; Davenport and Stam 2009).
My main data sources are Rwandan population censuses and health and fertility surveys; Rwandan government reports and briefings about the genocide; and scholarly estimates of genocide victims and the size of the Tutsi diaspora.
The article is organised as follows. Section II is a brief history of the conflict. Section III draws lessons from the controversy about the death toll of the Bosnian war. Section IV reviews scholarly estimates of Tutsi victims. Section V criticises these estimates. Section VI, the main thrust of the article, examines government censuses of genocide victims and survivors. Section VII presents my own calculation. Section VIII argues that ‘more than a million’ is not an actual count but a political statement. Section IX concludes the article.
II. Brief history of the conflict
The Rwandan civil war began in October 1990 when the armed wing of the Rwandan Patriotic Front (RPF) invaded Rwanda from neighbouring Uganda. The RPF was (and is) dominated by Rwandan Tutsi who had grown up in exile. The stated goal of the RPF was to end the Hutu-controlled one-party state and allow the return of Tutsi refugees.
A fragile, internationally brokered peace agreement held from August 1993 until 6 April 1994 when unidentified persons shot down an airplane carrying Rwandan president Juvénal Habyarimana, Burundian president Cyprien Ntaryamira, the chief of staff of the Rwandan armed forces, and three cabinet ministers (all Hutu). The attack triggered the resumption of the civil war and the beginning of the genocide against the Tutsi. Both ended that July when the RPF routed the government army and Hutu militias loyal to it. Paul Kagame, the commander of the RPF, and a small group of former Tutsi exiles have ruled Rwanda ever since.
A United Nations Commission of Experts reported in December 1994 that ‘an estimated 500,000 unarmed civilians have been murdered in Rwanda. That estimate indeed may err on the conservative side for [ … ] some reliable estimates put the number of dead at close to 1 million’ (United Nations 1994, para. 57). It concluded ‘that acts of genocide against the Tutsi group were perpetrated by Hutu elements in a concerted, planned, systematic and methodical way’ (Ibid., para. 183); and that ‘individuals from both sides to the armed conflict perpetrated crimes against humanity’ (Ibid., para. 182). The Commission said little about the ethnicity of the victims, other than that they were ‘mostly either of Tutsi origin or were Hutus considered to be moderate’ (Ibid., para. 40, emphasis added). Coincidentally, that same month the Rwandan Interior Ministry reported that more than 2 million had died (Guichaoua 2010, 435).
The next section draws lessons about the numbers game surrounding the Bosnian conflict. A comparison is instructive because both were ethnic civil wars in small countries with populations under 10 million; they overlapped in time; international criminal tribunals were established for both; and international organisations were on the ground after the conflict, though for much longer in Bosnia than in Rwanda.
III. Civil war, genocide, and death tolls: lessons from Bosnia
One of the wars that tore apart Yugoslavia in the 1990s was the three-way conflict in Bosnia and Herzegovina (‘Bosnia’) between Orthodox Serbs, Catholic Croats and Muslim Bosniaks. Western media in the summer of 1992 reported ‘ethnic cleansing’ and Serb ‘concentration camps’. The conflict raged until late 1995 and many civilians died, but not nearly as many as originally estimated.
The numbers game apparently began when Haris Silajdžić, the (Muslim) foreign minister of newly independent Bosnia, stated that if war breaks out, ‘200,000 to 300,000 people [could be] slaughtered within a few months’ (Gutman 1993, 7). War broke out in April 1992 and a year later the number of 200,000 deaths (‘mostly Muslim civilians’) began circulating in the Western media. The UN Security Council appointed a Commission of Experts to investigate possible war crimes. Its chairman reported in April 1995 that the Commission had ‘identified approximately 200,000 people killed’ (Bassiouni 1995, 24). The war formally ended in November 1995 with the signing of the Dayton Peace Agreement.
Two hundred thousand (civilian) deaths, mostly Muslim, became the consensus in Western media. Samantha Power (2002), for example, whose reporting from Bosnia earned her a Pulitzer Prize, wrote that ‘the ethnic cleansing and genocide against the country’s Muslims proceeded apace, and more than 200,000 Bosnians were killed’ (Ibid., 327).
The number would probably still stand had it not been for two initiatives. In March 1993 the UN Security Council had established the International Criminal Tribunal for the former Yugoslavia (ICTY). Once the tribunal was up and running, the chief prosecutor created a demographic unit within her office (Spirer and Seltzer 2008, 219). The unit’s final report counted 104,732 war-related deaths, of whom 25,609 were Muslim civilians (Zwierzchowski and Tabeau 2010, 17). Nearly a third (7000–8000) died in the Srebrenica massacre of July 1995.
The other initiative came from a non-governmental organisation (NGO) based in the Bosnian capital, Sarajevo. Launched in 2004 in response to the ‘playing with numbers’, the Research and Documentation Center (RDC) set out ‘to identify each single victim and prevent any type of manipulation of numbers’ (Ahmetasevic 2017). In 2007 RDC published The Bosnian book of the dead. The database contains 96,895 cases, each related to one victim who was killed, died another way in war-related circumstances or disappeared during the war. Mirsad Tokaca, the (Muslim) founder of RDC, received threats for his work. ‘It’s easier to manipulate numbers than names. [ … ] I’ve touched the myth that has begun to be created among my own nation’, he commented (Halimovic 2008; Nettlefield 2010). Experts commissioned by the project’s foreign sponsors have validated The Bosnian book of the dead (Ball, Tabeau, and Verwimp 2007).
The numbers games around the Bosnian war offer two lessons. The first is that the fog of war obscures. Most Western journalists covered the war from Sarajevo, where they may have overly relied on Bosniak sources. UN experts too are not immune to disinformation, or may lack expertise (Tabeau and Bijak 2005, 192). Government estimates call for even greater caution. The leadership in Sarajevo played the numbers game to win international support and delegitimise its enemies.
The other lesson is that the truth can be established if investigators are granted unimpeded access to the territory, people and official records. Therein lies the difference between Bosnia and Rwanda. Post-war Bosnia was – and still is – a de facto international protectorate after an internationally brokered peace agreement ended the war. The authorities in Sarajevo had little choice but cooperate with international administrators. The RPF, in contrast, decisively won its war, while the UN was seen as having ‘betrayed’ Rwanda during the genocide (e.g. Melvern 2000; Barnett 2002). If the international community wanted to play a role in post-genocide Rwanda, it would be on Kigali’s terms.
However, the RPF-dominated government unsuccessfully opposed the establishment of the International Criminal Tribunal for Rwanda (ICTR) (Scheffer 2012, 69–86). The ICTR in Arusha was institutionally linked with the ICTY in The Hague through a common chief prosecutor. William Seltzer, consultant to the ICTR, writes:
I offered a set of recommendations on the use of data and quantitative methods in the work of the Tribunal. These recommendations were essentially ignored, and the ICTR did not attempt to formally assemble quantitative evidence or commission demographic studies to complement the other evidence before the Tribunal. (Spirer and Seltzer 2008, 218)
IV. Early scholarly estimates: 507,000 to 800,000 Tutsi genocide victims
A half dozen scholars over the years have come up with casualty figures for the Rwandan tragedy, with those of Gérard Prunier (1996), Filip Reyntjens (1997) and Alison Des Forges (1999) becoming reference points for others such as Verpoorten (2005, 2014) and Verwimp (2013). Reyntjens and Des Forges were established Rwanda experts who advised the ICTR, while Prunier was a self-confessed newcomer. Rwandan government counts were not available when they published their estimates.
The baseline for their estimations is Rwanda’s general population and housing census of August 1991. RGPH2, as the census is known, put the total population two and a half years before the genocide at 7,158,000 (rounded) and the ethnic composition at 91.1% Hutu; 8.4% Tutsi; 0.4% Twa; and 0.1% ‘others’ (Rwanda 1994, 124).2
Prunier extrapolates the 7,158,000 of August 1991 to 7,776,000 in April 1994 (annual growth rate ±3.0%). This number is uncontroversial, but Prunier claims that the proportion of Tutsi was as high as 12%,
because the government systematically tried to underestimate the Tutsi population in order to keep its school and employment quotas low, and secondly because the Tutsi themselves often tried to pass themselves off as Hutu [ … ] to avoid discrimination. (Prunier 1996, 264)
Thus, according to Prunier, the Tutsi population in April 1994 was 930,000. Three months later, only 130,000 were left, and so Prunier arrives at around 800,000 Tutsi killed (1996, 264). The number stuck because he was ahead of Reyntjens and Des Forges; his book was in English and widely available; and 800,000 seemed a reasonable compromise between the 500,000 and one million mentioned by the UN Commission of Experts.
Reyntjens (1997) puts the total population in April 1994 at around 7,775,000 (annual growth rate ±3.0%) and the Tutsi population at around 800,000. Based on samplings at the communal and familial level, he estimates that 25% or 200,000 survived. His best estimate is 600,000 Tutsi killed.
Des Forges, the lead author of Leave none to tell the story: genocide in Rwanda, writes:
Although frequently said, no documentation has been presented to support this allegation [of undercounting Tutsi]. The 1991 data show Tutsi as forming 8.4% of the total population. This figure seems to accord with extrapolations from the generally accepted census data of 1952, taking into account the population loss due to death and flight during the 1960s and the birth rate, which was lower for Tutsi than for Hutu. (Des Forges 1999, 15)
Des Forges thus accepts the official 8.4% and estimates the number of Tutsi in Rwanda in April 1994 at 657,000 (annual growth rate ±3.0%). As for survivors, she takes the 130,000 from Prunier and adds 20,000 who fled to Congo and Tanzania. Her final number is 507,000 Tutsi genocide victims. It may be noted that Des Forges consulted about the death toll with William Seltzer, a former Chief of Demographic and Social Statistics in the UN Statistics Division (Spirer and Seltzer 2008, 218).
V. Critique of early scholarly estimates
The fog of genocide was still hanging over Rwanda when Prunier, Reyntjens and Des Forges published their estimates, and calculating the death toll was only a small part of their research. Mindful of these limitations, I want to point out some minor flaws in their numbers. My goal is not so much to criticise them as to lay the foundation for my examination of the government numbers in section VI and for my own estimation in section VII.
My first criticism concerns the population baseline and expected growth rate. The total Rwandan population in August 1991 was not 7,158,000 but 7,290,000 (including 50,000 foreigners). The higher number was established by the Post Enumeration Survey held one month after the general census (Rwanda 1993, 44). On the other hand, an estimated average annual growth of 3.0% is probably too high, for two reasons. First, while the growth rate between 1978 and 1991 had been 3.1%, a nationwide reproductive health survey conducted in 1992 showed a ‘brutal decline’ of the fertility rate in the previous four years. The number of births per woman dropped 17%, from 5.2 in 1988 to 4.3 in 1992 (Rwanda 1992, 32). Second, the RPF invasion and ensuing guerrilla war exacerbated an already dire humanitarian situation caused by crop failure, small-scale massacres, a severe economic crisis, and government austerity measures. By March 1993, one in eight Rwandans was internally displaced. Infant, maternal, and senior deaths must have been on the rise. I suggest that a combination of lower fertility and higher mortality reduced the estimated growth rate for the early 1990s to under 3.0%. At a rate of 2.75% the Rwandan population would have been 7.7–7.8 million by April 1994 (including 50,000 foreigners). The new number is very close to the ±7,775,000 of Prunier and Reyntjens: in both cases, their too low baseline is cancelled out by a slightly too high estimated growth rate. Though the errors are immaterial I mention them for the sake of accuracy.
My second critique concerns Prunier’s numbers. Despite being a newcomer to Rwanda, he was among the first to publish a book on the genocide. His estimate that 800,000 Tutsi died in the genocide is 30% and 60% higher, respectively, than the estimates of Reyntjens and Des Forges. On closer examination his appears to be the most speculative.
Prunier does not offer any proof of manipulation of the census results. RGPH2 (like RGPH1 of 1978) was conducted with considerable international technical assistance and oversight, leaving little room for systematic tampering. More importantly, a census answer did not determine one’s ethnicity. Census forms typically are destroyed once the field data have been coded and entered into a database. If there was ‘cheating’ with ethnicity, it was at the local level: Tutsi who managed to pass off as Hutu before communal officials received a ‘Hutu’ identity card from them. Those who carried such a card in 1994 reportedly escaped the slaughter (Uvin 2001, 166–167; Des Forges 1999, 238, 485, 538; Burnett 2009, 85). Hence, the question of how many were able to register as Hutu is irrelevant to estimating the death toll.
Nonetheless, there is reason to be sceptical about the official ethnic distribution in 1991. In 1978, 9.8% of census respondents identified as Tutsi (Rwanda 1982, 19); in 1991, only 8.4%. A decline of 15% in 13 years can neither be explained by the noticeably lower birth rates of Tutsi (see section VII), nor by out-migration, since there was no observable out-migration between 1978 and 1991. What was different, though, was the political climate in which the respective censuses took place. RGPH1 of 1978 was conducted during a time of relative political stability, whereas RGPH2 was held after the RPF invasion. Tutsi inside Rwanda were fearful, and some may have believed that it was safer to identify as Hutu. It is noteworthy that the authors of the 1991 census report themselves acknowledge this possibility (Rwanda 1994, 122, 126). I submit, therefore, that the ethnic data in RGPH1 are more reliable than in RGPH2, but that given the lower birth rate of Tutsi women during the intercensal period, the Tutsi share of the population dropped from 9.8% in 1978 to 9.0–9.5% in 1994, as opposed to the 12% suggested by Prunier, or the 8.4% according to RGPH2.
Also speculative is Prunier’s claim that 130,000 Tutsi survived the genocide. That number was the sum of two estimates: 105,000 counted in camps inside Rwanda by an anonymous international civil servant, and 25,000 ‘who did not go to camps’ (Prunier 1996, 264). The basis for the latter claim is unclear. How did he or his source identify 25,000 survivors among the hundreds of thousands of Tutsi returnees and millions of internally displaced Hutu? The combination of an incorrect baseline (930,000) and a speculative number of survivors (130,000) renders Prunier’s death toll estimate of 800,000 questionable. Written in less than year, The Rwanda crisis: history of a genocide certainly was a tour de force, but it was also bound to contain factual errors. Others have noted that the numbers in the book do not always add up (e.g. Mamdani 2002, 318–319). Nevertheless, Prunier’s 800,000 resonated more than Des Forges' prudent 507,000. In A problem from hell, Samantha Power writes that ‘Hutu nationalists exterminated some 800,000 Tutsi’ (2002, 91).
This section has reviewed the early estimates of three scholars in order to lay the foundation for my own examination of the government’s numbers in the next section and for my estimation of Tutsi genocide victims in section VII.
VI. Government counts: 937,000 to 1,952,087 Tutsi genocide victims
The Rwandan government has organised and/or funded at least four victim counts and two survivor counts.3 The paragraphs below provide introductory political context.
Hyping numbers was part of the RPF’s communications strategy from the beginning of the civil war. Émigré circles in the early 1990s claimed that during and after the ‘Hutu Revolution’ (1959–1962) ‘several hundred thousands of Tutsi were savagely slaughtered’ (Prunier 1996, 53), and that the Tutsi diaspora had reached two million (Guichaoua 1992, 18). I come back to these fantastic claims in section VII.
Given the catastrophic humanitarian situation in post-war Rwanda, the new government could ill afford to refuse the deployment of UN peacekeepers and aid workers. International agencies with expertise in demographic data collection reportedly offered to conduct a population survey (Prunier 1996, 362; Reyntjens 2017, 78). Accurate demographic data would indeed seem useful, if not indispensable, to any relief and reconstruction effort. Somewhat surprisingly, the government replied that it was ‘not interested’ (Reyntjens 2017, 78).
In March 1996, the ICTR chief prosecutor requested funding to hire forensic investigators and a statistical/demographic advisor. ‘Forensic analysis is critical to the investigation of the Tribunal. It is proposed to establish a Forensic Unit to undertake scientific analysis relating to the mass murders’ (United Nations 1996, para. 40, emphasis added). However, a month later the Tribunal’s registrar announced that ‘Acting on the advice of the Government of Rwanda and with due respect to the wishes of the families of the deceased, no further mass graves will be exhumed by the Office of the Prosecutor’ (Adede 1996, emphasis added).
What happened? An ICTR excavation in January–February 1996 of a mass grave in Kibuye, the first of its kind, had met with a street protest in the capital and disapproval from the government. According to a former tribunal official, excavations were ‘not something that the Rwandan government was happy with’ (quoted in O’Brien 2011, 168; see also Korman 2015, 203–220). Investigators reportedly had discovered ‘two layers’ of bodies, one of Tutsi genocide victims and one of Hutu civilians killed by the RPF (Guichaoua 2020, 132). As noted earlier, an investigative commission appointed by the UN Security Council had concluded that both sides to the armed conflict had perpetrated crimes against humanity.
The government then began progressively to exhume suspected mass graves and reinter any human remains found therein at designated genocide cemeteries and memorials (Korman 2015, 205; Jessee 2012; Rwanda 2008b). Individual identification and burial were not allowed because
For the state, the principal concern is the collective identification of victims. Victims are thus identified purely as victims of the genocide and for the state this anonymity is a reflection of the identity of the crime itself. (Korman 2015, 207)
I now turn to the core of the article: the Rwandan government’s various victim and survivor counts. Scholars have mostly ignored them, which in itself seems quite telling. Yet, they are worth examining if only to demonstrate how truth is fabricated in the new Rwanda.
Census no. 1 (1996): 1.2 million victims
Few would have known about the first victim census report had its principal author not been called to testify in 2003 before the ICTR. Eric Rousseau was in his twenties when the genocide happened. The Belgian had befriended Rwandan exiles in Brussels and when they returned to Rwanda in 1994, he joined them, and his involvement with the new government began.4
Rousseau began working for cabinet minister Jacques Bihozagara, a cofounder of the RPF and former RPF representative in Belgium. Rousseau, who had studied ‘marketing, advertising and public relations’, proposed the creation of a Commission for the Memorial of the Genocide and Massacres. The plan was accepted and an inter-ministerial commission created. Except for Rousseau, all six members were Tutsi returnees.
With financial support from the German government, the commission set out to identify massacre sites, count the victims and identify the perpetrators. Even with plenty of resources and expertise, just counting the victims would be a gigantic and time-consuming undertaking. Rousseau and his amateur team completed their field investigation in 56 days. A month later the ‘Rousseau report’ was ready.
Its long official title is ‘Preliminary report on the identification of sites of the genocide and massacres that took place in Rwanda from April to July 1994’ (Rwanda 1996). It seems odd to distinguish between sites of ‘genocide’ and ‘massacres’ because what else is a genocide site than a massacre site? In Rwandan discourse, however, Tutsi were victims of genocide while Hutu who died in intra-Hutu violence were victims of massacres. Adding ‘massacres’ to the title, therefore, was a way to acknowledge Hutu deaths. The April–July 1994 time frame, on the other hand, excluded the tens of thousands of Hutu civilians allegedly killed by the RPF after it seized power.
The report presents precise victim estimates for all 145 communes but leaves the reader guessing about the total death toll. In court, Rousseau clarified:
We were able to get to an estimate based on the graves that we opened, the skulls that we counted, collation of testimonies. We arrived at an overall estimate [of] about 1.2 million. (ICTR 2003a, 48)
A study by Philip Verwimp (2004) of the killings in the Bisesero hills illustrates the problems with the report. Using a database produced from a house-to-house survey of victims by an organisation of genocide survivors, Verwimp estimates the number of Tutsi killed in five communes in Bisesero (Gishyita, Gisovu, Gitesi, Mabanza, Rwamatamu) at around 13,000. Rousseau and his team counted 67,500 for these same communes.
Census no. 2 (1998): 282,804 survivors
In the years after the genocide, Tutsi survivors (rescapés) who felt neglected and marginalised by the new government organised to defend their interests (Rombouts 2004, 164–176). Responding to their demands, the government in 1998 created a survivors’ fund known by its French acronym FARG (Fonds national pour l’assistance aux victimes les plus nécessiteuses du génocide et des massacres perpétrés au Rwanda). The name of the fund seemed inclusive (‘victims of genocide and massacres’), and so was the definition of survivor:
‘Survivor’ means a person who survived genocide and massacres committed in Rwanda between October 1, 1990, and December 31, 1994, consisting in acts meant to:
ii) exterminate persons and destroy their property because of their opinion or of those of close relatives who were against genocide and massacres. (Rwanda 1998, art. 14)
The definition extends the category of genocide survivors (and implicitly also victims) to persons persecuted because they opposed the genocide, the so-called moderate Hutus. Tutsis who were in Rwanda during the genocide and survived automatically are rescapés (presumption of persecution), whereas Hutu had to prove individual persecution because of their ‘opinion’. In practical terms, moderate Hutu were members of certain opposition parties (nationally and locally); officials who refused to give the order to kill; and ordinary citizens who refused to kill or who helped Tutsi hide or escape. The challenge for Hutu was to prove their ‘justness’. As Rombouts (2004) writes, ‘the best proof for a moderate Hutu that he was persecuted during the genocide is being dead’ (Ibid., 212). The very fact of having survived reversed the presumption: a living Hutu was not ‘moderate’. Only 40 Hutu have officially been recognised as ‘just’ or ‘righteous’ (Rothbart and Cooley 2016, 77). Any pretence of inclusivity was abandoned when the FARG law redefined ‘survivors’ to mean ‘survivors of the genocide against the Tutsi committed between 01 October 1990 and 31 December 1994’ (Rwanda 2008a, art. 3.1)
FARG conducted a survivor census in 1998. The report was published in Kinyarwanda only and copies appear to be rare.5 The census counted 282,804 ‘neediest’ survivors (Rombouts 2004, 381). Conditions in Rwanda after the war and genocide were such that mortality must have been extremely high. If 282,804 Tutsi were alive in 1998, then the original surviving population was close to 300,000.
According to Rombouts, the census was a ‘raw count’ without taking names. ‘When assistance is offered, many will try to benefit from it especially in a poor country. [ … ] Individual rescapés eligible for support still had to be identified after the census’ (2004, 382). Therefore, it would be wrong to consider 282,804 an official figure. Undoubtedly there were cheaters and double counts, but survivor organisations (of which the umbrella organisation IBUKA is the best known) probably would have protested if the number was grossly inflated, as was the case in 2007 (see below). Assuming for the sake of my argument that the FARG census was only 25% inflated, the number declines from 300,000 to 225,000.
But 225,000 still is 75,000 more than Des Forges’ estimate. What can explain the difference? One possibility is that the FARG count included Tutsi who owe their survival to their registration as Hutu. ‘Survivor’ in the FARG law is defined so vaguely that such persons could be included or excluded. Another possibility is that the scholarly estimate was wrong. It is useful to recall that 150,000 was the sum of three estimates: 105,000 counted in refugee camps by an anonymous international civil servant, to which Prunier added 25,000 ‘who did not go to camps’, and Des Forges another 20,000 who fled to Congo and Tanzania. Thus, the estimate hinges on the reliability of the three numbers. Prunier’s claim that 25,000 ‘did not go to camps’ seems the most speculative (see section V). Des Forges’ estimate that 20,000 Tutsi fled to Congo and Tanzania is also unsourced. However, counting refugees in camps across the border seems far less speculative than identifying and counting survivors among millions of internally displaced. Rather than tying myself down to a specific figure, I consider 150,000 survivors a minimum and 225,000 a maximum.
Census no. 3 (2000): 1,074,017 victims
The Rousseau report being preliminary (note its official title), another victim count was held in July 2000. The new census was in fact the continuation of a non-governmental initiative launched in 1996. With financial support from the Dutch government and technical assistance from Belgian academic Philip Verwimp, the survivor organisation IBUKA conducted a victim count in Kibuye prefecture. Three years later, IBUKA published the Dictionnaire nominatif des victimes du génocide en Préfecture de Kibuye, a volume comparable to The Bosnian book of the dead.6
The Dictionnaire contains the names of some 59,000 Tutsi residents from Kibuye killed in 1994. IBUKA planned to replicate the project for the whole country (1999, 6–7) but the sometimes-critical NGO was reined in (Rombouts 2004, 366–370; Reyntjens 2013, 220) and its founding leaders fled abroad.
Using the questionnaire created by IBUKA the government proceeded with its own new census. The results were published in a report titled Dénombrement des victimes du génocide (Rwanda 2004). The title illustrates the shifting discourse about the events of 1994. By deleting the word ‘massacres’, Hutu victims are no longer acknowledged. The redefinition of ‘survivors’ in the FARG law (see above) is part of the same attempt to write them out of history. All that happened in 1994 was a genocide against the Tutsi.7
The report states that the government has compiled a list of the names of all victims but that ‘the law on the secrecy of statistics’ prohibits publication (Rwanda 2004, 19). However, laws on the secrecy of statistics do not protect death records; they protect the privacy of citizens whose personal data have been collected in a survey. Moreover, the names of thousands of victims are engraved in the Wall of Names at the Kigali Genocide Memorial, and a book commissioned by the government on the genocide in Murambi contains thousands more names (African Rights 2007). The government’s concern for the privacy of the dead also cannot be reconciled with its policy of burying them in mass graves or publicly displaying their mummified bodies (Vidal 2004; Burnett 2009; Jessee 2012; Dumas and Korman 2011; Korman 2015). A government confident of its work probably would want to publish the names of all victims. Finding donors to fund a Rwandan book of the dead should not have been too difficult.
As for numbers, the report speaks of 934,218 ‘confirmed’ victims and 1,074,017 ‘declared’ victims, and comments: ‘The majority of victims was killed because they were identified as Tutsi (93.7%)’ (Rwanda 2004, 38). That most others were Hutu the report does not say. As far as the National Commission for the Fight against Genocide (CNLG) is concerned, all victims were Tutsi. As of 1 June 2020, its website states that ‘about 1,070,014 [sic] Tutsi [were] killed in only 100 days’ (CNLG n.d., section 5.3).
| Prefecturea | Estimated Tutsi in April 1994b | Estimated survival ratesc (%) | Estimated survivors | Estimated victims | Government victim censusd |
|---|---|---|---|---|---|
| Butare | 147,000 | 25 (MV) | 37,000 | 110,000 | 221,000 |
| Byumba | 21,000 | 25 | 5000 | 15,000 | 21,000 |
| Cyangugu | 68,000 | 25 | 17,000 | 51,000 | 60,000 |
| Gikongoro | 68,000 | 25 (MV) | 17,000 | 51,000 | 107,000 |
| Gisenyi | 30,000 | 25 | 8000 | 23,000 | 38,000 |
| Gitarama | 92,000 | 50 (MV) | 46,000 | 46,000 | 129,000 |
| Kibungo | 60,000 | 25 | 15,000 | 45,000 | 102,000 |
| Kibuye | 79,000 | 25 (IBUKA) | 20,000 | 59,000e | 84,000 |
| Kigali Ngali | 94,000 | 25 | 24,000 | 71,000 | 165,000 |
| Ville de Kigali | 45,000 | 25 | 11,000 | 33,000 | 130,000 |
| Ruhengeri | 12,000 | 25 | 3000 | 9000 | 16,000 |
| RWANDA | 716,000 | 28 | 203,000 | 513,000 | 1,073,000 f |
aThis column lists the 11 prefectures that existed in 1994. In 1996 a new prefecture, Umutura, was carved out of the territory of Kibungo and Byumba. The official victim count for Umutura is 26,690. I have evenly distributed that number to Kibungo (7473 + 13,345) and Byumba (88,612 + 13,345). All numbers are rounded to the nearest thousand.
bSource: Rwanda 1994, 124. Increased by 1% (average 9.4%) + 2.5% annual growth.
cSurvival rates for Butare, Gikongoro, and Gitarama are based on field research by Marijke Verpoorten (MV) (Verpoorten 2012a, 554); for Kibuye, on the IBUKA census (IBUKA 1999). Survival rates for the other prefectures are not available, but I assume a rate not lower than in Kibuye, where the genocide reportedly was the most intense.
dSource: Rwanda 2004, 21.
eActual count by IBUKA (1999).
f The difference from the official number is the result of rounding.
As Scott Straus notes, ‘There are serious problems with the [report]’ (2008, 53). The total death toll far exceeds the size of the Tutsi population inside Rwanda in 1994. The government counted 25,000 more victims than IBUKA did in Kibuye; in Gitarama, the prefecture with the highest survival rate (50%), it counted nearly three times more victims than estimated. The results for the city of Kigali are also remarkable. The government counted 130,000 genocide victims, whereas the pre-genocide Tutsi population was only 45,000. These implausible numbers do not render this report any more credible than that of Rousseau. The original census cards are kept at the CNLG, but its archives are off limits to researchers unless they partner with an official. See, for example, the publications of US academics Hollie Nyseth Brehm and Christopher Uggen co-authored with CNLG Executive Secretary Jean-Damascène Gasanabo. My own numerous requests for access to the archives received vague answers, or none at all.
The report ends by saying that the numbers presented are minimal. ‘It is therefore recommended that this count be complemented with other more in-depth studies’ (Rwanda 2004, 38).
Column 6 in Table 1 shows the official death toll per prefecture. I added columns 2 to 5 to allow verification. In column 2 we find the estimated number of Tutsi per prefecture in April 1994; column 3 shows the estimated survival rates of Tutsi in 1994; columns 4 and 5 contain the numbers of survivors and victims respectively that should have been found in each prefecture, based on the data in columns 2 and 3.
Three more censuses would follow, but none can be considered ‘in-depth’ by any standards.
Census no. 4 (2004): 937,000 victims
All we know about this census comes from a short media report published on the eve of the 10th commemoration of the genocide in 2004:
A census carried out by Rwanda’s Ministry of Youth, Culture and Sports found that 937,000 Tutsi and politically moderate Hutus died during the 1994 genocide, an official announced on Thursday. ‘These are the people who died during the 100 days [April–June 1994] of mayhem and who we were able to find out their names, age and their places of birth’, [sic] Robert Bayigamba [ … ] said at a news conference.
He said the death toll could increase when the Gacaca justice system becomes fully operational as many perpetrators of the genocide were expected to testify about the people they killed. The Gacaca trials, based on traditional communal justice, are expected to begin later this year. [ … ] We shall come up with the exact figure after the Gacaca courts complete their work’, he said. (IRIN 2004, emphasis added)
The number 937,000 is close to the totals of the previous census but the lack of a report with names, or any report at all, makes verification impossible.
Gacaca trials of genocide suspects ended in 2012 but the promised ‘exact figure’ has yet to be announced. Access to the Gacaca archive, which has been digitalised with donor funds, reportedly is reserved to researchers vetted by the authorities (Anonymous 2015). My own requests for access were stonewalled.
Census no. 5 (2007): 309,368 survivors
With financial support from the European Union and the United Nations Development Programme, FARG organised a new survivor census in April 2007. Logically, the number of ‘surviving survivors’ should have been lower but the new census counted 309,368 rescapés, or 26,564 more than in 1998 (Rwanda 2008b). If 309,368 rescapés were alive in 2007, then their original number was +350,000.
The new FARG census produced a glossy report (Rwanda 2008c) but the survivor organisation IBUKA denounced it as a ‘failure’ and 50% inflated (Hirondelle News 2007). The fraud seems to have been the work of citizens scheming with local and/or FARG officials to obtain rescapé status and the attending social benefits (New Times 2011). Another possibility is that FARG grossly miscalculated. A table in the report shows survivors per five-year age group. The total is between 150,000 and 157,000 survivors (Rwanda 2008c, 9). The FARG count can also not be reconciled with the ±202,000 survivors recorded in 2005 by the Gacaca. Verpoorten (2012b, 5) considers the Gacaca count more reliable ‘since there are no clear no clear motives for over- or under-reporting’.
Census no. 6 (2008): 1,952,078 victims
In October 2008, while Gacaca trials were in full swing, the Kigali-based New Times reported the results of a new victim census funded by the government and organised by the Association of Student Genocide Survivors (Association des Etudiants Rescapés du Genocide, or AERG):
According to the [AERG] report, 390 Genocide Memorial Sites have been set up across the country since 1994. Those sites accommodate remains of 1,002,755 victims of the mayhem while 851,756 others were buried in different cemeteries across the country. To these numbers should be added the remains of 97,567 people who have been identified but have not yet been given decent burial, bringing the total to 1,952,078 genocide victims. (Musoni 2008)
The AERG report apparently was never published but one must wonder why the results were made public and why the government funded still another victim census in the first place. The CNLG website sticks to the results of the 2000 census: ‘about 1,070,014 [sic] Tutsi killed in only 100 days’. Rwandans who dare to challenge the official number expose themselves to prosecution under the laws against ‘divisionism’ and ‘genocide ideology’ (Human Rights Watch 2008, 34–41).
This section has provided insight into the Rwandan government’s peculiar way of counting the dead. Donors who funded the counts should ask themselves what these have accomplished. In the next section I propose an alternative method for calculating the death toll of the genocide, one that sidesteps the question of alleged undercounting of Tutsi in the 1991 census.
VII. My estimate: 489,000–564,000 Tutsi genocide victims
Back in the 1990s, official Rwandan demographic reports and health and fertility surveys were only available at selected depositories. Nowadays a dozen volumes can be readily downloaded from the Internet (e.g. websites of IREDA n.d.; ODSEF n.d.). I should like to emphasise that the goal of my demonstration is not to come up with an exact figure but to show that ‘more than a million’ is impossible. Here is what the various reports tell us.
First, a 1956 report from the Belgian colonial administration shows the following distribution for a population of 2,374,000: Hutu 82.74%, Tutsi 16.59%, and Twa 0.67% (Belgium 1957, 299–303). Thus, in 1956 the number of Rwandan Tutsi stood at around 394,000. Rwanda scholars agree that these data are reliable because the authorities had no reason to undercount Tutsi, and the latter had no reason to pass off as Hutu. Second, the population more than tripled between 1956 and 1991, from 2,374,000 to 7,290,000 (including 50,000 foreigners). The increase corresponds to a net average annual growth rate of 3.2%. And third, surveys from 1970 and 1983 reveal that fertility rates of Hutu women were consistently higher than those of Tutsi women (see below).
Let us now briefly consider the turbulent pre- and post-independence years. The ‘Hutu Revolution’ of 1959–1962 and accompanying violence triggered an exodus of Tutsi elites to neighbouring countries. Among the first wave of refugees were an infant Paul Kagame and his parents. When in 1963 a band of Tutsi exiles invaded Rwanda to restore Tutsi rule, the government responded with brutal repression. Between 10,000 and 14,000 Tutsi civilians were massacred, causing a second exodus. A political crisis in 1972–73 triggered a third and final wave of Tutsi refugees (Lemarchand 1970, 225; Prunier 1996, 41–74; Mamdani 2002, 103–31; Waugh 2013, 7–26; Reyntjens 2017, 16–23).
Two independently commissioned studies calculated the size of the Tutsi diaspora in the early 1990s. Catherine Watson for the US Committee for Refugees arrived at ‘probably half a million’ in February 1991 (Watson 1991); or ±550,000 in April 1994 at a 3% growth rate. André Guichaoua for the UN Refugee Agency estimated the diaspora at ±550,000 in March 1991 (Guichaoua 1992); or ±600,000 in April 1994 at a 3% growth rate. Prunier, on the other hand, counted 550,000 exiles in 1993 but 600,000–700,000 a year later (Prunier 1993, 122; 1996, 63). Whoever is right, these estimates undermine the claim of émigré circles that the Tutsi diaspora had reached two million (see above).
As noted earlier, the average annual growth rate of the Rwandan population between 1956 and 1991 was 3.2% despite massacres and exodus. Health surveys suggest that the Hutu and Tutsi populations had different fertility rates. In 1970 Tutsi women had a ±17% lower birth rate than Hutu women; from the late 1970s until the mid 1980s the difference was ±31% (Rwanda 1970, 87–113; 1983, 140–143). After 1987 the difference became negligible due to a ‘brutal decline’ of the overall fertility rate (see section V). Researchers concluded that ‘being Tutsi has no effect on fertility in the 1987–1992 period, while it has a negative and significant effect on fertility in both the 1982–1992 and 1977–1992 periods’ (Kraehnert et al. 2019, 18).
It would be logical to infer that for a substantial period of time between 1956 and 1991 Rwanda’s Tutsi population grew at a lower rate than the Hutu. However, three caveats are called for. First, the lower birth rates of Tutsi were partially offset by higher survival rates (Rwanda 1970, 97). Second, the quality of the surveys is debatable. The 1983 survey was the first conducted by the then newly established National Population Office. Third, the surveys do not take into account intermarriage between Hutu and Tutsi. We do not know how prevalent it was and whether women in mixed marriages had the same fertility as their ethnic kin.8 The only thing we can be reasonably certain about is that the growth rate of the Tutsi population in Rwanda did not exceed that of Hutu.
Comparable data about the Tutsi diaspora do not seem available. Given that growth of the Rwandan population in the 1970s and early 1980s was among the highest in Africa, we can also be reasonably sure that the growth rate of the diaspora did not exceed that of Hutu in Rwanda.
For the sake of my argument, I (liberally) assume that Tutsi in Rwanda and in the diaspora had the same growth rate as Hutu (3.2%) and that the total Tutsi population had increased from 394,000 in 1956 to around 1,264,000 in April 1994.9 If we subtract 550,000 diaspora Tutsi (Watson’s figure) we are left with 714,000 inside Rwanda; 600,000 diaspora Tutsi (Guichaoua’s figure) leaves 664,000 in Rwanda. But if the diaspora counted 700,000 Tutsi (Prunier’s figure), then the number of Tutsi in Rwanda before the genocide decreases to 564,000. Table 2 below shows these three possible distributions of the Rwandan Tutsi population in 1994.
| Hypothesis no. | Totala | Diaspora | Rwanda |
|---|---|---|---|
| 1 | 1,264,000 | 550,000 (Watson) | 714,000 (9.2%)b |
| 2 | 1,264,000 | 600,000 (Guichaoua) | 664,000 (8.6%) |
| 3 | 1,264,000 | 700,000 (Prunier) | 564,000 (7.2%) |
a1,264,000 is a maximum demographic estimate based on a 3.2% average annual growth rate since 1956.
bPercent of the estimated Rwandan population of 7,750,000 in April 1994 (excluding 50,000 foreigners).
| Total | Diaspora | Inside Rwanda | Survivors | Victims |
|---|---|---|---|---|
| 1,264,000a | 550,000 | 714,000 | 150,000–225,000 | 489,000–564,000 |
a1,264,000 is a maximum demographic estimate based on a 3.2% average annual growth rate since 1956.
Hypothesis no. 1 is the most probable because 9.2% Tutsi is within the 9.0%–9.5% range of the Tutsi share of the Rwandan population in 1994 (see section V). Assuming that 714,000 Tutsi lived in Rwanda in April 1994 and that 150,000–225,000 survived, the death toll from the genocide is between 489,000 and 564,000, as shown in Table 3.
As stated earlier, the goal of my exercise is not to establish a definitive figure but to show that the official death toll is inflated. My mid-point estimate of 526,000 is a little higher than the 506,000 mid-point estimate proposed by Omar McDoom (2020) in a forum on Rwanda in the Journal of Genocide Research. Other contributors to the forum suggest a death toll of 560,000–660,000 (Verpoorten 2020) and ‘roughly 600,000’ (Tissot 2020). Importantly, using different methodologies we all produce outcomes in the 500,000–600,000 range.
If 500,000–600,000 Tutsi died in the genocide and 1.7 million human remains are buried at genocide memorial sites (Barugahare 2019), then who are the other victims? What is hidden behind this contradiction? Scott Straus, an expert on the genocide, recently wrote that it is time for scholars to ‘discuss, recognise, study, and account for the other, non-genocide crimes against Rwandans in the 1990s’ (Straus 2019, 3–4).
VIII. ‘More than a million’ as a political statement
I submit that notwithstanding the various government censuses, ‘more than a million’ is not an actual count but a political statement. In its positive aspect, it could be argued that ‘more than a million’ is used to acknowledge that something unimaginable happened. To honour all victims, it is safer to err on the high side than to underestimate. In the Rwandan context, however, where Hutu deaths may not even be mentioned, ‘more than a million’ should be taken for what it is: propaganda.
There are interesting historical parallels between the RPF and the Algerian Front de Libération Nationale. The FNL fought a long and bloody war of independence from France (1954–1962), and in its aftermath murdered tens of thousands of native Muslim Algerians who had served in the French army. The RPF framed its struggle as a ‘war of liberation’ against what it considered a neo-colonial regime that perpetuated the divisions introduced by the coloniser; when the battle was over it reportedly killed tens of thousands of Hutu civilians. In Algeria the spoils of the war went to FNL cadres, in Rwanda to members of the RPF.
Post-war Algeria adopted a constitution with the following preamble: ‘The war of extermination waged by French imperialism was intensified and more than a million martyrs paid with their lives for the love of their country and liberty’ (Algerian Constitution 1963, para. 3, emphasis added). The preamble to Rwanda’s constitution is similarly worded: ‘In the wake of the genocide [ … ] that decimated more than a million sons and daughters of Rwanda’ (para. 1, emphasis added). As in Rwanda, it is unclear how the Algerian government arrived at ‘more than a million’. The best scholarly estimates of Algerian deaths range between 200,000 and 300,000 (Yacono 1982, 128; Ageron 2005, 661). In both countries the victors used the death of ‘more than a million’ countrymen and women to exculpate themselves and lay the foundation for radically new societies under their control.
The numbers game in Rwanda should be seen also against the background of its invasions of Zaire/DRC in 1996 and 1998. Waged under the pretext of preventing elements of the former Rwandan army and Hutu militia from completing the genocide of Rwandan Tutsi, the first invasion ended in overthrowing the Mobutu regime and installing a pro-Rwanda government. The second, purportedly launched to protect Congolese Tutsi, triggered a four-year long regional war which caused untold suffering and death. The international community largely condoned Rwanda’s actions. Who was to criticise a government that claims to represent ‘more than a million’ genocide victims and has made ‘fighting genocide’ a central tenet of its ideology?
IX. Conclusion
Casualty counts after a humanitarian catastrophe often are fraught with methodological and political pitfalls. In the case of the genocide against Rwandan Tutsi, the methodological pitfalls were minor because population registers were available and nearly all excess mortality of Tutsi in 1994 is directly related to the genocide. The political pitfalls, on the other hand, were enormous, given that the international community was seen as having been a ‘bystander’ to an unfolding genocide, and a bystander has no right to question or investigate. The new government dominated by the victorious RPF quickly made clear that it neither appreciated nor needed foreign help with counting the dead. Unlike in post-war Bosnia, there would be no forensic investigations and, for that matter, also no demographic/statistical studies by outsiders.
The government proceeded with its own counts. Other than their final tallies very little is known about them. The processes were extremely opaque; lists of names of victims are kept secret; and reports, if published, are filled with data of questionable relevance. Information that would allow meaningful verification was left out, and data files are kept secret. The definition of ‘victim’ and the time frame of the conflict were revised to include some dead and exclude others. The tallies themselves are highly problematic as well. Four victim counts, four remarkably precise outcomes, but none that can possibly be reconciled with demographic data. The government eventually settled on ‘more than a million’, a claim which few outside Rwanda have taken seriously.
The death of ‘more than a million’ Tutsi became the foundation of the new Rwanda, where former exiles hold a monopoly on power. It also created the socio-political environment for the mass criminalisation of Hutu. Gacaca courts eventually tried more than a million (Nyseth Brehm, Uggen, and Gasanabo 2016), which led President Kagame to suggest that all Hutu bear responsibility and should apologise (Benda 2017, 13). Thus the new Rwanda is built not only on the death of ‘more than a million’ Tutsi but also on the collective guilt of Hutu.10 This state of affairs is in no one’s interests except the regime’s. An honest accounting of the past must acknowledge every victim. Post-war Bosnia has shown that it can be done if the authorities do not obstruct.