Citation
Journal of Genocide Research
ISSN: 1462-3528 (Print) 1469-9494 (Online) Journal homepage: https://www.tandfonline.com/loi/cjgr20
Contested Counting: Toward a Rigorous Estimate
of the Death Toll in the Rwandan Genocide
Omar Shahabudin McDoom
To cite this article: Omar Shahabudin McDoom (2020): Contested Counting: Toward a Rigorous
Estimate of the Death Toll in the Rwandan Genocide, Journal of Genocide Research, DOI:
10.1080/14623528.2019.1703252
To link to this article: https://doi.org/10.1080/14623528.2019.1703252
Published online: 01 Jan 2020.
Submit your article to this journal
Article views: 34
View related articles
View Crossmark data
Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=cjgr20
JOURNAL OF GENOCIDE RESEARCH
https://doi.org/10.1080/14623528.2019.1703252
FORUM: CALCULATING MORTALITY IN THE RWANDAN GENOCIDE
Contested Counting: Toward a Rigorous Estimate of the Death
Toll in the Rwandan Genocide
Omar Shahabudin McDoom
Department of Government, London School of Economics, London, UK
As a world-historical event that has become indelibly etched into the conscience of
humanity, the Rwandan genocide unequivocally merits scholarly attention to establishing
its death toll. The historical record demands it. Preserving the memory of the victims,
recognizing the enormity of the crime, and attributing full responsibility to the perpetrators are also compelling reasons to account for the lives taken. Justice also demands it.
Yet in the twenty-five years since the genocide, dramatically divergent claims have
emerged in respect of two fundamental aspects of the death toll: the number and the
identity of the victims. How many Rwandans were killed? And how many of them were
Tutsi, and how many Hutu? If one examines the claims for the overall number killed, at
the higher end lies the figure of 1,074,017 Rwandan dead. This number originates with
the Rwandan government which conducted a nationwide census in July 2,000, six years
after the genocide.1 Toward the lower end lies an estimate from Human Rights Watch,
one of the first organizations on the ground to investigate the genocide, of 507,000
Tutsi killed.2
Differences in the claims concerning the victims’ ethnic identities are even more stark.
In the Rwandan government’s census, 93.67% or just over one million were Tutsi.3 In contrast, two contributors to this forum, Davenport and Stam, have in the past contended
somewhere between 250,000 and 890,000 Rwandans were killed and suggested –
among other possibilities – it is possible as few as 50,000 of them were Tutsi and as
many as 540,000 of them Hutu.4 If their claim that more Hutu were killed than Tutsi is
CONTACT Omar Shahabudin McDoom
o.s.mcdoom@lse.ac.uk
Department of Government, London School of
Economics, Houghton Street, London WC2A 2AE, UK
1
Government of Rwanda, Dénombrement Des Victimes Du Genocide: Rapport Final (Kigali: Ministère de L’Administration
Locale de L’Information et des Affaires Sociales, 2002). Higher estimates for the death toll exist. For instance, the Student
Genocide Survivors’ Association (AERG in French) estimated 1,952,078 victims. The Government of Rwanda has not
officially endorsed this figure. See https://www.newtimes.co.rw/section/read/5288 (accessed 15 June 2019).
2
Alison Des Forges, Leave None to Tell the Story: Genocide in Rwanda (Human Rights Watch; International Federation of
Human Rights, 1999), 17.
3
The government’s genocide census distinguished “declared” from “enumerated” victims. A victim is enumerated only if
the enumerator obtained responses for all questions on the questionnaire. The enumerated figure is slightly lower at
934,218 victims. The report is ambiguous, however, as to whether the 93.67% refers to the declared or enumerated total.
The estimate of over one million Tutsi dead is based on the declared total. If the enumerated total is used, the figure
becomes 875,000 Tutsi dead.
4
Christian Davenport cites a 2010 source jointly-authored with Allan Stam in a 2014 blog entry where these estimates are
listed. See http://politicalviolenceataglance.org/2014/10/24/measuring-denying-trivializing-deaths-in-the-case-ofrwanda/ (accessed 15 June 2019). In a 2012 Powerpoint presentation, Stam presented a different overall death toll
of between 200,000 and 1,250,000 (slide 25) and presented a different possible Tutsi victim count of 206,000 and, as
a plausible Bayesian-based estimate, a different possible Hutu victim count of 408,000 (slide 18). The presentation is
available on a project website listing Stam and Davenport as the principal investigators. See https://genodynamics.
weebly.com/presentations.html (accessed 15 June 2019).
© 2019 Informa UK Limited, trading as Taylor & Francis Group
2
O. S. MCDOOM
true, this would invite us to revise the common characterization of the violence as a genocide solely against Tutsi.5 It implies there may have been a double genocide or additionally a politicide, where individuals were targeted for their political beliefs rather than their
ethnic identity.
In this article, I present an estimate of between 491,000 and 522,000 Tutsi killed in the
period 6 April to 19 July 1994 within Rwanda. I take care to delimit this estimate both temporally and spatially and to specify it refers only to Tutsi. There have been other periods of
major violence in which Rwandans – both Tutsi and Hutu – lost their lives throughout the
1990s and the 2000s. Moreover, some of this violence occurred outside of Rwanda, in the
wider Great Lakes region and most notably in the Democratic Republic of Congo. The genocide and its death toll then could be understood in this broader historical and regional
context as these other episodes of violence were not unrelated.6 Adopting a wider lens,
however, should not alter our understanding of the fact that a genocide occurred
between April and July 1994 inside Rwanda.7 Nor should it lessen the moral responsibility
or negate the agency of those who organized and participated in it. The genocide remains
a shocking and tragic world-historical event in which the intensity and speed of the violence, and the scale of civilian participation, distinguished it. Nonetheless, situating the
genocide and its death toll in this broader context has merit insofar as it serves to underline that the life of every Rwandan taken through such violence deserves to be remembered and accounted for. Memory and justice cannot be selective.8
What Is It at Stake in Establishing the Genocide’s Death Toll
Both the politics and the emotions surrounding the genocide have provoked and shaped
the debate over its death toll. The number and the identities of those killed have become
intensely politicized questions in large part because they evidently drive the characterization of the violence. Did the violence targeting Tutsi constitute genocide, the crime of
crimes, or something less morally egregious? Should the killing of Hutu be seen as
morally less reprehensible or equivalent? The number and identities of the victims then
have become weapons in the battles waged between those who would deny, minimize,
or equate the genocide with other violence, and those who would insist on its moral
uniqueness.
These issues are politicized, however, in part because they are also central in another
morally charged debate. They shape our perceptions of who were the victims and the
aggressors in the violence. When the asymmetry in the loss of life is stark, the side with
In the BBC documentary Rwanda: The Untold Story (2014), Stam states (at minute 30.33): “If a million people died in
Rwanda in 1994, and that’s certainly possible, there is no way that the majority of them could be Tutsi.”
For an overview of the various episodes of violence in the Great Lakes in the postcolonial era, and the linkages between
them, see Omar Shahabudin McDoom, “War and Genocide in Africa’s Great Lakes since Independence,” in Oxford Handbook of Genocide Studies, ed. Donald Bloxham and A. Dirk Moses (Oxford: Oxford University Press, 2010).
7
Rwanda’s Organic law 08/96 on the Organization of Prosecutions for Offences constituting the Crime of Genocide or
Crimes against Humanity defines a genocide victim and sets temporal boundaries from 1 October 1990 to 31 December
1994. While there is scholarly disagreement as to when the genocide was first planned, it is generally recognized that a
state-sponsored policy to eliminate all Tutsi was not implemented until almost immediately following president Habyarimana’s assassination on 6 April 1994 and ended with the crossing of the interim government and government soldiers
into then Zaire and the establishment of a new government by the Rwandan Patriotic Front on 19 July 1994.
8
This point is powerfully made by Scott Straus, “The Limits of a Genocide Lens: Violence against Rwandans in the 1990s,”
Journal of Genocide Research 21, no. 4 (2019).
5
6
JOURNAL OF GENOCIDE RESEARCH
3
the higher civilian casualty count will be more likely viewed as having victim group status;
the side that sustained fewer civilian casualties more likely to be seen as the belligerent.
Recognition of victim status confers distinct political advantages. It elicits sympathy and
induces guilt in those who failed to prevent and stop the violence. Both sentiments
may be converted into rhetorical and material support for the victim group. It may also
grant the victim group some measure of immunity from criticism as critics risk being compared to the original perpetrators.9 In contrast, aggressor status is politically disadvantageous. It erodes legitimacy and casts doubt on claims of self-defense and any other
pretensions of a just cause to the violence. Aggressors must be held to account and punished for their misdeeds. More generally, contestation over victim status functions to
reduce the complexity of the violence into the basic binary division between good guys
and bad guys; and contestation over the victims’ identities latently functions to maintain
ethnicity as a politically salient force in society.
The contestation is not, however, driven solely by competing political logics and strategies. Discussion of genocide also evokes powerful emotions. It stirs feelings of anger, grievance, guilt, remorse, and even fear. Affect, as social psychologists have long established,
biases judgements and evaluations.10 Emotions, alongside politics, then represent another
force that distorts and obfuscates the true death toll. Disputes over the number and identity of the victims in genocides should be anticipated. Scholars would do well then to set
clear standards for evaluating the competing claims that they know will be made. I begin
then by proposing some criteria I believe should be considered when generating, and also
evaluating, estimates for the victims of genocide. These guidelines should certainly not be
seen as the definitive word on the subject. Rather, they should be considered part of an
evolving conversation. Better criteria will be developed in time.
Towards Some Guidelines for Establishing Genocide Death Tolls
First, as already suggested, the violence whose toll we are seeking to establish should itself
be carefully delimited. At a basic level this requires setting temporal and spatial boundaries. The more complex challenge are the conceptual boundaries. Multiple violent
events may occur within the same temporal and spatial limits but these events may comprise conceptually distinct forms of violence. In Rwanda, a genocide against Tutsi unambiguously occurred between 6 April and 19 July 1994. However, other types of violence
also took place: the assassinations of moderate Hutu and Tutsi political elites; retributive
killings of Hutu civilians by the RPF; personal score-settling among ordinary Rwandans;
and combatant deaths on both sides of a civil war. Conceptually subsuming all this violence within the label genocide not only reduces important complexity; it weakens the recognition of these other lives lost.
Second, there should be a degree of transparency in the data and methods sufficient to
be able to replicate and verify the estimate. The methodological assumptions and choices
should be set out and the data relied upon made available upon request. Third, a reasoned
judgement on the reliability of the data chosen and the suitability of the method selected
See Filip Reyntjens, “Rwanda, Ten Years on: From Genocide to Dictatorship,” African Affairs 103 (2004): 177–210, where
he argues the post-genocide Rwandan government skilfully exploited a “genocide credit” to silence criticism of itself.
10
See Dacher Keltner and Jennifer S Lerner, “Emotion,” in Handbook of Social Psychology, ed. Susan Fiske, Daniel Gilbert,
and Gardner Lindzey (New Jersey: John Wiley & Sons, 2010), 317–52.
9
4
O. S. MCDOOM
should be offered. As a basic standard, the author of the estimate should explain why they
opted for their particular data and, insofar as they are independent of the data, their
methods too. If the data were chosen because no alternatives were available to the
author, this should be made clear. As a higher standard, and a more complex task, the
author should offer a judgement on the potential biases in a data source. Competing
data sources should not be presented as equally reliable if there is reason to believe
bias may exist. Establishing estimates should not be simply a matter of calculating an
average across data sources. This rewards biased outliers.
Finally, the author should endeavour to situate their claim among the better-known
estimates already published to help the reader gauge how close it is to these others. Consensus-building is an important part of the process of establishing the historical record.
Methods and Data
Generally, two methods are commonly followed in establishing the genocide’s death toll:
counting the dead or counting the survivors. Notwithstanding the method chosen, the
reliability of all estimates depends on two crucial numbers: the number of Tutsi in
Rwanda on the eve of the genocide; and the number of Tutsi who survived the genocide.
How Many Tutsi Lived in Rwanda Before the Genocide?
According to the 1978 and 1991 population censuses, there were 467,587 and 597,459
Tutsi resident in Rwanda respectively at the time they were conducted. Tutsi then represented 9.8% and 8.4% of the population.11 If we extrapolate from these figures to
April 1994 using the historical average population growth rate specific to each of
Rwanda’s eleven administrative prefectures, this would mean there were between
714,000 and 647,000 Tutsi in Rwanda just before the genocide. However, it has been
claimed that both censuses significantly underestimated the Tutsi population.12 Two
reasons are usually given. First, the Habyarimana government (1973–1994) may have purposely understated the Tutsi population in order to justify ethnic quotas it had introduced
in the 1970s to reverse the longstanding over-representation of Tutsi in educational institutions and the civil service. Second, some Tutsi may have identified themselves as Hutu to
census enumerators to avoid state-sponsored discrimination during the Habyarimana era.
At first glance, the colonial census data appear to support the contention that the two
postcolonial censuses underestimated the Tutsi population. Tutsi represented 16.7%,
17.5%, and 16.6% in the 1933, 1952, and 1956 censuses respectively.13 However, a good
part of the difference between the colonial and postcolonial periods is explained by the
11
See Government of Rwanda, Récensement Général De La Population Et De L’habitat: Résultats Définitifs (Kigali: Bureau
National de Récensement, 1978); Récensement Général De La Population Et De L’habitat Au Août 1991: Résultats Définitifs
(Kigali: Ministère du Plan, 1994). The Tutsi proportion is calculated as a percentage of the population of Rwandan citizens
only. If non-Rwandans were included, the proportions would drop slightly to 9.7% and 8.3% for 1978 and 1991
respectively.
12
See for example Gérard Prunier, The Rwanda Crisis, 1959–1994 : History of a Genocide (London: Hurst and Company,
1995), 264.
13
For the 1933 data, see Mahmood Mamdani, When Victims Become Killers : Colonialism, Nativism and the Genocide in
Rwanda (Oxford: James Currey, 2001), 98. For the 1952 data, see Des Forges, Leave None to Tell the Story : Genocide
in Rwanda, 40. For the 1956 data, Filip Reyntjens, Pouvoir Et Droit Au Rwanda: Droit Public Et Évolution Politique,
1916–1973 (Tervuren: Musée Royal de l’Afrique Centrale, 1985), 28.
JOURNAL OF GENOCIDE RESEARCH
5
exodus of tens of thousands of Tutsi during and soon after the Hutu revolution (1959–62)
that ushered in Rwanda’s independence. A 1964 UNHCR census estimated 336,000 refugees, mainly Tutsi, had fled to Burundi, Uganda, Zaire, and Tanzania.14 Rwanda’s population stood at 2,634,000 in 1959 which would mean, at most, 461,000 Tutsi (17.5%)
were resident in Rwanda prior to the revolution.15 If the UNHCR census is accurate, projecting forward to 1994 would mean there would have been only some 350,000 Tutsi in
Rwanda on the eve of the genocide. As the national census, already believed to be an
underestimate, puts the Tutsi population at 597,000 in August 1991, this figure is implausibly low. Either the UNHCR data are unreliable, the 1959 population data are inaccurate, or
else tens of thousands Hutu also went into exile with Tutsi. Whatever the reason, it appears
dependence on the colonial-era data, which may have been reliable, to estimate the postcolonial Tutsi population depends on other data whose reliability is more questionable.
An alternative method for establishing the Tutsi baseline pre-genocide would be to
examine locally-collected administrative population data. These local data offer two
advantages over the national census data. First, they were collected more frequently,
often several times per year, allowing us to observe any unusual changes in the trend.
Second, individuals were less likely to be able to misrepresent their ethnicity. Before the
genocide, each of Rwanda’s 145 administrative communes maintained registers of
births, deaths, and migrations based on information provided either by the administrative
cell heads (responsables) or else by the heads of blocks of ten households (nyumbakumi).
These local administrative figures, who lived within the community, typically knew
members personally and it would have been very difficult for a Tutsi family to pretend
it was Hutu. The data in these commune registers were then used by each of Rwanda’s
eleven prefectures to produce annual administrative census reports.
How reliable are these local censuses? Local data collection capacity may well vary
between localities and over time. I analyzed the reported data in two localities and over
8 distinct time periods to probe their reliability. During fieldwork in 2003 I found in prefecture archives the original population reports for 1965, 1975, 1983, 1985, and 1988 for
Butare prefecture, in which Tutsi represented 18.3%, 18.5%, 18.3%, 18.3%, and 17.8%
respectively. I also found the reports from 1965, 1973, 1981, 1985, 1988, and 1991 for
Ruhengeri prefecture in which Tutsi represented 0.8%, 0.7%, 0.7%, 0.7%, 0.6%, and 0.6%
of the population respectively. The proportion of Tutsi reported in the local administrative
data then remained reasonably stable over twenty-five years. Importantly, the first reports,
from 1965, predate the introduction of Habyarimana’s ethnic quotas in the 1970s when
the ethnic balance would have become politicized. The trend in the local data then
does not suggest any sudden, large, or otherwise suspicious shift in the Tutsi proportion.
The observed gradual decline over time likely reflects the lower Tutsi female fertility rate.16
At least as an indicator of the Tutsi proportion, the local administrative data then appear
reasonably reliable.
14
Cited in André Guichaoua, Le Problème Des Réfugiés Rwandais Et Des Populations Banyarwanda Dans La Région Des
Grands Lacs Africains (Geneva: United Nations High Commission for Refugees, 1992), 20.
15
The 1959 population estimate comes from Office de l’Information et des Relations Publiques pour le Congo Belge et le
Ruanda-Urundi, (1959), 33, cited in Reyntjens, Pouvoir Et Droit Au Rwanda: Droit Public Et Évolution Politique, 1916–
1973, 21.
16
Using data from the Rwanda Demographic and Health Survey for 1991, I calculated Hutu women had on average 6.2
children in their lifetime, whereas Tutsi women had only 5.1.
6
O. S. MCDOOM
Did the 1991 National Census Underestimate the Tutsi Population?
I compared the national census data against the local data I had for Ruhengeri prefecture
for 31 December 1991, adjusted to 15–16 August 1991, the national census reference date.
To ensure data comparability, I counted only Rwandan citizens who were resident and
physically present in their household on the census reference date. Two main findings
emerge. First, the overall population matched reasonably well. The national census put
Ruhengeri’s population at 741,705; the date-adjusted local census data predicted a population of 734,697. The difference, 0.95%, is not inconsequential, though some of the difference may be attributable to the adjustment using the average historical population
growth rate (3.1% p.a. for Ruhengeri) or simply lower administrative capacity and
efficiency at the local level. Second, and more interestingly, the national census reported
the Hutu population higher and the Tutsi and Twa population significantly lower than in
the local data. In relative terms, the Hutu population was overstated by 1.07% and the Tutsi
and Twa understated by 14.48% and 9.20% respectively. Insofar as Ruhengeri prefecture
was not exceptional, it does appear then the national census did under-count the Tutsi.
Was this undercounting the result of deliberate central manipulation of the data or
individuals mispresenting their ethnicity to census enumerators? The data do not
permit a conclusive answer. However, it is worth noting that the number of ethnic
Twa were under-counted as well as the number of Tutsi. If the rationale for central
manipulation was to reverse Tutsi over-representation in educational establishments
and the civil service, this would suggest the cause was local self-identification as Hutu
rather than central government manipulation. The Twa were a historically marginalized
group and were likely under-represented in the public sector and education. Both Tutsi
and Twa had incentives to identify as Hutu because of the advantageous status it
afforded.
Establishing the Tutsi Pre-Genocide Baseline
Given the stability over time in the population share of Tutsi in the local administrative
data, they represent a potentially valuable alternative source to the national census for
estimating the pre-genocide Tutsi population. The only year for which I had administrative
population data for all eleven prefectures was 1983.17 In 1983, the national proportion
(excluding foreigners) of Tutsi was 10.8%. It is worth noting that the Tutsi proportions
for Butare and Ruhengeri prefectures were 18.34% and 0.66%, in line with the other
administrative censuses going back to 1965. The administrative data for 1983 then did
not appear unusual. If we assume the 1991 national census correctly stated Rwanda’s
overall population, and only misstated the ethnic balance between Hutu, Tutsi, and
Twa, then if we extrapolate the census data forward from 15th August 1991 to 6th April
1994 (966 days) using the national annual population growth rate of 3.1%, Rwanda’s
overall population on the eve of the genocide would have been approximately
7,415,000 persons.18 This figure refers to Rwandan citizens who were resident and
17
The 1983 administrative data were summarized and submitted in May 2005 as exhibit 2D49 in the trial of Bizimungu
et al., ICTR-99-50-T. The data were compiled for the Defense by expert witness, Rwandan anthropologist, Professor
Deo Mbonyinkebe.
18
The annual national population growth rate of 3.1% is calculated for the period 1978–1991. See Government of Rwanda,
Récensement Général De La Population Et De L’habitat Au Août 1991: Résultats Définitifs, 13.
JOURNAL OF GENOCIDE RESEARCH
7
present in their usual household on the census reference date.19 If the Tutsi population
were 10.8% of this figure, I estimate roughly 800,000 Tutsi were living within Rwanda
on the eve of the genocide.
How Many Tutsi Survivors Were There?
To my knowledge, there have been two widely-cited but significantly different survivor
estimates. The first estimate of 150,000 is the aggregation of three other estimates: (i)
an anonymous international civil servant’s estimate of Tutsi living in refugee camps in
July 1994 (105,000); (ii) an unsupported estimate of 25,000 Tutsi inside Rwanda who did
not go to the camps; and (iii) another unsupported estimate of 20,000 Tutsi who
escaped to Tanzania and Zaire.20 The second estimate of 309,000 is based on a 2006
national census, conducted by the National Institute of Statistics of Rwanda (NISR), a
Rwandan governmental body.21 It asked the local heads (responsables) of Rwanda’s smallest administrative unit, the imidugudu, to enumerate the survivor households in their
communities. To these two well-known estimates, I add a third estimate of 202,000
based on the Rwandan government’s gacaca data. Rwanda’s gacaca courts were an innovative community justice institution established by the government with the stated aims
of holding to account local perpetrators, establishing the truth of what transpired in those
communities, and assisting in reconciliation. Local communities compiled extensive information on the genocide as part of gacaca and the Rwandan government published a
summary of some of these data in 2007.22 They included the number of survivors in
local communities.
My assessment of the reliability of each estimate is varied. The first estimate of 150,000,
although the most longstanding, has the least well-documented empirical basis to it. Two
of the three estimates on which it depends – essentially the number of Tutsi outside of the
refugee camps – appear to be educated guesses with no evidence offered to support
them. I must regard it then as the least reliable. This leaves the two government estimates
between which the difference, 107,000, is still very large. I had field data that allowed me
to ground-truth the accuracy of the smaller, gacaca estimate of 202,000 survivors. Specifically, I had data on the number of Tutsi resident and the number of Tutsi killed in 20
administrative sectors in Ruhengeri and Butare prefectures that had been part of the
initial gacaca pilot exercise. Importantly, I had collected these data in 2003 directly from
the pilot communities themselves before the data had been passed to the nationallevel authorities charged with their compilation. These data were based on handwritten
lists drawn up by the community enumerating the entire sector population and also
19
I consider only Rwandans who were present on the census reference night of 15th August 1991 in order to ensure comparability with the 1983 administrative data. The summary of the administrative census data in my possession reported
only the population of Hutu, Tutsi, and Twa. Non-Rwandans were not reported. Administrative censuses also generally
reported only present individuals. Individuals temporarily absent (away for less than 6 months) were not reported. If both
Rwandan and non-Rwandans, and both individuals present and absent on the census reference night were counted, the
overall population of Rwanda on 6 April 1994 would have been approximately 7,761,000.
20
See Prunier, The Rwanda Crisis, 1959–1994, 264; Des Forges, Leave None to Tell the Story, 17.
21
See Government of Rwanda, Récensement Des Rescapés Du Genocide De 1994: Rapport Final (Kigali: Institut National de la
Statistique du Rwanda, 2008).
22
The Rwandan government’s webpage on which the gacaca summary report was released is no longer active. However,
Marijke Verpoorten, one of the contributors to this forum, downloaded the report before it went offline and entered the
data electronically. She generously shared the dataset with me.
8
O. S. MCDOOM
those killed by name. I viewed them as reliable. I could calculate with confidence how
many Tutsi survived in these particular 20 sectors. Comparing these numbers against
the 2007 national gacaca data, however, shows almost no correspondence. For 10
sectors, the national gacaca data understated the number of survivors by an average of
19.6%. For the other 10 sectors, the national data more troublingly overstated the
number by an average of 48.6%. I conclude then that the 2007 national gacaca estimate
of 202,000 survivors is also unreliable.
This leaves the reliability of the 2006 survivor census of 309,000 individuals to assess.
Government estimates will inevitably be vulnerable to the charge of manipulation for political purposes. In this instance, however, there may have been little incentive to inflate the
figure. The survivor census was not undertaken by the FARG, the government agency
charged with assisting survivors and whose budget would depend on their number. It
was conducted by the formally independent Rwandan statistics agency. A very high survivor number would also call into question the government’s claim of one million Tutsi
dead, a number that has greater political importance for it. In my view, the risk that the
survivor data had been exaggerated for political purposes was small. The main concern
with this figure, however, is its definition of a survivor. The census uses an expansive
definition: “Any person pursued for his ethnic affiliation, his ideology, or that of a family
member, against divisionism, exclusionism, or extermination, and who escaped massacres
and genocides committed between 1 October 1990 and 31 December 1994.”23 Survivors
then could include individuals killed before April 1994 and also persons who were not
Tutsi. One of the most important groups subsumed within this broad definition were
the Hutu widows and widowers of Tutsi partners. The census enumerated 31,776
widows and widowers, 10.3% of all survivors. It did not, however, distinguish their ethnicity. If we assume, albeit an unlikely possibility, all of them were Hutu, the lower bound for
the number of Tutsi survivors would become 277,592. My best overall estimate for the
number of Tutsi genocide survivors then is between 278,000 and 309,000 individuals.
How Many Tutsi Were Killed During the Genocide?
Subtracting the best estimate of the number of survivors from the best estimate of the pregenocide Tutsi population, I believe the number of Tutsi killed during the genocide lies
between 491,000 and 522,000 individuals. If accurate, then nearly two-thirds of
Rwanda’s Tutsi population were exterminated between April and July 1994. The finding
should leave little doubt as to whether the violence amounted to genocide.
How Many Hutu Were Killed During the Genocide?
Establishing the number of Hutu killed during the genocide matters for three key reasons:
it shapes the characterization of the violence; it raises questions of the experience of
memory and justice following the genocide; and, by extension, it has implications for
the possibilities of interethnic reconciliation. The number has been central then to
debates over whether the violence amounted to a double genocide or a politicide;
whether Rwanda has experienced victor’s justice and institutionalized an asymmetric
23
Government of Rwanda, Récensement Des Rescapés Du Genocide De 1994, 3. Author’s translation from French.
JOURNAL OF GENOCIDE RESEARCH
9
public commemoration of the victims; and whether the country has successfully escaped
the risk of renewed ethnic violence.
I do not have new data to improve upon the estimates already in the public domain for
the number of Hutu killed during the genocide period. Scholarly attention has focused in
particular on Hutu civilians killed by rebel forces, the Rwandan Patriotic Army (RPA), as they
advanced and captured territory. Two of the more widely-cited estimates, by the late
Alison Des Forges of Human Rights Watch, and Robert Gersony, a UN consultant who conducted a field investigation in 1994 but whose report was not officially released, put the
figure at between 25,000–30,000 and 25,000–45,000 respectively.24 These estimates
should be seen as a lower bound. Gersony visited only 41 of 145 communes inside
Rwanda. Davenport and Stam have also estimated as many as 80,000 persons may have
been killed in RPF-controlled areas.25 There is also evidence of Hutu killed for other
reasons in this time period: the assassinations of political moderates; combatant deaths
between two armed groups; interpersonal disputes, often land-related, between ordinary
Rwandans; and the targeting of individuals considered sympathetic to Tutsi. We have little
systematic evidence of these deaths. Verwimp conducted a survey that included 1,620
Hutu across three Rwandan prefectures. He found forty-three Hutu (2.7%) suffered
violent deaths of whom twenty-one were killed by the RPA, nine by Hutu militia, one by
Rwandan government soldiers, five by other, unspecified individuals, and eight by
authors unknown. In general, however, the deaths of all these non-Tutsi Rwandans in
this time period remain poorly documented and estimates will unlikely improve much
until investigations with a view to a judicial accounting for them are initiated. The
current and limited evidence we do have of Hutu civilian deaths in this time period
does not, in my view, support claims of a double genocide or a politicide. It also does
not challenge the evidence that a genocide against Tutsi took place. It does serve to
underscore, however, that forms of violence other than genocide occurred at the same
time in Rwanda for which there has been limited recognition. The exclusive focus on
the genocide unfortunately obscures this other violence.
I have purposely limited discussion of the Hutu death toll to the period 6 April to 19 July
1994, and to violence within Rwanda, as this is generally recognized as the time, and place,
in which the Rwandan genocide occurred. The estimate I offered for the Tutsi death toll
equally relates to this time period and to events inside Rwanda. However, it would be
remiss not to consider how the death toll would change if these particular temporal
and spatial boundaries were drawn differently. There have been major episodes of violence claiming the lives of non-Tutsi Rwandans both before and after the genocide, and
both inside and outside of Rwanda. These events did not occur independently of each
other, at least not in the minds of the parties involved in them. Straus has proposed a
useful periodization of this violence, along with possible death tolls.26 Six time periods
are distinguishable: (i) the civil war leading up to president Habyarimana’s assassination
(October 1990 to April 1994); (ii) the genocide itself (April to July 1994); (iii) the consolidation of power by the RPF (August 1994–1995); (iv) the pursuit of Hutu refugees in the D.R.C.
(1996–97); (v) the insurgency and counterinsurgency in Rwanda’s north-west (1997–98);
24
For both estimates, see Des Forges, Leave None to Tell the Story, 17 & 558.
See Christian Davenport and Allan Stam, “Rwandan Political Violence in Space and Time” (Unpublished manuscript).
Available at https://genodynamics.weebly.com/writings.html (accessed 17 June 2019).
26
See Straus, “The Limits of a Genocide Lens: Violence against Rwandans in the 1990s,” 7–12.
25
10
O. S. MCDOOM
and (vi) authoritarian repression inside Rwanda (2000 onwards). Aggregating the various
estimates offered for these episodes, excluding the genocide, places the death toll of nonTutsi Rwandans in the several hundreds of thousands. The most significant of these is the
systematic targeting of Hutu refugees in the D.R.C., for which the strongest claim of a
second genocide has been made.27 In comparison with the genocide against the Tutsi,
all these episodes have received far less recognition; the estimates of their death tolls
are much more imprecise; and, most troublingly, accountability for them has been woefully limited.
Conclusion
I have estimated between 491,000 and 522,000 Tutsi, nearly two thirds of Rwanda’s pregenocide Tutsi population, were killed between 6 April and 19 July 1994. I calculated
this death toll by subtracting my estimate of between 278,000 and 309,000 Tutsi survivors
from my estimate of a baseline Tutsi population of almost exactly 800,000, or 10.8% of the
overall population, on the eve of the genocide. My estimate is in line with Human Rights
Watch’s (HRW) early estimate of 507,000, though HRW relied on the 1991 national census
to estimate a pre-genocide Tutsi population of 657,000 and assumed there were 150,000
survivors. It is somewhat lower than Verpoorten’s figures of 512,000–662,000 Tutsi dead,
who estimated a pre-genocide Tutsi population of 812,000, using the 1987 administrative
census data where Tutsi represented 10.6% of the population, and that the number of survivors lay between 150,000 and 300,000.28 My estimate is also lower than Prunier’s figure
of 800,000–850,000, who assumed Tutsi represented 12% of the population and that only
130,000 Tutsi survived.29
In comparison with estimates at the higher and lower ends, my estimate is significantly
lower than the Government of Rwanda’s genocide census figure of 1,006,031 Tutsi killed. I
believe this number is not credible. If added to the government’s own survivor census of
309,000 persons (assuming all were Tutsi), it would require the Tutsi to have constituted
1.3 million or 17.8% of Rwanda’s pre-genocide population, a proportion slightly higher
than that in the colonial era. This in turn would imply the exodus of (mainly) Tutsi
during and after Rwanda’s revolution (1959–62) did not take place and would leave the
size of the Tutsi refugee population before the genocide unexplained. In contrast, my estimate is significantly higher than Davenport and Stam’s estimate of between 206,000 and
376,000 Tutsi killed. While they believe the number of survivors lies between 130,000 and
300,000 as several others do, they rely on a much lower pre-genocide Tutsi population of
27
The United Nations High Commission for Human Rights, Report of the Mapping Exercise Documenting the Most Serious
Violations of Human Rights and International Humanitarian Law Committed within the Territory of the Democratic Republic
of the Congo between March 1993 and June 2003 (2010), presents the results of the most serious investigation into the
violence in the D.R.C. in this time period so far. The report does not conclude whether the violence qualified as genocide,
stating this requires a judicial determination, but does state in paragraph 31 several incidents occurred “ … from which a
court could infer the intention to destroy the Hutu ethnic group in the DRC in part … ” In Straus, “The Limits of a Genocide Lens: Violence against Rwandans in the 1990s,” 11, a stronger assessment is offered: “In my view, of the four episodes of violence described in this essay, the sustained, large-scale violence against the Rwandan Hutu population in the
Democratic Republic of Congo is the one that could warrant the label of ‘genocide.’”
28
Verpoorten’s estimate and detailed explanation of how she derived it may be found in an online opinion piece at https://
africanarguments.org/2014/10/27/rwanda-why-davenport-and-stams-calculation-that-200000-tutsi-died-in-thegenocide-is-wrong-by-marijke-verpoorten/ (accessed 15 June 2019).
29
Prunier, The Rwanda Crisis, 1959–1994, 265.
JOURNAL OF GENOCIDE RESEARCH
11
506,000.30 This figure is extrapolated from the 1952 colonial census, taking into account
UNHCR data on refugees from the revolution, and using a population growth rate of 2.5%
per annum up until 1990 when the war begins. The figure of 506,000 Tutsi in Rwanda in
April 1994, however, is dubious. It is sensitive to the UNHCR estimates of the number of
Rwandan refugees in the region, and to the assumed population growth rate. The reliability
of the UNHCR data has already been investigated and questioned given dramatic fluctuations
in the figures over time that remain unexplained.31 It is also significantly lower than the
Rwandan national census’ figure which puts the number at 597,000 in August 1991 or
645,000 in April 1994. These figures are themselves already widely-seen as an under-estimate.
Genocide is an intrinsically exclusionary concept. The desire to reserve the unique
moral and abhorrent nature of the term genocide for the victim group, and to resist
claims by others who suffered violence also to be recognized as genocide victims, is
powerful. The Rwandan government’s decision to specify in the constitution the genocide
was specifically against the Tutsi exemplifies this exclusionary and proprietary urge. The
several estimates for the death toll offered in this forum, however, should leave little
doubt that a genocide targeting Tutsi occurred. The investment of significant scholarly
resources into establishing the genocide’s toll is a worthwhile exercise both for the production of a historical record and for reasons of memory, justice, and accountability.
However, the violence that claimed Rwandan lives in the 1990s was varied and
complex. Other forms of violence took place in Rwanda at the same time as the genocide.
Other forms of violence also preceded and followed the genocide, both inside and outside
of Rwanda, that are linked to it, at least in the minds of those who were its targets. In
setting a future research agenda, scholars should consider the value of committing
resources to documenting the victims of these other kinds of violence. They too
deserve to be remembered, and they too deserve justice.
ORCID
Omar Shahabudin McDoom
http://orcid.org/0000-0001-5660-1903
Disclosure Statement
No potential conflict of interest was reported by the author.
Notes on Contributor
Omar Shahabudin McDoom is an Assistant Professor of Comparative Politics at the London School of
Economics.
The survivor figures of 130,000–300,000 are cited in Davenport and Stam, “Rwandan Political Violence in Space and
Time.” The pre-genocide Tutsi population estimate of 506,000 is cited on their project website in a Powerpoint presentation Stam made in 2012 (slide 18). https://genodynamics.weebly.com/presentations.html (accessed 15 June 2019). In
the BBC documentary Rwanda’s Untold Story, Davenport and Stam are interviewed and Stam repeats the estimate that
200,000 Tutsi may have been killed (at minute 31.05) and the narrator also repeats their estimate of a pre-genocide Tutsi
population of 500,000. However, in an online magazine article, Davenport and Stam cite a different Tutsi death toll of
300,000–500,000 based on a different pre-genocide Tutsi population extrapolated from the 1991 national census. See
https://psmag.com/social-justice/what-really-happened-in-rwanda-3432 (accessed 17 June 2019).
31
See Guichaoua, Le Problème Des Réfugiés Rwandais Et Des Populations Banyarwanda Dans La Région Des Grands Lacs
Africains, 20–24.
30
