Journal of Genocide Research (2004), 6(1), mm Carfax Publishing
Vf/^F   Taylor & France Croup
March, 85-98 m
How many perpetrators were there in the Rwandan genocide? An estimate*
SCOTT STRAUS
Introduction
How many Rwandans participated in the 1994 genocide? Existing estimates are hugely discrepant and often not supported with evidence. Some Rwandan government officials claim there were three million perpetrators.1 Other observers claim there were "hundreds of thousands" (Des Forges, 1999, p 2; Mamdani, 2001, p 7; Scherrer, 2002, p 126; Waller, 2002, p 67).2 Still others estimate tens of thousands (Jones, 2001, p 41).3 The high-end estimate effectively criminalizes the entire adult Hutu population at the time of the genocide.4 The low-end estimate is equivalent to a small fraction of the adult male Hutu population. Which is right? Resolving the question is important not only for understanding the genocide's specifics. A reliable estimate of the number of perpetrators also has implications for the post-genocide environment in Rwanda. Is the current government facing a "criminal population," as some claim, or something far less than that? Calculating the exact number of perpetrators is probably impossible given the evidence that is currently available, but a better estimate than currently exists is possible and important. This article's principal aim is to do that—to provide an estimate that is both empirically based and systematically calculated.
Background
The question about the size of the perpetrator population stems from a larger set of research questions. Initial commentary on the Rwandan genocide claimed the violence was an outbreak of longstanding enmity between the Hutu and Tutsi
*This article emerges from my forthcoming dissertation entitled "The order of genocide: race, power, and war in Rwanda." I would like to thank Ben Valentino, Stephen Stedman, and Michael Mann for encouraging me to write this article. I also would like to thank Gerry Caplan, David Leonard, and Eric Markusen for comments on earlier drafts. Funding for the research and writing of this article comes from the Social Science Research Council International Dissertation Research Fellowship program and from the National Science Foundation Graduate Research Fellowship program. In Rwanda, the research depended on the generous permission of the Ministry of Interior and the Ministry of Local Affairs and Administration to conduct fieldwork in the country.
ISSN 1462-3528 print; ISSN 1469-9494 online/04/010085-14 © 2004 Research Network in Genocide Studies DOI: 10.1080/1462352042000194728
SCOTT STRAUS
ethnic groups.5 Some commentators also argued that the genocide was rooted in state "failure" and state "collapse" in Rwanda.6 Both views treated the violence as a mass, largely unorganized phenomenon without a specific history leading up to the event. These views also tended to elide the specificity of genocide: rather than seeing the violence as an intentional, state-driven campaign of annihilation, the commentary tended to treat the violence as spontaneous and amorphous.
In response, scholars with expertise on the region labored to show how wrong these viewpoints were. Two main themes emerged in this second wave of commentary. On the one hand, scholars stressed the ways in which ethnic identities had been shaped and constructed in the hundred years preceding the genocide. Much emphasis fell on Rwanda's colonial experience where Europeans were seen to have rigidified and racialized previously more fluid ethnic categories and to have privileged one ethnic category over another. This scholarship is rich and detailed but not my emphasis here.7 The point is that this scholarship showed that ethnicity had a specific history in Rwanda and that the ethnic categories on which Rwandans acted during the genocide were neither natural nor "ancient." Rather, the categories were modern and constructed.
On the other hand, scholars stressed the planning and state orchestration of the genocide. Not only did ethnicity have a history, scholars argued, but also political elites deliberately fomented and organized the violence. The analytic emphasis thus shifted to government-prepared assassination lists, to the diffusion of extremist propaganda, to radio incitement, to Rwanda's dense state administration, and, lastly, to the creation and training of militias. In short, rather than seeing the violence as chaotic, spontaneous, and the outcome of state collapse, scholars demonstrated the systematic and planned nature of the killing—and thereby showed that the Rwandan genocide was akin to other calculated genocides.8
Taken together, these two themes came to define a new status quo on the genocide—as a state-organized, planned extermination campaign that served elite interests and drew on constructed ethnic categories. While this part instrumentalist, part constructionist, part statist view made very significant gains in understanding the genocide, it shifted the analytic and empirical emphases to macro historical processes and to the actions of the most senior elites in the country. In so doing, the new consensus left important gaps in understanding and in evidence about how and why the genocide started and spread at the local level. If elites planned and promoted the genocide, how and why did they succeed in mobilizing large numbers of Hutu civilians to join the extermination campaign? What actually happened at the local level and how can researchers find evidence to evaluate that question? How many Rwandans participated in the genocide? And what are the theoretical and practical implications of answers to these and related questions?
My overall research is designed to address these issues. To wit, in 2002, I spent more than seven months in Rwanda researching the genocide's local-level dynamics. My research had three main components: (1) a survey of perpetrators in prisons; (2) case studies of the genocide in five communes; and (3) targeted
86
PERPETRATOR NUMBER IN RWANDAN GENOCIDE
interviews with local-level, alleged leaders of the genocide. This article draws primarily on the results from the first research phase—the survey—and the analysis is limited to one specific question—the size of the perpetrator population. However, that question is derivative of pre-existing scholarship on the genocide and is part of a broader, micro-level inquiry into the dynamics of the genocide.
Definitions
Estimating the number of perpetrators in the genocide requires clear definitions. One reason why existing estimates vary dramatically is because different commentators have different crimes and different events in mind when they refer to "perpetrators of genocide." That being the case, explicit statements on what "perpetrator" means, on the time period under question, on what "genocide" means, and on what "acts of genocide" mean all are critical.
I define a "perpetrator" as any person who participated in an attack against a civilian in order to kill or to inflict serious injury on that civilian. Perpetrators thus would be those who directly killed or assaulted civilians and those who participated in groups that killed or assaulted. During the genocide, there were other kinds of participation. Many Rwandans joined state-ordered civilian night patrols. Sometimes the members of these patrols attacked civilians, and if they did they would be "perpetrators" under my definition. However, if they did not, they would not be. There also was a significant amount of looting during the genocide. If killing accompanied the looting, then looters would be "perpetrators" under my definition. However, if Rwandans looted property, but did not join attacks where civilians were killed, they would not be perpetrators. The central idea behind this definition is that a perpetrator is someone who materially participated in the murder or attempted murder of a non-combatant.9
The time period during which a Rwandan could be considered a "perpetrator" also needs definition. Some observers date the onset of genocide to massacres that occurred after the rebel Rwandan Patriotic Front (RPF) invaded Rwanda on October 1, 1990. Indeed, the current Rwandan government uses that time period—from October 1, 1990 through December 31, 1994—in its legal delineation of genocide crimes and in its statistics on the genocide (Official Gazette, 1996, p 14; Republique Rwandaise, 2002, p 15). I take a narrower view of the time period during which the genocide was committed—as that period when state officials conducted a campaign to physically destroy Rwanda's Tutsi population. That period began immediately after President Juvenal Habyarimana was assassinated on April 6, 1994, and ended on July 19, 1994, when the RPF swore in a new government.
The delineation of this time period rests on a specific definition of "genocide," which also needs to be made clear. The concept of genocide is a much-contested one with scholars often disagreeing about what does and does not qualify as "genocide." Without entering into the details of that debate here, I want to specify the definition I have in mind because it directly affects how I estimate
87
SCOTT STRAUS
the size of the perpetrator population. I take "genocide" to be a time-specific, organized attempt to annihilate a group thought to have ethnic, racial, or otherwise organic properties.10 This definition conceptualizes "genocide" as an intentional effort to physically destroy a population said to share a common "genos"—even if shared biology is a myth, as it likely was for Rwanda's Tutsi population.11 Thus, I would not call a massacre that killed several hundred Tutsi civilians "genocide," but I would use that label for the extermination campaign that began on April 6, 1994, and took the lives of at least 500,000 civilians.12
This definition has other implications for estimating the size of the perpetrator population. During the genocide, Tutsi civilians were not the only ones targeted for murder. Those who attacked Tutsi also killed Hutu civilians for being in the political opposition and for refusing to participate in the genocide. At the same time, RPA soldiers also killed Hutu civilians while fighting government forces and securing control of areas they had conquered.13 My criterion for identifying a perpetrator is whether the attacks were consistent with genocide—with the attempt to annihilate the Tutsi population. Thus, those who participated in attacks against Hutu who resisted genocidal violence or who were in the opposition would be considered genocide perpetrators. By contrast, RPA soldiers who committed violence against civilians would not be.
Genocidal acts were any of the following: murder, attempted murder, sexual violence, torture, and a host of other particularly cruel acts. The latter include impaling, slicing tendons, cutting off breasts, and disemboweling pregnant women, among other acts of brutality (Taylor, 1999, p 140). The main means for carrying out these acts were grenade, gun, machete, impiri (club), sword, knife, drowning, arson, stick, rock, and barehanded assault. According to a government study of the genocide—whose accuracy cannot be verified—death by machete was the most common means of murder (37.9%), followed by killing by club (16.8%), followed by killing by firearm (14.8%) (Republique Rwandaise, 2002, p 26). Killing occurred in four main locations: (1) at central congregation points such as churches, schools, and government buildings; (2) at roadblocks; (3) during house-to-house searches; and (4) during searches through cultivated fields, wooded zones, and marshes (I base this assertion on my own field research in Rwanda). Most massacres (large-scale killings at one time and place) occurred at the central congregation points. A perpetrator is anyone who participated in any of these attacks, using any of these means, and in any of these locations.
Methodology
One method for estimating the number of genocide perpetrators is to do so on the average of two to three attackers per victim, in which case there would be two to three million perpetrators.14 By the same token, however, if one stipulates that there were 500,000 victims and that the average perpetrator killed 10 victims over the course of the genocide, then there would be 50,000 perpetrators.15 These calculations differ in part because each investigator employs a different
88
PERPETRATOR NUMBER IN RWANDAN GENOCIDE
estimate of the number of genocide victims. However, the greater discrepancy stems from a difference in how the genocide itself is conceptualized. For the former estimate, the model is residential killing: the base killing event entails neighbors attacking neighbors at one time and place. For the latter estimate, the model is mobile killing: the base killing event entails soldiers, paramilitaries, and hardcore killers attacking civilians at different times over the course of the genocide period. Judging from my field research in Rwanda, both models are right. In some instances, neighbors killed neighbors on a single day. In other instances, soldiers and paramilitaries killed civilians on different days in different locations.
Indeed, there were significant differences among perpetrators in the level of violence they committed. Many victims died at central congregation points in large massacres. Often these massacres involved military or paramilitary firearm use, such that a single person using several grenades or an AK-47 easily could have killed 10 civilians or more in a single attack. These soldiers and paramilitaries also tended to be mobile and, as such, tended to participate in more than one killing event during the 100-day genocide period. Moreover, during the house-to-house and field-to-field attacks in residential communities, specific individuals often took the lead in killing victims. Some killed 10 or more victims, but many more did not physically kill even if they participated in groups that killed.
In short, there were important, related differences in how the genocide was perpetrated and also in the number of victims that perpetrators tended to kill. This variation needs to be incorporated somehow into an estimate of the number of perpetrators. A calculation premised on an undifferentiated ratio of the average number of victims per perpetrator does not do this.16
A better alternative is to base an estimate of the perpetrator population on average group sizes during the genocide. My field research indicates that patterns of mobilization, particularly among civilians, were similar across Rwanda once genocidal violence started. I found considerable regional variation in when the violence began and in who controlled it. However, I also found that once violence took hold in a particular community, the ways in which civilians were mobilized were similar across the country. If this assumption can be made— namely, that the dynamics of civilian participation were consistent throughout Rwanda—then the key piece of evidence needed to calculate the number of perpetrators is an estimate of the number of perpetrators in every location where genocide occurred. Specifically, the evidence needed is (1) the size of the groups that attacked; (2) the groups' makeup; and (3) an estimate of the number of locations where genocide occurred.
Research design
How can this information be found? One way is to interview survivors, but the main problem with relying on survivor testimony is that their survival often depended on hiding from perpetrators. As such, many survivors do not have
89
SCOTT STRAUS
direct and detailed knowledge about the size or makeup of the groups that attacked them. Similarly, bystanders—those who neither were targeted nor participated in the violence—often rely on hearsay to answer specific questions about perpetrators. The best sources are the perpetrators themselves. However, relying on perpetrator testimony has its own set of problems, and thus care must be taken in how such testimony is collected and evaluated. This section discusses these concerns in detail.
The obvious locations to find genocide perpetrators are Rwandan detention facilities. In 2000, the government held 109,499 detainees on genocide charges (Office of the Prosecutor, 2002a).17 To choose among these, I relied on four criteria for selecting perpetrators to be interviewed. The first criterion was that detainees be sentenced, and in 2002 there were roughly 6,000 prisoners in this category (Office of the Prosecutor, 2002b).18 The principal advantage with interviewing sentenced detainees is that they had fewer incentives to lie than those detainees awaiting trial. Moreover, pre-existing guilty judgments served as a first-stage verification on the crimes committed; the crimes already had been discussed publicly in a court of law; and there was a public record that could be consulted to crosscheck interview testimony.
The second criterion was that the sample be chosen randomly, where possible. In Rwanda, prisoners with relative power and prison officials often choose which prisoners foreign visitors meet. While the intentions of these prisoners and officials are not necessarily bad, my objective was to avoid unintended biases introduced through non-random selection. The third criterion was that those interviewed had already pled guilty.19 In a pilot study, interviews I conducted with sentenced prisoners who denied participation were unhelpful, given that my questions concerned the specifics of how the violence was perpetrated. Finally, given regional variation—both in Rwandan politics and in when and how genocidal violence started—the fourth criterion was that the sample be national.
The method that I chose to incorporate these various criteria was stratified random sampling of sentenced confessees in all prisons where they were detained. In total, I interviewed 210 prisoners in 15 central prisons using this method. The procedure was the following: at each prison, I made an advance request to prison officials for a list of sentenced confessees. On an agreed date of arrival, I obtained the list and used computer-generated random numbers to select persons from this list. In some prisons, the number of sentenced confessees was too small to use a computer-generated number. In one prison, I interviewed all six sentenced confessees in that prison; in four other prisons, I interviewed every other prisoner on the list. For the remaining 10 prisons, I used computer-generated numbers. All interviews took place in a private office with a Rwandan research assistant.20
There are two main concerns with this research design. The first relates to reliability: can the information given by perpetrators be trusted? Perpetrators may outright lie or they may reconstruct events in such a way as to mitigate their own responsibility. My research design takes measures to reduce these risks— chiefly by interviewing sentenced perpetrators and by sampling in different
90
PERPETRATOR NUMBER IN RWANDAN GENOCIDE
prisons (thereby making it less likely that collective reconstruction pertaining to particular prisons or trials would bias the responses). Still, these measures are not foolproof and important reliability concerns remain.
That said, some parts of my interviews with perpetrators are more subject to distortion than others. I asked respondents numerous questions, ranging from their age, to their pre-genocide relations with their Tutsi neighbors, to their motivation to participate in the genocide, to how many people they killed, to the specifics of their attack. The parts of the interviews most subject to distortion concern motivation, interethnic relations, and the crimes committed. By contrast, perpetrators are less likely to distort their answers about their age: they have few incentives to do so. The same is largely true for their estimates of the number of people in their attack. As long as I did not ask them to name other perpetrators in their attack (which I did not do), respondents had few reasons to distort the size of the group in which they participated.
The second main concern relates to whether the sampled perpetrators are representative of the overall perpetrator population. The bottom line here is that I expect the sample to be biased away from the genocide's worst killers. There are three main reasons for this. First, many grands genocidaires never returned to Rwanda after fleeing in 1994. Some were killed in Congo while fighting pro-Rwandan government forces there. In 2002, when I conducted my field research, others were still fighting in Congo or living in exile there.21 Still other grands genocidaires are detained in Tanzania awaiting trial at the International Criminal Tribunal for Rwanda, which is responsible for prosecuting the genocide's major leaders. Second, anecdotal evidence suggests that many of the worst perpetrators who returned to Rwanda were killed when they did. Some also were executed after being sentenced to death. Third, the confessee prisoner population also is biased against the hardcore genocidaires. Judging from my field research, those who pled guilty tended not to be hardcore perpetrators. Moreover, in 2002, the sentenced confessee population did not include soldiers who had been on active duty during the genocide. The net effect is that my sample is biased against those perpetrators who had the greatest responsibility during the genocide. That bias does not derail the methodology, but the bias needs to be remembered when interpreting the results (more on this below).
Estimating the number of perpetrators
My method calls for estimating the number of perpetrators in a single location and then multiplying that number by the number of locations where genocide occurred. This analysis thus requires a consistent geographical unit, and the one I use is the cellule. In 1994, Rwanda had a centralized and hierarchical administration of which the cellule was the smallest unit that existed countrywide. There were approximately 6,300 cellules in 1994. The next administrative level up from the cellule was the sector, of which there were approximately 1,510 in 1994. The next level up from the sector was the commune (or district),
91
SCOTT STRAUS
Table 1. Perpetrator group sizes during the genocide
Group size (number Weighted		Group size per	Weighted
of persons)	frequency	cellule (number of	frequency
		persons per cellule)	
1-10	23.7%	1-10	44.1%
11-30	28.6%	11-20	25.4%
31-50	17.5%	21-30	15.2%
51-100	11.1%	31-40	7.2%
101-200	3.8%	41-50	7%
200 +	15.3%	50 +	1.1%
Weighted average	116 persons	Weighted average	22 persons
of which there were 145, followed by the prefecture (or province), of which there were ll.22
The size and makeup of perpetrator groups varied considerably during the genocide. To generate a base unit for the number of perpetrators per cellule, I first asked respondents how many people were in the attack in which they participated. I then asked them to estimate how many cellules, sectors, or communes were represented in the attack. With these two pieces of evidence, I then estimated the number of perpetrators per cellule by dividing the total group size by the number of cellules represented in the group. For example, an attack of 20 persons from one cellule results in a 20-person per cellule estimate. An attack of 50 persons from two cellules results a 25-person per cellule estimate. An attack of 300 persons from four sectors yields a 19-person per cellule estimate.
Not all respondents could estimate the size of the group in which they participated. Nor could all those who estimated their group size estimate their group's makeup. Of the total sample, 181 respondents estimated their group's size and 166 respondents estimated both their group's size and makeup. Table 1 reports the results.24
To be sure, these results demonstrate more than just average group sizes and the average number of perpetrators from one cellule in various attacks. They also demonstrate that violence tended to be committed in groups and often in groups of considerable size. Indeed, only one respondent in my entire sample said that he attacked on his own. That finding has important theoretical implications, but I want to limit the analysis here to estimating the number of perpetrators. The key result for that endeavor is that attacks consisted of, on average, 22 persons from a single cellule.
The next piece of needed information is the number of cellules in which genocide occurred. Based on currently available information, there is no certain way to derive this number. The method I propose is first to estimate the number of communes where genocide occurred and then to multiply that number by an average number of cellules per commune. Judging from my field research, genocide did not occur in communes under RPF control in 1994, in a demilita-
92
PERPETRATOR NUMBER IN RWANDAN GENOCIDE
rized zone, and in one commune—Giti—under government control.25 That leaves 133 communes where genocide occurred.26 If there were on average 44 cellules per commune, then genocide can be said to have occurred in 5,852 cellules. With this number, a base estimate for the number of perpetrators is 128,744 persons (22 X 5852).
This estimate needs modification. First, in general, respondents participated in civilian attacks and estimated civilian group sizes. If the population of non-residential regular and irregular armed forces totaled 40,000 men in Rwanda at the time of the genocide, a conservative estimate for the number of genocide perpetrators among them would be 10,000. Second, respondents' estimates were for single attacks, not for the genocide's duration. Often the same nucleus of perpetrators participated day-in, day-out in local communities, but group composition also changed over time. I know of no way to calculate this number systematically, but based on my research, my best estimate is an average of 30-35 perpetrators per cellule over the course of the genocide. These modifications suggest a revised estimate of between 185,560 and 214,820 perpetrators, inclusive of regular and irregular armed forces.
However, this number needs further revision. Several communes had few resident Tutsi prior to the genocide, and as such the relative number of attacks there likely would be less. For other communes, the number of attacks also was relatively few because RPF rebels seized those communes quickly. Even if the calculations above are based on a national average, there is a bias against areas with fewer perpetrators. I estimate an additional 11 communes that fit these descriptions.27 Based on this, I would revise the estimate to 15-20 participants per cellule in those communes, leading to a mild down revision of the above figures to 178,300-207,560 total perpetrators.
All told, I estimate between 175,000 and 210,000 active participants in the Rwandan genocide. The main advantages of this estimate are that it is based on (1) a national sample, (2) on direct perpetrator information, and (3) on an explicit methodology. Indeed, if any assumption made here is wrong or if new research yields different evidence, the estimate can be revised. The main weaknesses with the estimate are that it is based on (1) the quality of perpetrator observations about group size and makeup; (2) an assumption of broadly similar dynamics of mobilization across regions once genocidal violence started; and (3) informed guesswork about perpetrators and genocidal violence in those areas of Rwanda not well represented in my sample.
One way to triangulate this result is to look at sentencing trends, which show that about 19% of those tried are acquitted (Office of the Prosecutor, 2002b). The trials are subject to corruption and other problems, and thus are imperfect yardsticks of guilt. But the bias works both ways: some guilty are freed, and some innocent are sentenced. As such, the acquittal rate can be used to estimate the actual number of perpetrators in prison: with 110,000 prisoners, the acquittal rate indicates that 89,100 were perpetrators. In addition, there are perpetrators in Rwanda at large; there are perpetrators who have been killed; and there are perpetrators outside the country. Estimating these numbers is impossible at this
93
SCOTT STRAUS
stage, but it seems reasonable to expect that the uncounted populations would be roughly similar to the number of perpetrators in prison.28 Thus, by this method, an estimate of 175,000-210,000 perpetrators also seems plausible.
Before turning to the implications of this finding, a final word on the assumption of broadly similar patterns of participation across regions once the violence started. This assumption is key because it allows a generalized, national calculation. One way to test it is to examine regional variation in prisoner populations, and indeed the relative size of prisoner populations does vary between prefectures. However, this measure has a major validity problem: the number arrested in a particular area does not necessarily reflect the number of genocide perpetrators from that area. Regional variations in prisoner populations reflect local politics and policies, regional differences in population sizes, regional differences in refugee levels, and regional differences in the level of post-genocide killing.29
Another way to test the assumption is to analyze regional variations in my sample of the average group sizes per cellule. Those averages for Rwanda's 10 rural prefectures are as follows: 38 for Butare, 16 for Byumba, 15 for Cyangugu, 10 for Gikongoro, 18 for Gitarama, 15 for Gisenyi, 13 for Kibuye, 21 for Kibungo, 35 for Rural Kigali, and 12 for Ruhengeri.30 These results exhibit somewhat consistent per cellule average group sizes with two main exceptions: Butare and Rural Kigali prefectures, where the average per cellule group sizes are notably higher than those for other prefectures. However, the small number of observations per prefecture limits the inferences that can be derived from these numbers. The bottom line here is that these results do not scuttle the assumption that average per cellule group sizes were broadly similar countrywide. Rather, the results show there were some regional differences in my sample of the average number of perpetrators per cellule, but the significance of those differences remains an open question. More research on this issue is needed.
If the figure of 175,000 to 210,000 perpetrators holds up as more evidence becomes available, the estimate has important implications. The figure supports the claim that mass participation characterizes the Rwandan genocide. Compared to other genocides and most state-sponsored mobilization campaigns in Africa, some 200,000 perpetrators is an extraordinary number. However, the estimate detracts significantly from the allegation that the current authorities govern a "criminal population."31 If active adults are defined as 18-54 years old, 175,000 to 210,000 perpetrators equals 7% and 8% of the active adult Hutu population at the time of the genocide and 14% to 17% of the active adult male population.32
There is another implication, but one that first requires several assumptions. As mentioned above, there were significant differences among perpetrators in their levels of killing. In my sample, the average number of victims that respondents admit to killing was 0.7. If one accepts the following—that the perpetrators told the truth when saying how many people they did or did not kill, that my sample is representative of non-hardcore civilian perpetrators, and that
94
PERPETRATOR NUMBER IN RWANDAN GENOCIDE
there were 500,000 genocide victims—then my calculation of the total number of perpetrators would strongly suggest that a small minority of perpetrators did the majority of killing.
Assume, for example, that there were 200,000 perpetrators, of whom 180,000 were non-hardcore civilian perpetrators and 20,000 were soldiers, paramilitaries, and extremely zealous killers during the genocide. If the average killing rate for the former category of perpetrators were 0.7 deaths, then the participation of this category of perpetrators would account for 25% of all genocide deaths, even though this category represented 90% of all perpetrators. If true, that calculation would mean that the latter category of perpetrators accounted for 75% of all genocide deaths, even though this category represented 10% of all perpetrators. The exact numbers used in these calculations are likely not correct, but the idea driving the calculations probably is: even if mass participation characterizes the Rwandan genocide, a small number of armed perpetrators and especially zealous ones probably did the lion's share of the killing.
Conclusion
This article has been limited to a narrow, but important question: how many Rwandans participated in the 1994 genocide? The article seeks an estimate based on an explicit methodology and micro-level field research. The net calculation is a perpetrator population of around 200,000 Rwandans. That estimate has implications for post-genocide Rwanda because it suggests that collective blame of "the Hutus" should be eschewed. Not all Hutu adults were genocidaires. Moreover, not all genocidaires participated to the same degree. Some killed many victims; others did not kill. As such, when analyzing participation in the Rwandan genocide, the category of "perpetrators" should be disaggregated.
The findings in the article also have theoretical implications. Even if not all Hutus participated in the genocide, an enormous number did. This mass participation requires explanation. What drove so many Rwandans to participate in genocidal violence? How can evidence be found to assess that question? The analysis also indicates that not all perpetrators had the same level of participation and that a large number of adult Hutu did not participate in the genocide. Thus, why did some participate more than others and why did some participate while others did not? These are critical questions, but ones that I hope to answer elsewhere.
Notes and References
1. The figure of three million perpetrators has been cited to me in several formal interviews and informal conversations with senior government officials in Rwanda over the course of visits to the country in 1998, 2000, and 2002. See also Philip Gourevitch, We Wish to Inform You that Tomorrow We Will Be Killed with Our Families: Stories from Rwanda (New York: Farrar Straus and Giroux, 1998), p 244, where the author quotes a former presidential adviser who cites the figure of three million perpetrators.
2. Note that in addition to citing "hundreds of thousands" of perpetrators, Mahmood Mamdani also quotes as evidence of mass participation an RPF "political commissar" who claims the RPF faced a "criminal population" when it took power in 1994. Christian Scherrer claims that 40-66% of male Hutu farmers,
95
SCOTT STRAUS
60-80% of the higher professions, and "almost 100%" of Rwanda's civil servants participated in the genocide. No substantiation is offered for the latter claim, and a German reference is offered for the former two. If properly calculated, those numbers would total more than a million perpetrators. Indeed, Scherrer also claims that "every fourth person in Rwanda's Hutu population" was "probably directly involved"—a number that would total about 1.25 million—and that "millions rendered themselves indirectly responsible" (p 126). Scherrer elsewhere claims that "About half the male Hutu farming population ... must have been actively involved in the gang killings or the collective persecution of the victims" (p 115).
3. Note that Des Forges also claims there were "tens of thousands" of perpetrators who chose to participate "quickly and easily" (p 2). These are in contrast to "hundreds of thousands" who participated "reluctantly." Jones refers to 25,000 perpetrators as a low-end estimate and 100,000 as a high-end estimate.
4. According to the 1991 census, Rwanda had 2,813,232 citizens between 18 and 54. If, according to the census, 8.4% were Tutsi, then the adult Hutu population would have consisted of 2,576,920 persons. Because of population growth between 1991 and 1994, the actual number was higher, and some perpetrators were younger than 18 and older than 54. Still, the three million estimate would encompass the entire adult Hutu population. For the census figures, see Republique Rwandaise, "Recensement general de la population et de l'habitat au 15 Aout 1991: analyse des resultats definitifs," Kigali, April 1994, p 124.
5. There are many examples of journalists and commentators quoted in news reports citing "ancient tribal hatreds" while the genocide occurred. For an example, see a very revealing exchange reproduced in Samantha Power, "A Problem from Hell": America and the Age of Genocide (New York: Basic Books, 2002), pp 355-356. For another example, see Roger Rosenblatt, "The killer in the next tent: the surreal horror of the Rwanda refugees," New York Times Magazine, June 5, 1994, p 40.
6. For examples where commentators viewed the genocide as emerging from state "collapse" or "failure," see Eliane Sciolino, "For West, Rwanda is not worth the political candle," The New York Times, April 15, 1994 and I. William Zartman, "Introduction: posing the problem of state collapse," in I. William Zartman, ed., Collapsed States: The Disintegration and Restoration of Legitimate Authority (Boulder: Lynne Rienner, 1995), p 4.
7. Indeed, there is a great deal of work done in this vein. For examples, see Jean-Pierre Chretien, The Great Lakes of Africa: Two Thousand Years of History, trans. Scott Straus (New York: Zone Books, 2003); Mahmood Mamdani, When Victims Become Killers; and Gerard Prunier, The Rwanda Crisis: History of a Genocide (New York: Columbia University Press, 1995).
8. Most human rights accounts of the genocide take this view and stress the factors cited in this paragraph. The best of these books is Alison Des Forges, Leave None to Tell the Story. See also African Rights, Death, Despair, and Defiance, rev. ed. (London: African Rights, 1995); Article 19, Broadcasting Genocide: Censorship, Propaganda, & State-Sponsored Violence in Rwanda 1990—1994) (New York: Article 19, 1996); and Jean-Pierre Chretien et al., Les Medias du Genocide (Paris: Karthala, 1995).
9. With one exception, the definition of perpetrator that I use is consistent with the Rwandan government's definition of genocide criminals. Rwandan law lists four categories of criminals, and the difference between my definition and this law concerns the fourth category, which covers property crimes. My definition of a perpetrator excludes property crimes if the crime was not part of a physical attack on a civilian. The four categories in Rwandan law are as follows: Category one is for: "a) Persons whose criminal acts or whose acts of criminal participation place them among the planners, organizers, instigators, supervisors and leaders of the crime of genocide or of a crime against humanity; b) persons who acted in positions of authority at the national, prefectoral, communal, sector, or cell level, or in a political party, the army, religious organizations or in a militia and who perpetrated or fostered such crimes; c) Notorious murderers who by virtue of the zeal or excessive malice with which they committed atrocities, distinguished themselves in their areas of residence or where they passed; d) Persons who committed acts of sexual torture." Categories two is for: "Persons whose criminal acts or whose acts of criminal participation place them among perpetrators, conspirators, or accomplices of international homicide or of serious assault against the person causing death." Category three is for: "Persons whose criminal acts or whose acts of criminal participation make them guilty of other serious assaults against the person." Category four is for: "Persons who committed offences against property." Again, my definition of a perpetrator includes the first through third categories, but not the fourth. For the Rwandan government's definitions of genocide criminals, see Official Gazette of the Republic of Rwanda, "Organic law on the organization of prosecutions for offences constituting the crime of genocide or crimes against humanity committed since 1 October 1990," Kigali, September 1, 1996, p 15.
10. For an overview of different definitions of genocide and for an extended discussion of the basis for my definition, see Scott Straus, "Contested meanings and conflicting imperatives: a conceptual analysis of genocide," Journal of Genocide Research, Vol 3, No 3, 2001, pp 349-375.
11. Again, there is a large volume of work on this question, but the point I am making is that most scholars dispute the notion that Tutsi formed a race and many doubt that the Tutsi formed a contiguous ethnic group
96
PERPETRATOR NUMBER IN RWANDAN GENOCIDE
with a common ancestry. As indicated earlier, when Europeans colonized this region of Africa in the late nineteenth and twentieth centuries, they imposed racial categories on identities that were previously more fluid. They also elevated the importance of the Hutu/Tutsi cleavage, as opposed to other identity-based cleavages, such as clan, affiliation with the monarchy, region, and so forth. Moreover, Rwandan Hutu and Tutsi speak the same language and, by the time of the genocide, many Hutu and Tutsi had intermarried, such that the notion that "the Tutsi" formed a bounded, biologically related ethnic or racial group is probably false.
12. Estimates of the number of victims in the Rwandan genocide vary. The figure I cite is from Alison Des Forges, Leave None to Tell the Story, pp 15-16.
13. RPA stands for the Rwandan Patriotic Army, which was the armed wing of the RPF.
14. Although I have never seen this specific estimation in print, I have frequently heard this calculation during interviews with government officials and with Tutsi survivors in Rwanda during field trips in 2000 and 2002.
15. Bruce Jones uses this logic in his analysis of the size of the perpetrator population (2001, p41).
16. In theory, if one had a representative sample of the perpetrator population, and an average was taken of how many victims each killed, then a calculation of the number of perpetrators based on the number of victims would make sense. As of now, such data do not exist.
17. These statistics were the latest available when I conducted field research in Rwanda in 2002.
18. According to government statistics, by June 2002, 7,211 detainees had been judged, of whom 1,386 had been acquitted.
19. Rwanda's law for prosecuting genocide crimes includes a provision for reduced sentences if suspected perpetrators confess and plead guilty to their crimes. See Official Gazette of the Republic of Rwanda, "Organic law," pp 16-18.
20. For further details on these points, see my forthcoming dissertation.
21. After the RPF won the civil war and ended the genocide in 1994, some two million Rwandans fled the country as refugees. An estimated 1.2 million went to the Democratic Republic of Congo (then called Zaire), and the remainder went principally to Tanzania. In late 1996, the new, RPA-dominated Rwandan army invaded Congo to destroy the refugee camps there because, among other reasons, the behavior of exiled leaders in those camps posed a security threat. Rwandan government forces ultimately succeeded in that effort: the camps were broken up and many refugees returned home. However, many refugees fled westward—that is, deeper into Congo—and Rwandan government forces followed them there. In fact, Rwandan government forces went on to spearhead an insurgency in Congo that ultimately toppled the Congolese government in May 1997. A little more than a year later, the newly imposed Congolese leader, Laurent Kabila, fell out with his Rwandan backers and there began a second and more complicated war in Congo that lasts to the time of my writing this article (July 2003). During the first Congo war (1996-1997), Rwandan government forces killed a large number of Rwandan exiles and refugees. The exact number is not known, nor is it likely to be known, but estimates range from 10,000 to 200,000. Moreover, there remain an untold but significant number of exiles who continue to fight the current Rwandan government from Congo or who remain in Congo as refugees. Not all of Rwandans killed or exiled in Congo were perpetrators, but some were. Moreover, it is reasonable to assume that those who had the greatest responsibility in the genocide had the greatest incentive to fight and the fewest incentives to return to Rwanda.
22. The counts are based on the last administrative list printed before the genocide that I could find: Republique Rwandaise, "Annex a l'arrete presidentiel No 251/03 du novembre 1975," Journal Officiel, No 22, November 15, 1975. This decree lists 143 communes, 1,489 sectors, and 6,213 cellules, where the average number of sectors per commune is 10.4 and the average number of cellules per commune is 43.5. At least two communes that existed in 1994 were not listed in this decree—Kacyiru and Kicukiro—and so I have added 21 sectors and 87 cellules to the total listed.
23. For these calculations, I use the following ratios: 44 cellules per commune and four cellules per sector.
24. Note that I have weighted the results to reflect varying sampling ratios in each prison where I interviewed.
25. The government's official report on the genocide lists 154 communes where genocide occurred. However, this is a technical impossibility because that many communes did not exist in Rwanda in 1994. See Republique Rwandaise, "Denombremement des victimes du genocide."
26. In addition to Giti, those communes where I calculate that genocide did not occur include Kiyombe, Muvumba, Kivuye, Cyumba, Mukarange (Byumba), Kigombe, Kinigi, Butaro, Nkumba, Cyeru, and Kidaho (Ruhengeri).
27. The communes are: Tare, Rushashi, Musasa (Rural Kigali), Nyakinama, Nyamugali, Nyamutera, Ruhondo (Ruhengeri), Gaseke (Gisenyi), Tumba, Kinyami, and Rutare (Byumba). I identify these communes on the basis of my field research and a government document that lists the number of victims in every commune. The 11 communes that I list here existed in 1994 and had less than 2% of all genocide deaths per
97
SCOTT STRAUS
prefecture in the report. See Republique Rwandaise, "Denombremement des victimes du genocide," pp 41-176.
28. The main basis for this assertion concerns events in neighboring Democratic Republic of Congo, where many perpetrators were killed or where many remain. For more on that point, see note 23. As for the other uncounted populations—namely, Rwandans in Rwanda who are genocide perpetrators but who are not in prison and perpetrators who were killed in Rwanda either during the genocide or after—those numbers are impossible to estimate with any accuracy. But based on my observations in Rwanda, I would argue that the sum of these two uncounted populations is in the tens of thousands.
29. For details on these latter points, see notes 23 and 30.
30. Note that these estimates are weighted to reflect variations in my sampling.
31. As has been claimed in interviews conducted with Rwandan government officials and as cited in Mahmood Mamdani (2001, p7).
32. According to the census, Rwanda had 2,813,232 citizens between 18 and 54 in 1991. If, according to the census, 8.4% were Tutsi, then the Hutu population would be 2,576,920 individuals, of which 48.7% were active men. Thus, the total number of active adult Hutu men would be approximately 1,255,960. For the census figures, see Republique Rwandaise, Recensement general, pp 74, 124.
Bibliography
Des Forges, A. (1999) Leave None to Tell the Story: Genocide in Rwanda (New York: Human Rights Watch).
Jones, B. (2001) Peacemaking in Rwanda: The Dynamics of Failure (Boulder: Lynne Rienner).
Mamdani, M. (2001) When Victims Become Killers: Colonialism, Nativism, and the Genocide in Rwanda
(Princeton: Princeton University Press). Office of the Prosecutor (2002a) "Abantu Bafungiye mu Magereza Kasho na Burigade," Ministry of Justice,
Kigali, August.
Office of the Prosecutor (2002b) "Decisions judiciares rendues par annees (Dec. 1996-Juin 2002)," Ministry of Justice, Kigali, August.
Republique Rwandaise (2002) "Denombremement des victimes du genocide: rapport final," Ministere de 1'Administration locale, de l'lnformation, et des Affaires sociales, Kigali, November.
Scherrer, C. (2002) Genocide and Crisis in Central Africa: Conflict Roots, Mass Violence, and Regional War (Westport, CT: Praeger).
Taylor, C. (1999) Sacrifice as Terror: The Rwandan Genocide of 1994 (Oxford: Berg).
Waller, J. (2002) Becoming Evil: How Ordinary People Commit Genocide and Mass Killing (Oxford: Oxford University Press).
98