1) There were several forms of violence taking place at once
In certain respects, we are all quite familiar with what took place in Rwanda between April and July 1994 – 100 days that will be remembered for some time (for quick background click here). As discussed now globally, a group of extremist Hutu (members of the Rwandan Armed Forces [the FAR], the Presidential Guard, national police, the “Zero Network death squads” as well as affiliated militias: the Interahamwe and Impuzamugambi, hereafter the FAR+) targeted their ethnic rivals – the Tutsi (Gourevitch; Human Rights Watch; ICTR–Prosecution; Mamdani; Power; Prunier; Straus). The FAR+ systematically engaged in their abuse and killing. This readily and appropriately led to claims of genocide – the systematic attempt of political authorities in Rwanda to eliminate, in whole or in part, members of an ethnic group and, indeed, some observers referred to the events in question as the clearest example of the concept since the Holocaust. The only variation among those advocating this position was exactly how many people were involved with the killing. The controversy: Some highlighted a small clique whereas others highlighted a large proportion of the Hutu population.
A growing and diverse group of academics, politicians and students of African politics, however, disagree with this exclusive characterization, noting that other forms of killing were also occurring at the same time and place. Typically, the identity of perpetrators, victims and motivations differentiate these “other forms”.
For example, many of the researchers above generally acknowledge that ethnic Tutsi were not the only targets of the FAR+, identifying that politically “moderate” Hutu were targeted as well. This was done in order to eliminate resistance to the government (commonly referred to as politicide in the literature [Harff and Gurr; Harff]). Some note that both Hutu and Tutsi (combatants as well as civilians) were killed in the civil war being fought between the FAR+ and the Rwandan Patriotic Front (hereafter the RPF). These could alternatively be classified as battle-deaths, war crimes or crimes against humanity, depending upon the identity of victims, motives, circumstances and manner in which people were killed. There are others who note that there was a large degree of random political violence taking place or what is referred to as “wilding” (Fujii; Hatzfield). In this situation, ordinary/non-government affiliated Rwandans (both Hutu and Tutsi) squared off against other Rwandans (both Hutu and Tutsi) in an attempt to exact revenge for personal wrongs, financial gain or collective hatred – some ethnic, some political and some idiosyncratic in nature. More recently, individuals have argued that the RPF engaged in reprisal killings as they entered Rwanda (Gersony/UN; Des Forges/Human Rights Watch; Pean; Spanish court). As this is viewed as being disconnected from the military engagements undertaken during civil war but involved with state consolidation, these represent a distinct form of political violence. Finally, there are those who suggest that all of the different forms of conflict identified above were present (ICTR–Defense). Here, within a Hobbesian “war of all against all”, all hell broke loose.
What led us to this conclusion? Listening to what eyewitnesses could recall, what NGOs and INGOs recorded, what diverse scholars had written and putting it all together. Additionally, mapping the violence also suggested the same conclusion (i.e., creating a database of political violence and showing the violence move through space and time). This is discussed more under the next point.
2) The perpetrators include the Hutu government of 1994, ordinary Rwandans as well as the Rwandan Patriotic Front (RPF)
Another major insight from our research concerns the fact that there were different types of violence that took place that can only be determined from viewing the conflict spatially as well as temporally. Individuals were killed all throughout Rwanda during 1994 but exactly when and where these deaths were generated as well as who was involved in the violence and who was targeted is extremely important. These characteristics are crucial because the control over different locales varied over time. At certain points, the FAR (the extremist government) were in control of the relevant territory and they engaged in violence, in other times it is not clear who was in control (the rebels as well as the government were fighting to get the advantage) and in still other times the RPF was in control. This control issue is important for understanding and classifying the violence. For example, if the extremist Hutu (FAR) government was in control of territory where violence occurred, then the killings could be genocide, politicide or random violence = depending upon exactly who engaged in the killing. What is important is that the state/government was in control. If neither the state nor the rebels were in control, then killings would fall under the category of civil war. If the rebels (the RPF) were in control, however, then killings could be reprisal or random.
What led us to this conclusion? Definitions developed in social science research largely based in political science and sociology, occasionally overlapping with law were the principal reason. Essentially, different forms of violence are largely categorized by what perpetrators target which victims. Employing a conceptualization developed by Christian Davenport (below) one will see that Genocide involves government (Instigator/Perpetrator) targeting Civilians/Society/Societal Group. Other forms of violence are identified as well. For example, Civil War involves a government and challenger (political challenger) coercively engaging with one another. As civilians target other civilians this is referred to as "communal conflict", "(intra) group conflict" or "ethnic conflict". As one can see from the chart, violence in Rwanda can be categorized in many cells at the same time. This should prompt us to identify what was taking place as well as who was doing it, against whom and why.
In certain respects, we are all quite familiar with what took place in Rwanda between April and July 1994 – 100 days that will be remembered for some time (for quick background click here). As discussed now globally, a group of extremist Hutu (members of the Rwandan Armed Forces [the FAR], the Presidential Guard, national police, the “Zero Network death squads” as well as affiliated militias: the Interahamwe and Impuzamugambi, hereafter the FAR+) targeted their ethnic rivals – the Tutsi (Gourevitch; Human Rights Watch; ICTR–Prosecution; Mamdani; Power; Prunier; Straus). The FAR+ systematically engaged in their abuse and killing. This readily and appropriately led to claims of genocide – the systematic attempt of political authorities in Rwanda to eliminate, in whole or in part, members of an ethnic group and, indeed, some observers referred to the events in question as the clearest example of the concept since the Holocaust. The only variation among those advocating this position was exactly how many people were involved with the killing. The controversy: Some highlighted a small clique whereas others highlighted a large proportion of the Hutu population.
A growing and diverse group of academics, politicians and students of African politics, however, disagree with this exclusive characterization, noting that other forms of killing were also occurring at the same time and place. Typically, the identity of perpetrators, victims and motivations differentiate these “other forms”.
For example, many of the researchers above generally acknowledge that ethnic Tutsi were not the only targets of the FAR+, identifying that politically “moderate” Hutu were targeted as well. This was done in order to eliminate resistance to the government (commonly referred to as politicide in the literature [Harff and Gurr; Harff]). Some note that both Hutu and Tutsi (combatants as well as civilians) were killed in the civil war being fought between the FAR+ and the Rwandan Patriotic Front (hereafter the RPF). These could alternatively be classified as battle-deaths, war crimes or crimes against humanity, depending upon the identity of victims, motives, circumstances and manner in which people were killed. There are others who note that there was a large degree of random political violence taking place or what is referred to as “wilding” (Fujii; Hatzfield). In this situation, ordinary/non-government affiliated Rwandans (both Hutu and Tutsi) squared off against other Rwandans (both Hutu and Tutsi) in an attempt to exact revenge for personal wrongs, financial gain or collective hatred – some ethnic, some political and some idiosyncratic in nature. More recently, individuals have argued that the RPF engaged in reprisal killings as they entered Rwanda (Gersony/UN; Des Forges/Human Rights Watch; Pean; Spanish court). As this is viewed as being disconnected from the military engagements undertaken during civil war but involved with state consolidation, these represent a distinct form of political violence. Finally, there are those who suggest that all of the different forms of conflict identified above were present (ICTR–Defense). Here, within a Hobbesian “war of all against all”, all hell broke loose.
What led us to this conclusion? Listening to what eyewitnesses could recall, what NGOs and INGOs recorded, what diverse scholars had written and putting it all together. Additionally, mapping the violence also suggested the same conclusion (i.e., creating a database of political violence and showing the violence move through space and time). This is discussed more under the next point.
2) The perpetrators include the Hutu government of 1994, ordinary Rwandans as well as the Rwandan Patriotic Front (RPF)
Another major insight from our research concerns the fact that there were different types of violence that took place that can only be determined from viewing the conflict spatially as well as temporally. Individuals were killed all throughout Rwanda during 1994 but exactly when and where these deaths were generated as well as who was involved in the violence and who was targeted is extremely important. These characteristics are crucial because the control over different locales varied over time. At certain points, the FAR (the extremist government) were in control of the relevant territory and they engaged in violence, in other times it is not clear who was in control (the rebels as well as the government were fighting to get the advantage) and in still other times the RPF was in control. This control issue is important for understanding and classifying the violence. For example, if the extremist Hutu (FAR) government was in control of territory where violence occurred, then the killings could be genocide, politicide or random violence = depending upon exactly who engaged in the killing. What is important is that the state/government was in control. If neither the state nor the rebels were in control, then killings would fall under the category of civil war. If the rebels (the RPF) were in control, however, then killings could be reprisal or random.
What led us to this conclusion? Definitions developed in social science research largely based in political science and sociology, occasionally overlapping with law were the principal reason. Essentially, different forms of violence are largely categorized by what perpetrators target which victims. Employing a conceptualization developed by Christian Davenport (below) one will see that Genocide involves government (Instigator/Perpetrator) targeting Civilians/Society/Societal Group. Other forms of violence are identified as well. For example, Civil War involves a government and challenger (political challenger) coercively engaging with one another. As civilians target other civilians this is referred to as "communal conflict", "(intra) group conflict" or "ethnic conflict". As one can see from the chart, violence in Rwanda can be categorized in many cells at the same time. This should prompt us to identify what was taking place as well as who was doing it, against whom and why.
With the objectives of understanding what happened in mind, we focused on the three broad categories of violence that have received the most attention in the Rwandan case: 1) state (FAR)-affiliated (between Government against Civilians), 2) on the battle front (between Government against Political Challenger [and vice versa]) and 3) rebel (RPF)-affiliated (between Political Challenger against Civilians). Using the data we collected (described below), we "mapped" the violence (i.e., we estimated violent activity by location/time/actor and stitched together separate images to create one unified film). The following animations illustrate these efforts, beginning with an animation of battlefronts, an animation of violent deaths and then an overlay of deaths with battlefronts. More animations are found on the Data Animations/Films page.
Animation of Battle-fronts (.mp4)
Animation of Violent Deaths and Battle-fronts
What is clear from the animations is that violence took place in three distinct areas or zones, under the jurisdiction of three distinct contexts: 1) under the control of the FAR, 2) under the control of the RPF and 3) on the battlefront between the FAR and the RPF.
3) The majority of victims were likely Hutu and not Tutsi
According to conventional wisdom, government personnel targeted citizens – either directly or indirectly through militias and citizens targeted their neighbors. In addition to this, we add two other dynamics: government personnel targeting rebels and rebels targeting citizens. Unfortunately, the distribution of activities across categories is never discussed. Much is made of citizens turning on their family members but given the large size of many of the killings identified above, this was clearly not the way that the majority of individuals in Rwanda died. An emphasis on the inter-personal dynamics makes for great book and article titles but it does match up with the evidence.
The information on behavioral distributions is important because it assists one in understanding why activities were undertaken. If, for example, the majority of the victims were produced from government targeting citizens – which was the case above, then it compels the researcher to discover exactly what type of relationship existed between citizens and authorities. On this point, there appears to be diverse opinions. In one view, governments are killing a citizenry that is clearly identified, categorized and defined. In another view, however, governments are killing a citizenry that they generally find unclear, mobile and alien. We support the latter position and thus suggest that the killings were more driven by the clearer defined enemy from without than to the less clearly defined enemy within; this actually also explains the behavior of the RPF as they came into the country.
What leads us to this conclusion? There are several pieces of information.
First, there was massive population displacement within the country which hindered state as well as rebel observational capacity. As the International Panel of Eminent Personalities notes (page 37), between 1990-1992, RPF
Second, there has been much discussion of the ability of individuals within Rwanda to discern between Hutus and Tutsis. Essentially, these are distributional issues like that identified below in the figure (developed by Allan Stam):
Animation of Battle-fronts (.mp4)
- Brief Powerpoint discussion of how maps were created
- For the location of troops we did the following: 1) we began with Defense estimates noted in Kuperman, 2) we adjusted these in accordance to declassified CIA documents and 3) we compared these against estimates provided by then RPF staff member Ruzibiza to develop an initial movement. We then had 4 former members of the RPF check the animation and followed this with an evaluation of several FAR defendants at the ICTR.
Animation of Violent Deaths and Battle-fronts
- Median estimates with troop overlay (daily counter)
- Median estimates with troop overlay (dual counters)
- Upper estimates with troop overlay (dual counters)
What is clear from the animations is that violence took place in three distinct areas or zones, under the jurisdiction of three distinct contexts: 1) under the control of the FAR, 2) under the control of the RPF and 3) on the battlefront between the FAR and the RPF.
3) The majority of victims were likely Hutu and not Tutsi
According to conventional wisdom, government personnel targeted citizens – either directly or indirectly through militias and citizens targeted their neighbors. In addition to this, we add two other dynamics: government personnel targeting rebels and rebels targeting citizens. Unfortunately, the distribution of activities across categories is never discussed. Much is made of citizens turning on their family members but given the large size of many of the killings identified above, this was clearly not the way that the majority of individuals in Rwanda died. An emphasis on the inter-personal dynamics makes for great book and article titles but it does match up with the evidence.
The information on behavioral distributions is important because it assists one in understanding why activities were undertaken. If, for example, the majority of the victims were produced from government targeting citizens – which was the case above, then it compels the researcher to discover exactly what type of relationship existed between citizens and authorities. On this point, there appears to be diverse opinions. In one view, governments are killing a citizenry that is clearly identified, categorized and defined. In another view, however, governments are killing a citizenry that they generally find unclear, mobile and alien. We support the latter position and thus suggest that the killings were more driven by the clearer defined enemy from without than to the less clearly defined enemy within; this actually also explains the behavior of the RPF as they came into the country.
What leads us to this conclusion? There are several pieces of information.
First, there was massive population displacement within the country which hindered state as well as rebel observational capacity. As the International Panel of Eminent Personalities notes (page 37), between 1990-1992, RPF
- advances through the north and north-east, combined with the government’s ... anti-Tutsi propaganda, produced a massive movement of terrified Hutu into settlement camps in the centre of the country. In a short time, close to 300,000 Rwandans, mostly Hutu, had been driven from or had fled their land to become “internally displaced persons” (the term used to distinguish refugee groups who do not flee across national boundaries) within their own country. In early 1993, another large-scale RPF attack led to a further million, again mostly Hutu, being displaced (emphasis added).
Second, there has been much discussion of the ability of individuals within Rwanda to discern between Hutus and Tutsis. Essentially, these are distributional issues like that identified below in the figure (developed by Allan Stam):
Some suggest that there are clear distinctions that could be made between Rwanda ethnic groups. Drawing upon an older, admittedly racist Belgium phrenological approach, this generally falls upon the stereotypical conception of Hutus (as short, darker and stocky) and Tutsi (as taller, lighter skin and thin). These are identified on the left side of the figure. In this view of clear differentiation (depicted at the bottom-right of the figure), there is low variability in ethnic identification; if asked to select, one could easily do so. Violence undertaken in this context directed against members of the rival community is likely straightforward. In this context, activities could be enacted with pinpoint accuracy. Others suggest that there is a moderate degree of variability in Rwandan ethnic identity (depicted at the top-right of the figure). Here, it is a bit more difficult to separate between groups and violence would be a bit less precise. Still others (and we would agree with this group) suggest that the degree of intermarriage and similarities across groups is so great that most individuals could not be distinguished (depicted in the middle-right of the figure). In this context, those engaged in violent activities would be the most careless as well as costly and the death tolls would be extremely high.
The last point suggests that the targets of Rwandan state violence might be a bit less clear than discussed within the conventional position. Our consideration of existing data suggests that this is most likely the case.
Considering the number of Tutsi survivors after Rwandan political violence (between 130,000 to 300,000 by our best estimate) compared to the number that existed before relevant activity (506, 000 by our estimation), we find there is a range of Hutu victims between 64,000 and 1,000,000+ – depending upon which source(s) one utilized. This suggests that the killing was a lot less precise than many account for. Clearly evidence of genocide is evident within this analysis (as large-scale violence is undertaken in territory under the control of the FAR) but it is also clear that a variety of other activities exist as well which merit discussion and consideration within journalistic, scholarly, legal and political circles. Part of the reason for suggesting what we do is that no clear evidence has been presented regarding the victims of each killing, across the whole country, by the day. This type of information is essential for classifying the different types of violence that took place (e.g., identifying the ethnicity and/or political identity of the victim) as well as information about the perpetrator. Our effort attempts to draw conclusions of the likely identity of both the perpetrator and victim derive from battle fronts as well as census records and projected movements of the population.
The last point suggests that the targets of Rwandan state violence might be a bit less clear than discussed within the conventional position. Our consideration of existing data suggests that this is most likely the case.
Considering the number of Tutsi survivors after Rwandan political violence (between 130,000 to 300,000 by our best estimate) compared to the number that existed before relevant activity (506, 000 by our estimation), we find there is a range of Hutu victims between 64,000 and 1,000,000+ – depending upon which source(s) one utilized. This suggests that the killing was a lot less precise than many account for. Clearly evidence of genocide is evident within this analysis (as large-scale violence is undertaken in territory under the control of the FAR) but it is also clear that a variety of other activities exist as well which merit discussion and consideration within journalistic, scholarly, legal and political circles. Part of the reason for suggesting what we do is that no clear evidence has been presented regarding the victims of each killing, across the whole country, by the day. This type of information is essential for classifying the different types of violence that took place (e.g., identifying the ethnicity and/or political identity of the victim) as well as information about the perpetrator. Our effort attempts to draw conclusions of the likely identity of both the perpetrator and victim derive from battle fronts as well as census records and projected movements of the population.
How do we arrive at specific figures? This is somewhat detailed and is discussed thoroughly here. We offer a shorter description below.
Essentially, we rely upon a technique (called Bayes) where information is taken from a variety of different sources, overlaps are identified and then an estimation is provided with some general sense of error. This proves to be ideal in many respects for different sources each have their strengths as well as weaknesses and they cover diverse units of observation. For example, some cover the whole country but only for specific types of activity (e.g., the report from the Rwandan Ministry of Education). Some cover specific parts of the country for a variety of different types of activities (e.g., the report from the Rwandan human rights group African Rights). These are displayed below.
What is useful about the method applied is that we are able to leverage all information that is available. As more information becomes available, this can also be placed into the program providing us with the best estimate. Indeed, the more information the better.
Specifically, our estimates appear below (the one to the left was developed by Allan Stam and the one on the right by Christian Davenport and David Armstrong):
Specifically, our estimates appear below (the one to the left was developed by Allan Stam and the one on the right by Christian Davenport and David Armstrong):
4) The RPF did not stop the killing. Rather, they followed the termination of violence.
From our data collection (described above) and analyses (viewing each location one at a time) we observe that the Rwandan Patriotic Front advances into communes generally after the violence has ended. This is best seen in the following figures (developed by Prof. Stam):
From our data collection (described above) and analyses (viewing each location one at a time) we observe that the Rwandan Patriotic Front advances into communes generally after the violence has ended. This is best seen in the following figures (developed by Prof. Stam):
Location after location we see violence (in red) ending (some of which is most likely genocide but some other violence as well). Following this cessation of activities, the RPF moves in (in blue). If the RPF had been involved in stopping violence, then we should have seen RPF movements into the relevant area before the red had vanished. We provide selected figures of a few locations here but in the book manuscript we provide information for every locale and see a similar pattern. This has implications for not only how we understand the end of the violence (indeed we now need to search for what ended it) but it also has implications for how we evaluate a government whose identity and legitimacy is largely founded upon the claim that they ended genocide.
We offer an alternative figure below (also created by Prof. Stam) here the star represents when the RPF got into a location, locations are along the left, the magnitude of violence is along the right and time in the specific location moves from left to right:
We offer an alternative figure below (also created by Prof. Stam) here the star represents when the RPF got into a location, locations are along the left, the magnitude of violence is along the right and time in the specific location moves from left to right:
Another figure was tried but we did not think that it worked very well. Here, when the RPF started in a location is along the top. Each cell represents the location on the left. The magnitude of violence is along the right and time moves from left to right.