Part V Genocide Denial – D: Statistical Analysis

Part V Genocide Denial – D: Statistical Analysis


Howard Adelman


The BBC video is available at


The most important effort of Allan Stam and Christian Davenport (S&D) was not spent on conceptualization or on history, but on their statistical analysis. In that effort, the most significant part was not the aggregation of numbers to establish that just over one million people were killed in Rwanda – a figure I for one accept and which is not very controversial – but the disaggregation of that total to make the following claims:

  1. Far more Hutu than Tutsi were killed;
  2. The number of Tutsi killed was 200,000;
  3. Only half that number were killed with the intent of extermination.

This revisionist thesis of the proportion of Hutu and Tutsi killed has to be suspect. It would mean that the three mass graves that we investigated of over seventy identified at the time with thousands buried (see the 1995 Lansat map below – less than 5% of the sites of mass graves) held 33% of the Tutsi victims, a preposterous assumption. The absurdity can be put another way. Those three graves held two-thirds of the S&D proposed 100,000 Tutsi murdered for genocidal reasons. And we only examined areas that had recently been excavated to assess accuracy of counting, not totals; more were uncovered later. We had no doubts that virtually all those killed in those three sites had been Tutsi who had been forced together in churches and schools and systematically slaughtered over a period of days.

This is a 1995 Landsat TM mosaic for the country of Rwanda after the genocide.

The national border in the Lansat map is shown in white. Genocide sites of mass graves (“lieus publics“) are shown in blue. Memorials (“lieux de culte“) are shown in red. Resistance sites (“collines de résistance“) are depicted in green.

The Rwandan government’s documentation centre tried to identify as many of the victims as possible using photos, audio-visual interviews, and other official records based on methods pioneered in the Israeli memorial in Jerusalem at Yad Vashem. We never examined the vast majority of sites. In the BBC documentary, Murambi is identified as a site where 15,000 Tutsi were killed. The official numbers are actually 45,000 killed there and 20,000 killed trying to escape. Nor did we examine sites at Nymata, Ntarama, Bisesero, Nyanza and Nyarabuye. We did not go through the Aegis Trust interviews, but only read early summaries of a small number of individual interviews. Further, 85% of graves were classified as mass graves if they only held 20 people, while we focused on the 15% of the 466 sites where corpses were calculated in the hundreds and thousands.

Our excuse in not being as thorough as we would have liked is because that was not the focus of our study. Yet we seem to have paid attention to sources that S&D did not, though they claim in their study that they had access to a wide variety of documentation in Rwanda. Further, they had access to much more sophisticated sources that were developed subsequently to our study. The surprise is that they did not use these to attempt to falsify their own presumptions.

In their methodology employing event cataloguing, in which they coded the events during the 100 days of massacres using indicators such as time, place, perpetrators, victims, the weapons used and the particular form of military action, and correlated them to discover a pattern, they claimed to uncover an original interpretation – that the violence did not take place all at once but took place in a sequence, though they did not initially make any claims about the pattern of that sequence. The truth is that every single scholar who studied the genocide has said that the killing did not take place all at once. Alison Des Forges in the first large scale study documented that sequence and explained it in part mostly because of the politics of place. For example, the prefecture in certain places like Butare refused to go along with the genocidal campaign of the new military government run by extremists. The prefect was eventually dismissed from his post and the genocide subsequently went forth in that area.

The two academics argued (not in the BBC program but in their scholarship) that the vast majority of the population had been on the move in 1993 and 1994. It is true that, as we and many other had documented, large numbers of Hutu in territories conquered by the RPF fled the advancing military force, believing the extremist Hutu propaganda that they would be slaughtered. Others were on the move because they were forced by the killing machine of the militias into schools and churches to be slaughtered. The number S&D offer is 450,000 to 750,000. This too was not a new revelation, only the twist S&D gave it that there was at least as many Hutu and Tutsi in these mass graves. S&D created maps that showed that, while killing took place in different parts of the country, it did so at different rates and magnitudes. Again, this was not a new discovery, though using dynamic and layered mapping to show it was helpful. But as I have shown, more static maps without the temporal dimension had previously been created.

The conclusion that most of the over a million killed in the genocide were Hutu is contentious at the very least. The calculation is not based on any dissection of those killed to determine who was a Hutu, who was a Tutsi and who was of mixed background, or who was Twa for that matter. It was primarily based on the argument that the number of Tutsi in Rwanda at the beginning of the war was only 500,000. S&D use a figure of over 300,000 Tutsi who survived the war. (Gérard Prunier claimed the figure to have been 130,000, though others claimed 150,000), Therefore, according to S&D, only 200,000 were killed. The other over 800,000 had to have been Hutu.

According to S&D, the number of Tutsi in the country prior to the war was only 500,000. Tutsi made up only 6.5% of the population, whereas demographer William Seltzer using the same faulty 1991 census data, arrived at a figure of 657,000 or 8.4% of the population, again versus a large number of scholars critical of the 1991 census who focus on the “hidden” or self-disguised Tutsi, and estimated that Tutsi constituted 12-14% of the population. The total pre-genocide population in 1990 was said to be over 7.12 million in 1990 and by 1994 before the genocide was said to have reached 7.6-8 million. Critics assert that the number of Tutsi was underreported in that census and in the prior census of 1978 because the Habyarimana government wanted to minimize the importance of Tutsi in the population.

Whether or not census data were purposely altered to reduce the number of Tutsi, the figures underestimated the Tutsi population because an undetermined number of Tutsi arranged to register as Hutu in order to avoid discrimination and harassment. Although many Rwandans know of such cases, there is at present no basis for estimating how many Tutsi were counted as Hutu because of this. Nevertheless, the 1991 data show Tutsi as forming 8.4 percent of the total population, not 6.4%. This figure seems to accord with extrapolations from the generally accepted census data of 1952, taking into account the population loss due to death and flight during the 1960s and the birth rate, which was lower for Tutsi than for  .

If the 14% figure is used out of a total of 7.8 million, then the pre-genocide Tutsi population was almost 1,100,000. If the 12% figure is used out of a total of 7.8 million, then the pre-genocide Tutsi population was 936,000. If the 8.4% of the 7.8 million population is used, then the pre-genocide population was over 650,000. If the number of Tutsi survivors is 150,000, then the number of Tutsi killed would be 950,000 based on the pre-genocide figure of 14%, 786,000 based on the pre-genocide figure of 12%, and 500,000 based on the pre-genocide figure of 6.4%. This excludes those of mixed parentage who constituted a significant part of the Rwandan population who were also targeted by the genocidaires. S&D offer an even lower percentage of the population of Tutsi in Rwanda before the genocide and a much higher number of survivors. They justify the latter by misrepresenting their source, which used the figure of 300,000 to refer to both Hutu moderates who survived and Tutsi. Combining the misrepresentations of both the original Tutsi population (diminished) and the number of Tutsi survivors (enhanced), leads to their conclusion that only 200,000 Tutsi died. Of these they guess – not estimate – that Hutu were targeted as Tutsi and that, of the 200,000 Tutsi, only half were targeted.

S&D, in spite of their insistence on being evidence-based and their stress on quantitative analysis based on data, ignored most of the caveats and guidelines in the William Seltzer and Margo Anderson 2001 classic paper, “The Dark Side of Numbers: The Role of Population Data Systems in Human Rights Abuses”. More specifically, the use of the 1991 census is usually accompanied by a disqualifier about its unreliability vis-à-vis the distribution of the population between Hutu and Tutsi.

One fundamental error is that S&D use as their base line the 1991 Rwanda census. Even though many demographers use the 1991 census, the vast majority, including William Seltzer, consider the 1991 census as unreliable, especially with respect to the number of Tutsi. The government of Habyarimana had a vested interest in minimizing the count of the numbers of Tutsis in the country to offset the accusations against him by the extremists that he was soft on the fifth column within the country. The Tutsi themselves had a vested interest in disguising their identity since anti-Tutsi extremism was already mushrooming in Rwanda in 1991 following the RPF invasion. As Alison Des Forges wrote, “Whether or not census data were purposely altered to reduce the number of Tutsi, the figures underestimated the Tutsi population because an undetermined number of Tutsi arranged to register as Hutu in order to avoid discrimination and harassment.”

In contrast, S&D claim that they used other sources to confirm the accuracy of the 1991 base line as their foundation. “(W)e took information from before the questionable census of 1991 and projected forward diverse population growth rates (which is standard practice in demography) and found figures that were comparable to what was discussed in 1991 therefore allowing us to use it in our estimations.” They then compounded their error about the ethnic disaggregation of the pre-genocide data with misrepresenting their sources as referring to Tutsi survivors, whereas the definition included Hutu political survivors. To be charitable and unwilling to claim bad faith, the two were so busy counting trees that they missed seeing the forest, an inversion of a more typical error.

Further, did the RPF kill up to a 150,000 Hutu within Rwanda? Or did the RPF’s aggressive pincer- like military advance propel the FAR into killing as many Tutsi as possible before they were forced to withdraw. And, as S&D so correctly point out, the RPF had a singular focus on winning the war, not on saving Rwandan Tutsi. As so many scholars have concluded, the genocidaires made their priority killing Tutsi, not winning the war.

However, the two American scholars had a different message; they concluded that a significant proportion of the killings took place in RPF controlled territory, a much more expansive thesis than simply arguing that pogroms against the Tutsi were provoked by the RPF, initially by their invasion in 1990 and then by the flight, forced displacement or ethnic cleansing – depending on how you interpret the Hutu population movement in the territory it captured. Kuperman’s conclusions may at first appear to be similar to S&D’s since S&D imitate and expand upon his methods. However, S&D state that, “The killings in the zone controlled by the FAR seemed to escalate as the RPF moved into the country and acquired more territory. When the RPF advanced, large-scale killings escalated. When the RPF stopped, large-scale killings largely decreased.” However, in Kuperman’s book and articles, he noted that some of the initial massacres took place in southwest Rwanda furthest from the RPF front line. Yet S&D claim that the data revealed in their maps was consistent with FAR claims that it would have stopped much of the killing if the RPF had simply called a halt to its invasion. The illogic of such a conclusion is astounding!

The data indicates no such thing. It is an inference. Many others could be made using the same data. If I drink four glasses of gin on ice one day and get drunk, then drink four glasses of rum on ice the second day and get drunk, then four glasses of scotch on the rocks on the third day and get drunk, then four glasses of rye whiskey on ice on the fourth day and get drunk, can I conclude that the it was the ice that made me drunk? Further, even if one assumes that the RPF advances instigated an increase in the pace and amount of killing, this does not mean that the RPF was responsible for the increased numbers killed. The correlation could have been the result of FAR trying to finish its work before fleeing.

S&D had a second string in their quiver echoed in the BBC documentary. Their thesis was a counter to the regime’s contention that the civil war was continued to stop the genocide. However, every decent scholar that I have read knows and writes that the RPF was single-mindedly focused on defeating the FAR and the Interahamwe militias. Stopping the genocide was a by-product of that success. The faster defeat came, the quicker the slaughters would stop. The strategy of using a pincer movement against a stronger, more numerous and better equipped military foe would almost certainly have imploded if the RPF diverted their energies to countering the genocide directly. The RPF had a much stronger explanation for not dispersing their troops to stop the genocide, quite aside from Kagame’s willingness to sacrifice Tutsi lives for his goals. The Kagame regime may now say that the invasion was intended to stop the genocide, but that was one of the overall goals and not the intent of specific operations. The RPF victory did bring an end to the genocide even if that was neither a prime original intention nor a continuing immediate objective.

Though S&D – and many other scholars – hold Paul Kagame to be guilty of war crimes and crimes against humanity, they do not deny that a genocide took place. They do not deny the enormity of the genocide targeting Tutsi. Unfortunately, accusations against Kagame and genocide denial often go together and the two charges are often compressed into a single claim. S&D do not argue that the genocide did not take place, merely that it was only part of the story. “We have never denied that a genocide took place; we just noted that genocide was only one among several forms of violence that occurred at the time.”

It certainly was. But in their own calculations, mass graves consisted of 400,000 to 750,000 corpses, with most scholars leaning towards the upper end. To get around this problem, S&D claim that half the bodies in these mass graves were Hutu, but they provide no evidence for their claim except that this would have to follow from their other false premises instead of falsifying the conclusions based on those false premises. However, such a claim is so inconsistent with direct observations and testimony of survivors that it is a wonder that they can hold their heads up and make such a claim.

Thus, though they claim that their data is “only as precise as their background data,” their inattention to that background data is grossly negligent. Then they admit that, “Sometimes relative accuracy…is better than absolute accuracy.” Their final estimate of the Tutsi population in Rwanda in 1990 is based on an educated guess, not a reasonable calculus, and the estimated percentage of that population killed is an outright guess.

Further, if their estimate of the total number killed is over 1,000,000, and if the totals of the disaggregated ranges range from 625,000 to 1,350,000, it would be extremely difficult to arrive at a figure of 100,000 Tutsi killed. The calculation would be as follows:

S&D                             HA

Range                          Mid-point     Hutu   Tutsi

Those individually named                               25,000 – 100,000          62,500     50,000 12,500

Those killed at roadblocks                               25,000 – 100,000          62,500     31,250  31,250

Mass killings                                                     400,000 – 750,000         575,000    57,500 517,500

Local violence                                                    25,000 – 100,000          62,500       31,250  31,250

RPF violence (an admitted total guess)       100,000 – 150,000       125,000    125,000        0

Totals                                                               575,000 – 1,200,000       887,500     245,000 580,000

The large spread in the ranges already make the figures very suspect. Guessing those killed by the RPF with no foundation is a statistical prohibition. The first estimate of numbers slain by the RPF was made by Gersony in his supposedly suppressed 1994 report. He concluded that the RPF killed between 25,000 and 45,000, not 100,000 to 150,000. Seth Sendashonga, former minister of the interior and an early member of the RPF, estimated that the RPF in Rwanda killed some 60,000 people between April 1994 and August 1995, with more than half killed in the first four months of that period. He never suggested that many were killed under the earlier occupation. Further, these much lower estimates could include persons killed in the course of combat, both civilians and militia.

The total figures cannot be reconciled. Analyzing any suggests a different distribution. Named lists consisted mostly of moderate Hutu rather than Tutsi. On the other hand, S&D’s suggestion that, since the killers in the Interahamwe largely could not read the ID slips of paper, since it was often dark at roadblocks, therefore the proportion of those killed at those roadblocks should be taken roughly as the Tutsi and Hutu proportion of the population or, at most, equal numbers of each. This is pathetic reasoning. However, for the sake of argument, use the equal figure. On the other hand, the vast majority of those murdered en masse were Tutsi – I presume above 90%. Assume equal numbers were killed in local violence, again an assumption that veers strongly in favour of enhancing Hutu totals. There is no way one can arrive at a figure that a majority of those killed were Hutu. The calculations are totally misleading. Even in S&D’s best case, almost two-thirds had to have been Tutsi. And if the grand total is upped to one million, at least 630,000 Tutsi were killed in the genocide.

However, this is a mugs game that ends up mostly with guesses, perhaps some educated ones, when counting bodies in mass graves is probably the best indicator of the number of Tutsi killed since evidence indicated the vast majority of these were Tutsi. And that yields an even higher total than 630,000.

How do S&D reduce their total to 100,000 Tutsi killed through a deliberate genocide? They argue that those killed were a mixed category of all other types of random killing, revenge killing, mistaken killing, etc., but do not disclose anywhere that I could find the basis for such reasoning. Further, the counts of the mass graves, a source they seem to totally dilute with the use of those other methods, belie such a conclusion. They never seem to question the reliability of witness testimony, the results from interviews and focus groups, which they used as an important method in their calculations. How many Germans after WWII admitted to participating directly or indirectly in the slaughter of the Jews? Or even admitted they knew of the death camps? More importantly, look at how many said that most of those killed were results of war and violence and not deliberate efforts at extermination. If the “evaluation of killing designated by territorial control” concludes that the vast majority of those killed resided in FAR controlled territory and not in either the battle zone or the RPF-controlled areas, then the real issue is disaggregating that figure. The authors have no methodology that can do that with any degree of rigour.

If “we cannot be clear on how much of the violence (i.e., what proportion of all acts considered were legally classified as genocide), and that is because making such a judgment requires extremely detailed information about the victims, the perpetrators and their motives, motives, naturally being the hardest to document as they reside within an individual’s mind,” then presumably that should make us very cautious about reaching a number like 100,000, especially when the authors have used the very narrowest definition of genocide, one that excludes actions by non-state actors, such as so-called voluntary militias under the control of extremists, and especially when the authors ignore most other sources of information about the perpetrators and their motives.

Their reasoning is just fallacious, quite aside for the myriad of unsupported assertions. Most scholars refuse to accept the questionable 1991 census as a base line. The large team we worked with uniformly rejected such a presumption. Further, S&D conflate the criticism of the 1991 census data with the argument that this data was needed to enumerate and/or identify targeted Tutsi victims. As they say elsewhere in their work, the identification of the Tutsi targeted for killing was carried out by locals based on intimate local knowledge of who was a Tutsi and who was a Hutu. This had nothing to do with the issue of large numbers of refugees on the move. Those refugees were overwhelmingly Hutu fleeing areas conquered by the RPF.

Deliberate misrepresentation of ethnicity complicates assessing how many of the victims were actually Tutsi. At a reburial ceremony for a family slain during the genocide, the only two survivors, both priests, talked separately to S&D’s researchers. One maintained that his family was Tutsi, but claimed to be Hutu looked like Tutsi, while the other declared that the family was really Hutu, but was said to be Tutsi by neighbors who coveted their wealth. All this can be admitted without the need to deform the overall proportions so as to minimize the number of Tutsi in the population prior to the genocide and maximize the number of survivors.

Does this gross misuse of statistics and numbers constitute genocide denial?






One comment on “Part V Genocide Denial – D: Statistical Analysis

  1. Alex Zisman says:

    Howard, I am starting to get a bit confused with the sequence of parts in Roman numbers and the letters that follow. How do you plan to assemble everything? Alex

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s