In 2012, while working on a humanitarian base in Haiti, I found a copy of Dr. James Orbinki’s book An Imperfect Offering on our community bookshelf. As a doctor, humanitarian, and a past president of MSF, he shares deep insights through the telling of often heartbreaking stories. This week I decided to revisit his book. While there are many pages that force anyone with empathy to give pause, one line stopped me and has stuck with me.
“The genocide in Rwanda was the most efficient genocide of the twentieth century...”
The word “efficient” stopped me. I use it all of the time. I read about, advise, and optimize for it all of the time. I have never used it as a descriptor in such a shameful context. It makes complete sense, however. One of the great tragedies of genocides is the pure “efficiency” of it all. Considering this, I began wondering if there is an opportunity to reduce the brutal efficiency.
I have summarized my thoughts here and shared my full process below. I would be grateful for feedback and discourse. My quick exploration went roughly as follows:
- UN definition of genocide
- A “back of the envelope” calculation of how “efficient” the world’s modern genocides had been
- Identify the steps of genocide
- Identify opportunities to interfere
- Use two “case studies” or “vignettes” to:
- consider a process for intervening
- estimate the possible impact
- consider the cost of an intervention
- Note the weaknesses of this post
My current thoughts following this < 50-hour exploration of the topic:
- The shortest events tend to be the most “efficient.” With two exceptions, the top ten most efficient genocidal events lasted less than 9 months. I speculate that these condensed events likely have the highest ROI potential for any intervention. They also likely suffer from the slow bureaucracy of international agencies.
- Of the entities attempting to identify the causes and stages of genocides, I found Dr. Gregory Stanton’s "Ten Stages of Genocide" to capture the consensus and be the most distinctly identifiable. (His organization also has a genocide watch and warning function.)
- During the Rwandan genocide, the US military passed on the opportunity to save lives at the cost of ~$40,800 per life saved (completely ignoring all of the DALY’s, spillover effects, and other suffering). That is ~2.25x the “Cost-effectiveness of counterfactual government health spending” estimate used by GiveWell.
- Speculative opportunities for civilian intervention:
- Communication Interference:
- In a future situation of increased immediate violence incited and organized by a dominant media outlet (similar to the Rwandan genocide), communications interference could prove an effective and efficient intervention. In the age of social media and cell phones, the costs could be lower and the opportunities to intervene could be greater.
- Likely illegal (though not as illegal as genocide, or even a single murder), it could be possible to jam a radio signal with limited technology. (And at a lower cost than the 1994 rate of $8500 per flight hour for a US military signal-jamming plane.)
- Another option would be for a DMI-style emergency response arm to offer a counter-voice. (Cited as an option by the US military during the Rwandan genocide.)
- Facebook’s role in the Myanmar Genocide of the country’s mostly Muslim Rohingya minority group instigated considerations into the platform’s role in shutting down violent and targeted rhetoric (credit to Aaron Gertler for sharing this insight). Is there a role for a ML/AI system to improve the detection of the fake accounts used by Myanmar? Is there value to an organized troll/disinformation scheme to disrupt their attempts to organize and instigate on the platform? From an operations perspective, it could be an impactful career to work with these platforms to have a playbook in place to respond to these events.
- Activist Investing:
- While the likelihood of having an impact is uncertain, there may be opportunities to use infrastructure interests (aligned interests) to lobby state actors to either intervene or reduce the conflict.
- The cost of the intervention is tied to the investment. So it could be negative (positive economic outcome) or up to the value invested (company fails). With relatively “safe” investments in well-established companies, the risk could be mitigated.
- Communication Interference:
If you are interested in further exploring my full thought process, I have shared it below:
(1) Defining Genocide:
Genocide is defined as...
“... any of the following acts committed with intent to destroy, in whole or in part, a national, ethnical, racial or religious group, as such:
(a) Killing members of the group;
(b) Causing serious bodily or mental harm to members of the group;
(c) Deliberately inflicting on the group conditions of life calculated to bring about its physical destruction in whole or in part;
(d) Imposing measures intended to prevent births within the group;
(e) Forcibly transferring children of the group to another group.”
— Convention on the Prevention and Punishment of the Crime of Genocide, Article 2
(2) Estimating Efficiency
I’ve created a Google Spreadsheet to create a rough calculation of the efficiency of genocides, as measured by the number of Individuals Killed Per Day. My calculation should be considered a very rough “back of the envelope” calculation. I started with the 70 entries found on Wikipedia’s List of genocides by death toll. I reduced the data set to 48 entries by removing incomplete entries and all entries before 1894. My intention is that this trimming will reduce problematic errors and improve the modern contextual relevance, all while keeping the data set as large as reasonable.
Process for calculations:
- Estimate the total duration of the genocides in days (end day included)
- Divided the low and high estimates of fatalities by the number of days
- Calculate a simple mean: (L+H)/2
- Reorder by mean
Once ordered by the mean, by my rough calculations, it appears that Dr. Orbiniski was mistaken. Tragically, four other genocidal events could be categorized as more “efficient” killing occurrences. Worth noting though, one of those events was inextricably tied to the events in Rwanda.
(3) Stages of a Genocide
I then moved on to what factors can identify a genocidal event. I found the UN and other organizations to offer unproductively vague identification factors. In my personal opinion, Dr. Gregory Stanton: "Ten Stages of Genocide" to be the most identifiable.
It is worth going on a brief tangent to highlight his tools, “Genocide Watch,” an interactive international map that identifies countries engaged in the Ten Stages, and Genocide Alerts, which does what it sounds like it does.
His ten stages include:
Many organizations work to reduce the tensions (stages 1-6) that can spark genocidal events. There are organizations that lobby to recognize events as genocide and ensure that they are taught historically as such. I have not found, in my limited search, any organization beyond the UN that provides operational responses to the Persecution and Extermination stages.
(4) Opportunities to Intervene (vignettes)
Rwanda: Radio Disruption
When considering interventions, I was immediately curious about the Rwandan Genocide. The violence was primarily limited to 100 days and the radio was a significant coordinator and instigator. Throughout the conflict, there were international calls to stop the broadcasts. So influential were the radio broadcasts, that in 2003, three “media leaders” were convicted of “ genocide, incitement to genocide, conspiracy, and crimes against humanity, extermination, and persecution” by the UN International Criminal Tribunal. I was even able to find a 2014 Harvard study that sought to quantify the effects of the RTLM radio broadcasts on the violence. It states that “approximately 10% of the overall violence, can be attributed to the station.” It also estimates that “approximately 6.5 percent (28,000 persons) of the individual violence and 29 percent (22,000 persons) of the collective violence can be attributed to the broadcasts.”(2012 Working Paper PDF)
One research article casts doubt on the proxy used in the Harvard study.
Beyond spending months cultivating toxic and divisive rhetoric that contributed to the culture of violence, the radio station took direct steps in organizing the killings (i.e. - broadcasting names and addresses of individuals). If that result is true, and (falsely) assuming a uniform distribution of violence, then 50,000 - 107,100 lives could have been spared if the radio had been prevented from broadcasting. In Foreign Affairs, Alan Kuperman proposed that President Clinton "could not have known that a nationwide genocide was underway" until about two weeks (14%) into the killing. An action then still could have had considerable impact.
There were multiple reasons that the Americans failed to act, but concerning the radio station, it may be easiest to cite The Atlantic analysis directly;
“The country best equipped to prevent the genocide planners from broadcasting murderous instructions directly to the population was the United States. Marley offered three possibilities. The United States could destroy the antenna. It could transmit "counter-broadcasts" urging perpetrators to stop the genocide. Or it could jam the hate radio station's broadcasts. This could have been done from an airborne platform such as the Air Force's Commando Solo airplane. Anthony Lake raised the matter with Secretary of Defense William Perry at the end of April. Pentagon officials considered all the proposals non-starters. On May 5 Frank Wisner, the undersecretary of defense for policy, prepared a memo for Sandy Berger, then the deputy national-security adviser. Wisner's memo testifies to the unwillingness of the U.S. government to make even financial sacrifices to diminish the killing.
“We have looked at options to stop the broadcasts within the Pentagon, discussed them interagency and concluded jamming is an ineffective and expensive mechanism that will not accomplish the objective the NSC Advisor seeks.
International legal conventions complicate airborne or ground based jamming and the mountainous terrain reduces the effectiveness of either option. Commando Solo, an Air National Guard asset, is the only suitable DOD jamming platform. It costs approximately $8500 per flight hour and requires a semi-secure area of operations due to its vulnerability and limited self-protection.
I believe it would be wiser to use air to assist in Rwanda in the [food] relief effort ...”
The plane would have needed to remain in Rwandan airspace while it waited for radio transmissions to begin. "First we would have had to figure out whether it made sense to use Commando Solo," Wisner recalls. "Then we had to get it from where it was already and be sure it could be moved. Then we would have needed flight clearance from all the countries nearby. And then we would need the political go-ahead. By the time we got all this, weeks would have passed. And it was not going to solve the fundamental problem...”
Just working with the $8500 per flight hour, that number works out to be $204,000 per day, assuming the radio station played non-stop. Given that the conflict lasted 100 days, and 6.5% of the individual violence (ignoring spill-over effects) could be attributed to the broadcasts, if you could disrupt 1 day (1%) of transmissions, you could plausibly reduce 6.5% of the individual violence by 1%. Using the mean, that equals ~5 lives saved per day of interrupted transmission. Saving 5 lives for $205,000 works out to $40,800 per life saved (~10x the current cost of life saved through the Helen Keller Initiative ($4,016) per GiveWell estimates).
(In response to radio jamming being brought up yet again at a meeting, one Pentagon official responded; "Pru, radios don't kill people. People kill people!")
There are certainly weaknesses and limitations to this thought exercise. For one, the killings were not uniformly distributed. Secondly, the radio station began broadcasting in July 1993 and the serious killing did not begin until April 2004, allowing the station to “plant seeds” of divisive rhetoric, especially among young men. Additionally, it is unclear what response (i.e. backlash), if any, there would have been to attempts to stop the broadcasts.
Somalia: Investment & Corporate Lobbying
Between May 1988 and March 1989 there was a state-sponsored genocide of 50,000-100,000+ Isaaqs, a clan in the north. Notably, there was considerable American influence in the area. Throughout the 1980s US oil companies were finding deposits in Somalia. While they remained undeveloped the US established naval and military bases in the country, overseeing the Persian Gulf and the Red Sea. The Times reported that around two-thirds of Somalia was allocated for oil exploration, specifically by four US oil companies (Conoco, Amoco, Chevron, and Phillips) In 1988, the GDP per capita of Somalia was ~$148 (current US$ WB). The investment by the oil companies and the US military was a significant financial support to the country.
In August 1988, Amnesty International began running stories of the killings that would turn into a genocide. At the time (August 15, 1988), an investor could have purchased shares in Chevron for $11.25. Over the remainder of the genocide, the stock had an annual return rate of ~50%. An investor (or group of investors) could have purchased enough shares to advocate for the company to speak out to the relevant parties (investment activism). Further exploration of the strategy and efficacy of investment activism would be highly valuable in such a strategy. It is likely highly uncertain what effect, if any, this would have had. Would the oil companies have been influenced by activist investors (to complain/protest/call on the US/divest from the region)? Would the applicable government(s) have been influenced by any action taken by the oil companies? Even if the impact was effectively zero, the financial costs would effectively be negative (a positive economic outcome). This case study looks at what holding the investment would look like 25 years later (though I am confused about the cited share price).
(A more modern context: Aaron Gertler referred me to Facebook’s role in the Myanmar genocide. Since October 09, 2016, when the genocide is cited to have begun, Facebook’s stock has had an ARR of 22.19% and a $10,000 investment would have more than doubled to $22,625.70. On Nov. 08, Aung San Suu Kyi's ruling party won re-election.)
Weaknesses of this post:
- The killings were not uniformly distributed during the windows analyzed. It is possible that smaller windows would have considerably higher “efficiency” rates.
- My curating of the dates and numbers was based on quick research with little follow-up. There are likely flaws.
- I have not yet made a calculation that considers the proportion of the targeted population.
- My considerations completely neglect the impact on those who survived with injury, trauma, or other loss.
- The mean calculation is unweighted.
- I am not fully confident in my calculation of the value of the investment considered.
- My 1% calculation is certainly over-simplified to provide a perspective in a complicated situation.
[I would like to thank Aaron Gertler for his encouragement at EA conferences to write on the Forum, including taking the time to host workshops, share resources, and offer his personal assistance. It is likely that I would not have written this, my first post, without his nudging (at least not in the same time frame). Your time and effort influenced behavior change. Also, thanks to the kind folks over on the FB Effective Altruism Editing and Review page who shared their thoughts (specifically those who messaged me privately).]
Hi, Nice post. I found this article and it had some criticism of the effect of the radio broadcasts.
It will be interesting to estimate the effect of genocide on future outcomes, like economic growth or political stability. For example, Somaliland, where the Isaaqs were killed, is currently more stable than other parts of Somalia, but that could be due to factors completely orthogonal to the genocide.
EDIT: I tried to find studies on the effects of genocide on the economy and a brief glance makes it seem that it's very bad. I also found an interesting report on Somaliland. It looks like the situation in Somalia was worse than in Somaliland, which might explain their differences, but I'm very uncertain and I haven't finished reading the report.
Interesting work on a very important topic, good job. I was especially surprised to see that it took two weeks for the US to learn about the genocide; surely the US ambassador should have noticed?
I think you are a little harsh on the US decision not to use the radio blocking technology. It sounds like money wasn't their only (main?) objection: it was also logistically difficult to use the radio blocking plane, and require a substantial escort. Perhaps keeping it safe might even have required destroying Rwandan SAMs:
Looking at your spreadsheet, it seems the Rwandan genocide in some ways represents a best case scenario for intervention, as it was implemented in a somewhat decentralized way with civilians in a third world country. Many of the other genocides occurred under the direct orders of more powerful states, which would prevent such interventions from working - even if the US could have blocked German radio in 1943, for example, the holocaust would have continued.
Some other techniques that might be useful:
Thanks for reading my post and sharing your thoughts.
With regard to any criticism you may have derived from the post, I am assuming you are referring to my choice of the phrase “failed to act.” As an American, I expect my elected representatives to intervene in genocides or other human tragedies. I understand that there are rational limitations to this opinion and that other voters do not agree. (I also have no interest in sparking a political debate here.) I hold no judgment around the process of how to intervene, as it is well beyond my expertise. I do believe that any genocide or event of mass suffering is caused, at least in some part, by a failure on the part of those with the ability to act.
In regards to the plane, it very well could have been a poor idea for many reasons. I found it interesting because it was a number offered by a credible source that allowed for a calculation.
I agree that Rwanda is certainly an obvious case for intervention. As you mentioned, there would likely have been limited international risk from a global power (save France). It’s recency also allows some benefits from perspective and analysis; familiarity with the context, news coverage, recent UN trials, etc.
I came to the same conclusion the protracted conflict with world powers (i.e. the Holocaust) would certainly be more complicated to intervene in.
Thoughts on your other possible techniques:
Sorry, I didn't mean to suggest the US shouldn't have intervened - I think quite possibly we should have! I just meant the costs would likely have been higher than you estimated, because it's not just the per-hour cost of the radio jamming plane. Political capital with neighbors is costly, and protecting the plane could have been quite expensive. Wikipedia suggests Rwanda had some (old) Russian fighter jets, so they might have needed to be shot down, and they may also have had SAMs which would require neutralization.
Yeah, I was thinking about things like the role of civilian firearms as a defence against lynching in the US south, where they seem to have been somewhat effective:
But it's not exactly the same case because lynching is quite different from genocide, and the total number killed was quite small - probably under 5,000 over many decades.
Perhaps a more similar case was the decision by the Albanian government to arm the northern civilian population to help protect them from the south:
Again, this is not a perfect example, because we don't know what would have happened if they had not been armed.
We do know that many historical genocides were preceded by the disarming of the victims. For example, prior to the Armenian Genocide:
Similarly, prior to the Soviet genocides:
Similarly, Weimar Germany had relatively strict regulation of firearms, and the Nazis banned Jewish firearm ownership prior to the holocaust.
Of course, once a government has decided to disarm a population, presumably they would not be willing to allow outsiders to re-arm that population. So it might be more effective to educate at-risk groups about how to conceal firearms and avoid confiscation.
I think you raise a good point about governments arming groups that they later go on to fight - the US arming the mujahideen in Afghanistan is a classic example. But my impression is that these cases generally involve the supply of anti-tank weapons, anti-air weapons, and other pieces of relatively heavy-duty equipment. If you aim is to simply make genocide more difficult, small arms are likely sufficient. The Rwandan genocide, for example, made widespread use of machetes to murder victims - ownership of even relatively small caliber weapons, common among ordinary civilians in the US, could have likely prevented much of this.
You could (and likely are) correct that the US military intervening would have cost (in cash and political capital) more than the number ($8500) cited by Frank Wisner, the undersecretary of defense for policy. I have no idea what all is covered by his sum, but it does not seem a stretch to me that it was not all-inclusive.
Interesting citations regarding firearm ownership. Thanks for sharing. I would not have thought of that one myself. You also made a good point about the difference between military-grade armament and small arms possession.
I think there are a few complicating dynamics:
Types of Injury
There were certainly guns present and used during the Rwandan Genocide, especially by the military/militia groups. That said, the distribution and use of machetes was pervasive and brutal. Dr. Orbinski observes that the Interahamwe utilized machetes for violent acts that would have been near impossible for a gun to inflict. What would a greater prevalence of guns have done to the types of injuries inflicted?
Unfortunately, the Rwandan Genocide was not the first violent tension between the Tutsi and Hutu populations. So it seems unclear to me what effect arming the Tutsi minority would have had. This is also one difference from the African American population. In Rwanda, the Tutsi’s were traditionally the empowered population.
Access (guns or otherwise) requires infrastructure. Infrastructure takes time and coordination. Once the violence started, the bulk of it only lasted 100 days. So any promotion of ownership would have had to have happed in advance. What tensions would that have caused?
Long-run spill-over effects
Beyond the historical tensions, there would likely be effects of increased gun ownership. How would more prevalent gun ownership affect suicide rates, domestic violence, accidental injury, school violence, etc.?
At the risk of being unproductively vague: Who would develop the markets for gun ownership? The US? The private market? Who regulates it? How does regulation affect corruption? Where is the line between self-defense and violence as preventative action?
I appreciate you keeping the conversation going. You have raised multiple points I would have missed.