Interview with early warning forecasting expert Dr. Charles Butcher

Welcome to the first edition in a new podcast series from the Auschwitz Institute for Peace and Reconciliation. My name is Samuel Gillespie and today I will be discussing early warning forecasting, a new strategy in the field of genocide prevention. Joining me today is Charles Butcher PhD, a lecturer at the National Centre for Peace and Conflict Studies out of the University of Otago in New Zealand. As part of his focus on genocide forecasting and group participation in violent and nonviolent conflict, Dr. Butcher coauthored a report titled “Understanding and Forecasting Political Instability and Genocide for Early Warning” published in 2012. The report drew international attention after the forecasting model accurately predicted the mass atrocities in the Central African Republic. How are you today Charles?  

Well thanks, thank you for having me on.

Thank you for joining us. What sparked your interest in early warning forecasting?

Well I was very fortunate to have associate professor Ben Goldsmith as my doctoral supervisor at the University of Sydney. My PhD dissertation was on understanding the spatial dimensions of civil war violence, so in other words, why some places experience a lot of violence in civil wars, and others not so much. So I was familiar with the existing research on armed conflict and then at the end of my dissertation Ben brought me on board with the Atrocity Forecasting project with the University of Sydney, which also involved some colleagues at the University of New South Wales. At this point, the idea of even predicting mass atrocities was fairly new and interesting to me, so just the idea of prediction at that point brought me on board. The more I’ve worked on this over time, I’ve become increasingly convinced that the early warning of mass atrocities can have a real preventative impact and there have been a couple of recent studies that have solidified my opinion on this, by Matt Krain and Jacqueline Demara. They seem to show that naming and shaming of perpetrators and potential perpetrators can reduce the severity of civilian killings or prevent them from beginning. These might not be huge impacts, but small impacts really matter in these situations where we’re talking about thousands and thousands of people being killed and even a greater magnitude being displaced. So I thought that knowing where to look for instances of mass killing before they occur is a pretty obvious public good in this regard and given our tools, forecasting tools can be constructed with data that’s already publicly available and demonstrate a relatively high degree of accuracy, say in comparison to forecasting civil wars, makes early warning and public dissemination of those warnings a fairly low cost policy option with the potential to save lives. And there of course many other benefits.

And we’ll get into those. For those not familiar with early warning forecasting, what are two or three of the most important factors that are indicative of a genocide or a mass atrocity?

Okay so I’m just speaking from our own research and our own report here and your listeners should know that there are other projects out there that aim to forecast mass atrocities of different kinds and they emphasize slightly different aspects. This is the work of Barbara Halfand the political instability task force, also Jay Ulfelder and the Center for the Prevention of Genocide, I would encourage your listeners to take a look at these sites as well. In our own work, some of the factors that place countries at an elevated risk of genocide over a five-year period include the time since the last genocide episode, the presence of systematic discrimination of minority groups, the infant mortality rate as a broad indicator of the quality of governance and of poverty, and the number of neighboring states experiencing civil wars. I think I listed five of the most potent predictors in our model, but I would just add that these are the most powerful indicators in our model, such that when we take them out of our models, it has the largest negative impact on forecasting accuracy. There are actually 16 factors in the model that we use and they all do improve forecasting capability but they don’t all have quite the same impact. So a nation can be placed at risk by having risky values on their most potent predictors or also by having more moderately risky values across a number of dimensions.

While authoring your report, what countries exhibited signs of a nation that was vulnerable to a mass atrocity, based on these factors?

 The top five in our report, which was authored in 2012 and the forecasts were based on information from 2010, were the Central African Republic, Democratic Republic of Congo, Chad, Somalia and Angola. The Central African Republic ranked highly because of the ongoing civil war in 2010, a very high infant mortality rate and a large number of states around it also experiencing violent conflict. So in 2010, this was the DRC, Chad and the Sudan. Chad also ranked quite highly, it came in at third on our list but for slightly different reasons. It more moderately risky values across a lot of different categories, so a fairly large number of states, neighboring states, in violent conflict; recent militarization, so that means an increase in the number of people in the military proportional to the whole population; the occurrence of an election period and the presence of a peacekeeping mission, and this last one might seem a little bit strange and we’re at pains to point out that we don’t believe peacekeepers cause genocide but more that they tend to get deployed in places where the likelihood of large scale political violence and genocide are already high so they tend to get deployed to places that are very likely to experience serious political instability. More generally as well, not all the factors in our model also reflect causal connections to genocide. We think some do, but this is not necessary when the primary task is to forecast.

As an academic, what role do you believe academia in informing national governments and the international community about regions or countries that are vulnerable to crises? Because we do have great studies and great projects, such as your report, but how do we get that to the policy level?

 I think that academics have a couple of responsibilities in this regard. The first and very simply, academics should make the predictions that we make publicly available both for consumption by policy makers but also for the general public and advocacy organizations. I think this is especially the case given the capability of this type of information to raise the costs to governments and other sub-state actors of using mass atrocity even if its only in a very small way, I think that’s still an important responsibility. But going to couple with that, I think academics also have a responsibility to present the caveats and uncertainties for forecasting these kinds of events. Of course we’re not talking about forecasting the weather, but the actions of human agents that learn, they’re strategic and take into account what other people think they’re going to do and they think about what other people are going to do. And of course we have limitations related to the data we use and statistical modeling techniques, these all add additional uncertainty. So I think that’s a second responsibility, and really happily, I think the small group of people working on these problems do a really good job of these two things. In our own project, we’ve actively lobbied government to come and present some of our results, and we have done that. And I also know that in the U.S. there is a relatively close connection between people that are working on this kind of stuff and policy.

Do you believe there has been a recent, or in the last couple years, a shift in national governments to fund these sorts of projects and studies? Do you think that’s going to come or perhaps already started to come?

 I don’t know a lot about a move or a shift. I do know that the U.S. has funded some early warning efforts and I believe continue to do so to an extent. Our project was funded by, what was then, the Australian Aid Agency through a fund that was called the Responsibility to Protect fund, and we’re of course very, very grateful for that support. There’s still continued interest from universities in Australia and organizations with the prevention of genocide, such as your organization and also the Holocaust museum. So this research does get funded but I think like many things in academia, finding funding for research is a bit of an ongoing challenge so we have to apply for everything that we can. To my mind, it’s a little bit of a shame because I think a very small outlay of money can make significant improvements and I think if we consider some of the benefits that publicly available forecasts of mass atrocities can have, in relation to what I think are relatively low cost to produce them, this is quite effective. You could even consider it a public health intervention.

And my last question, you’d mentioned it earlier about modern technology playing a role in early warning forecasting, there’s a project called the Sentinel Project which I’m sure you’ve heard of, that uses satellite surveillance to monitor regions that are susceptible to genocide or mass atrocity. Do you believe that modern technology, such as the satellite surveillance and other sorts might improve the accuracy of the forecasting model?                                                                 

Actually this is one of the things we would like to do in the future is to combine our longer term forecasts, which are predictions over the next five years so in the report that you’re talking about its predictions between 2011-2015. We have to now update those forecasts very soon. We’re pretty familiar with the satellite Sentinel’s work and think it’s very valuable. Where we think we can combine these two types, combine forecasting and monitoring, is because while forecasts are for longer periods of time they don’t really capture those short term triggers, those structural trigger events that might lead to the onset of a genocide. That is exactly what projects like the satellite Sentinel Project do, but of course its really expensive, you can’t monitor all countries at all times in the world. So we think it would be quite useful to combine these efforts since we can provide where to look and then the more expensive and finer grain data that can be picked up with these monitoring efforts might be useful in that regard.

Well thank you very much for your time. We look forward to revisiting this issue with you at some point down the road. Thanks very much.

You’re welcome. Great to talk to you.

Leave a Reply