AOAV: all our reports

Thoughts on violence: Michael Spagat, University of London


Michael Spagat is Professor of Economics and Head of the Department at the Royal Holloway, University of London

Professor Michael Spagat is the Head of the Economics Department at the Royal Holloway, University of London. He holds a PhD in Economics from Harvard and a BA in Economics and Mathematical Methods in the Social Sciences from Northwestern.  He is also a research fellow of the CEPR, the Davidson Institute and the Households in Conflict Network.

AOAV: What is your main academic focus of interest? 

Michael Spagat: My specialism is armed conflict. Originally I did my PhD on the soviet economy and then also worked on the transition from communism but that was a topic that was ending – so I eventually transitioned to armed conflict.

AOAV: And, within armed conflict, what are your main areas of interest?

MS: One would be the dynamics of conflict. A lot of people, especially economists, focus on the causes of conflicts. They just look at moments in which conflict breaks out and what factors are causing conflicts to break out.

I am more interested in how conflict evolves over time. My prime, most successful, work has been done with an interdisciplinary team – consisting mostly of physicists.  We looked at event data for a bunch of different conflicts. By event I’m referring to a discrete incident like, say, a suicide bomb or battle, where we looked at the number of people killed in the event.

In some cases we had data on injuries as well, so sometimes the size could be the killings and supposed injuries, but usually it’s just killings. We also looked at the timing of the events and how they spaced out over time. We found some remarkable regularities in size distribution and timing.

We were rather shocked to discover that the conflict we looked at had a very regular, “well-behaved” mathematically size distribution of the event.

The larger the size of the event the rarer they are. It is always the smaller ones that are most common.

But the issue was to see how quickly the frequency of the event drops off as you move to the larger sizes. You might imagine that with the drop in, let’s say, a informally chaotic way there would be no real pattern: but it drops off in a regular straight line pattern.

The second thing we found that at least to a first approximation, the way this line looks is almost the same as you move from conflict to conflict.

So the line of the Colombia conflict looks pretty much the same as the of line of the Iraq conflict. If you look close enough there are some differences between these conflicts, but at first sight they look very similar.

AOAV: This challenges a possible notion of war being chaos?

MS: Yes, I used the word chaos with some hesitation just because in physics and mathematics the term has a particular, technical meaning. In the conventional use of the word “chaos” I would say yes. I would say that it challenges that. It seems to have much more of an attained or well-behaved pattern.

AOAV: This results in a theory that maybe there is an anthropological response to violence where humans respond in prescribed manners?

MS: Yes, I think this is exactly what it suggests. It suggests that there is something about the way people handle armed conflict.

Comparing the Iraq conflict with the Colombian conflict, they look extremely different on a lot of basic levels.

In the Iraq war you see a lot about Sunni and Shias.

In Colombia you get a lot of emphasis on the fact that Colombia is covered with jungle and lots of mountains.  Rebels can use those to hide in, and that’s been an important part of the conflict. And since practically the whole population of Colombia is Catholic, the conflict doesn’t seem to be about religion.

These conflicts look very different, but then on the other hand you look at the basic measures, like the size distribution of the conflicts’ events and they very look similar after all. They suggest there is something about the way humans fight: this actually remains constant across conflicts.

AOAV: Does this mean that intervention could help to disrupt the patterns that you may come to expect in conflicts? 

MS: I think that’s possible. Yes, we developed a mathematical model of armed conflict that can generate the patterns that were observed in the data. There are plausible candidates for capturing the way the actual dynamics play out. Within these models you can do some experiments to see how an intervention might play out.

I would say that research at this point is still at a fairly rudimentary stage but I think that going forward we might be able to understand the impact of different kinds of interventions a little better.

I mentioned earlier that we did some work on the timing of the events. You look at the number of days in which there are no conflict incidents, how many days are there in which there is one or two or more incidents.

That way you can see what percentage of days is very busy and what percentage is very light and then try to build a model that would generate that size distribution.

The idea of the model is actually about competition over media attention.

There are a lot of armed groups out there and they would all like to generate a story on an event they are responsible for. The idea is that, from a perspective of such a group, if they have fairly spectacular event on a certain day and there aren’t any spectacular events from other groups, they will then monopolise the media coverage of the conflict on a particular day.

On the other hand, they would rather not have their big attack happen on a day when there are a bunch of other attacks.

There are many armed groups out there, probably without means of communication or coordination, all trying to guess when the others are going make the attack. They are trying to find a quiet space in which they can make the attack and get all the attention. It’s a competition over scarce resources: but not oil, food or water.

Paradoxically, although the objective is try to find a little quiet space where you can do your attack, we do see many all trying to attack at the same time.  They have a tendency to generate a surprising number of very busy days when they are all attacking at once without getting that much media attention.

This model, the “complexity theory”, was developed largely by physicists and has been widely used, for example to explain things like traffic jams. Imagine you there are two different ways you can drive to work in the morning and a bunch of other drivers were driving down similar routes. Every day, it depends on whether you decide to take route A or route B. All of you have a rule of thought that you are following. So maybe your rule of thought is that if you do route A today and you encounter a huge traffic jam, you don’t repeat the same mistake tomorrow. Maybe, my rule of through is that a lot of people are going to be thinking the same the next day and they are going to do route B as well, so I’ll just come right back with route A – trying to out guess everyone else. Other people may have similar rules.

There are suggestions that this is similar with attacking patterns of armed groups. Armed groups might be thinking: “OK, Monday has been a quiet day for the last few weeks, let’s attack on a Monday.” But maybe three other armed groups were thinking exactly the same thing, so suddenly Monday is a busy day…

With models of the sort we are looking at, we can generate patterns that look very much like the empirical patterns that we see in actual conflicts.

AOAV: You suggest that the desire to get media attention has a degree of influence on the type, intensity or spacing of armed violence attacks. Could that observation be contaminated by the fact that the evidence you are getting is often based on the data itself from media reports – you may have been missing some attacks that are not covered by the media?

MS: This raises a lot of issues – some of which have been debated fairly vigorously already.

My first response is that if we think that this idea of competition for media attention is a driving force in determining the timing of an attack then attacks that don’t get reported in the media are not really that relevant to armed groups. What they would be predicting is media reported events and they would base their decisions on those.

If we step back and ask about conflict data, event data and the dynamics of how events get reported in the media, the data we gather is based largely on media reports. We need to pause and think about that in conflict research. There is evidence that proves that the media are more likely to report large events than small events.

We saw this actually from a very early stage in the Iraq Body Count data. One thing that came out very clearly was that if you just list the number of media recorded conflict data for events of different sizes, it’s very clear that the larger the event the more media there will be: events in which one person is killed may draw one report, possibly two or three reports; events in which 10 people are killed will draw dozens of reports.

Once you recognise this, you realise that there must be events around small sizes that are drawing zero media reports. Events that just don’t get reported by the media.

This point was further clarified when Wikileaks’ Iraq War Logs came out. There we had a completely different source of event listings, reports from soldiers in the field rather than the media.

We took a random selection of events that appeared in the soldier’s reports and tried to see whether they were also in the IBC’s databases.  So some of them were, some of them weren’t. It was also very clear that the smaller the event the more likely it is that it didn’t appear already in the IBC’s (media sourced) database.

We estimated that once the Iraq war logs were fully processed we will find about 15,000 deaths that appear in the war logs that had not already come to the attention of IBC.

So, if you have largely media reported databases – and there are a lot them – you can expect that you are going to do a much better job in capturing large rather than small events.

This thing about how more events being missed at the low-end rather than at the high-end, and in particular in the case of Iraq, is often presented as a critique of the IBC database. But IBC was actually the first to point out this out.

Now that the Iraq war logs are public, we have a lot more information now about how much might be missing at the low-end now. It’s still difficult, virtually impossible, to definitively settle this issue because there are going to be a certain number of events that are recorded neither in the IBC nor in the war logs.

The question is: “How can you find those?”

But generally you can’t, they are lost forever. Though there have been some successful efforts, particularly in the Balkan wars in the 90s and beyond, in doing really exhaustive victim-by-victim list of every single person killed.

I am involved in a project trying to evaluate the quality of one of these efforts, which is known as the “Kosovo Memory Book”. In this, they tried to list exhaustively every single victim of the Kosovo conflict. They are up to about 13,000 to 14,000 victims killed.

It really does look like it is pretty close to being an exhaustive list at this point. There are various ways in which we can know this. For example, there was a sample survey that was done at the time that made an estimate based on knocking on doors of households and asking how many people from the household were killed in the conflict. Then you would extrapolate from this result. That estimate was around 12,000, but there was a very wide margin of uncertainty around that.

AOAV: In terms of understanding armed violence, I can very much see the utility of an incident driven analysis in terms of being able to predict future conflict in the way they may turn out. What is the utility, above and beyond the moral imperative, about a victim-led book of the deaths?

MS: Good question. But my answer would pretty much be along the theme of moral imperative.

I think families of victims want to see the fate of their family members recognised publicly. If you have a child killed in a conflict, I think people benefit psychologically from seeing that the death is acknowledged and their child has appeared on the list. I think it can also potentially help in a healing process.

What seems to also happen almost inevitably in every conflict is that exaggerated stories take hold of what actually happened. Normally there will be wildly exaggerating casualty numbers circulating.

AOAV: From an armed violence reduction perspective, is there a moral argument for an exaggeration of casualties figures?

MS: People make that argument sometimes more explicitly than in other cases. I think sometimes fairly explicitly and sometimes kind of implicitly, people feel that the larger the number the better.

All people have different motivations. Often, they are trying to generate an outcome and generally tend to feel that higher numbers are better for the outcomes they are trying to generate. Either they are trying to motivate interventions or to save lives.  So it’s better from their perspective if more people are reported dying, they can potentially save more lives through intervention in that way.

In some cases, like in the case of Iraq, where the US and UK have already intervened, I think some of the very high and exaggerated numbers were meant to try to get the US and the UK to leave. The idea being that the US and UK are causing all these deaths.

I’ve seen attempts recently in the context of Syria, where people are anti-interventionist and do not want the US, the UK, NATO or any group like that to intervene in Syria.

They are trying to play down the numbers there, thinking that the higher those numbers the more likely, at some point, this sort of liberal interventionist argument is going to succeed.

AOAV: There may even be a propaganda argument: boost the numbers being killed, so as to impact the morale of the enemy?

MS: That could happen as well. I have a former PhD student of mine, who is Colombian and some government office contacted him and said: “Look, we’ve been collecting information for a few years now, using press releases by the FARC about the FARC battling the Colombian Army. We think that there is a lot of interesting information here. If you want, you can put it together and maybe make some reports and draw some conclusions.”

So he started to compile them as a database rather than a collection of press reports but it quickly became obvious that these press releases were pretty much just fabricated from bottom to top.

We don’t do anywhere near as good a job figuring out when civilians have been killed in conflicts.

If a British soldier has been killed we know. There is exhaustive recording and in Colombia it turned out to be the same. So the FARC was making all these claims about killing all these Colombian soldiers and it was clear that this was just not happening. For them this helped bolster the morale.

I think a similar thing will be happening in the Syrian conflict, where you know that there is actually not a lot of media present and a lot of reports are coming from the field. It is great that we have groups out there that are compiling that information. But I do wonder, whether a lot of false information is actually creeping in for the reasons you suggested.

I just wonder if, in part, some of the data is showing that because we are just getting lots of claims of successful attacks against the Syrian government – just like in the case of the FARC. If you just read reports of the Syrian conflict incident-by-incident, you don’t feel, at least in recent months, that the rebels are getting the best of it right now.

AOAV: You’ve addressed a lot of different complexities in getting data on civilians and military and war, and you’ve addressed issues of different approaches to analyse these numbers. Maybe intervention could be a tool whereby you could then intervene at a certain point and try to disrupt possible future attacks and use your analysis as a way of doing so. A lot of your work, I assume, is historical analysis. Do you think that it could ever be a real-time analysis of casualties in a conflict using some of the findings you’ve done out of Iraq and Colombia which would enable armed violence intervention as the violence happens to occur?

MS: I don’t think we are anywhere near the point where I could envision something like effective intervention. I would not exclude as a possibility for the future, but I do not want promise too much.

AOAV: What do you see as the benefits of carrying out casualty recording?

MS: There is one very basic thing that is connected to some work I did in the past. A former medic, now a Colonel in the British Army, who was based in Afghanistan, told me that by analysing events in which civilians were killed by ISAF troops they were able to introduce some improvements in some of the procedures: for example in ‘Escalation of Force’-type incidents. He claimed that it had an impact in reducing such events were ISAF soldiers were killing civilians.

AOAV: Iraq Body Count is doing an analysis of interventions on Escalation of Force in Iraq.  Perhaps we can see from their findings if there is a possible virtue in real-time analysis by military of casualties…

MS: Yes, I agree with that. I don’t think it had anything to do with fitting any curves to data, it was just taking a look at the incidents and trying to see what is the common denominator was.

The Colonel I’ve mentioned also said that firing warning shot is a really bad idea because quite a few people in Afghanistan could quite easily have weapons in their car.

Clashes with cops are surprisingly prevalent in Afghanistan – hearing a warning shot is generally not taken a reason to calm down and people might take it as a provocation so might they start firing back.

I can understand, some military people thinking that just firing a shot into the air might be helpful. But, I gather it is not, really.

AOAV: It was similar in Iraq. I think, they realised relatively quickly that once an American soldier opened fire, the impulse of Iraqi drivers was to speed up and try to get the hell out of there. For the Americans, that’s an indication of a possible suicide attack and they would take the car out. This resulted in many civilians being killed. So, maybe the analysis of data reveals cultural misunderstandings that may occur between troops and locals.

There has been a massive rise around the world of casualty recordings and seems to be much more commonplace now than it was 20 years ago. What do you think of the most exciting developments in terms of approaches towards casualty recording and quantitative data at the moment?

MS: Although a moment ago I was a little sceptical of some of the data coming out of Syria, I still think this is a really exciting development. Despite the really low-level of media presence in Syria right now, we are able to extract lots of data, lots of information. I think we need to develop techniques to assess the reliability of the data. And I think that some of the work I described earlier on the size distribution of events have the potential to address that issue.

AOAV: So if something strikes you as a major anomaly, then you may say “Well possibly this isn’t reliable” because if look at what you’d expect to see on a predictive curve you can say: ‘this is standing out’?

MS: Yes, that’s exactly what I am saying. In a recent paper I published we found very regular patterns on the size distribution of the conflict. We have been able to extend that a lot, to look at more conflicts since then.  And, I think, painting with a broad brush, we can say that we a have a pretty good idea now about the spectrum of what we observe for the size distribution of the event in a quite range of conflicts now.

So if we receive conflict data from a new conflict and it falls completely outside that spectrum, I don’t know if I can just blow the whistle and say “That’s not reality”, but I would be very sceptical.

I think that I would need to see some pretty strong argumentation why this conflict looks different from what we normally see and quite a lot of other conflicts. We need to think more about that and perhaps research this subject specifically. But I would be fairly confident at this point saying that we can do something significant and important along those lines.

There have been some minor changes to this article since its publications following a request from Professor Spagat. These changes have been carried out for the purposes of clarity.