Skip to 0 minutes and 13 secondsHello, everybody. I'm here now with Bethany Lacina, from the University of Rochester, and we're going to talk about something called the PRIO battle deaths data set. So perhaps, you could start off by explaining, first of all, what is PRIO, which would be easy. And then maybe more complicated after that, what is a battle death? OK. So the PRIO is the Peace Research Institute in Oslo. A battle death, for the purposes of this data set, first of all, had to take place in the context of a state-based armed conflict, as defined by all the sister data sets. And then a battle death was someone killed in the course of combat. It could be a civilian, could be a soldier.
Skip to 0 minutes and 58 secondsBut they had to be killed in combat, as opposed to an indirect non-violent death, say, like a death due to disease or violence that was unambiguously one-sided, like the execution of a prisoner. Recording these things called battle deaths-- and then that is really only part of all of the death that might be going on within a war. So maybe you can pursue that a little bit. So think about there being battle deaths, which are-- I suppose the most restrictive thing that you might look at are the deaths of soldiers. Battle deaths would be a superset, that would include civilians killed in combat, as well.
Skip to 1 minute and 53 secondsThen you would possibly include people killed in non-combat situations, like the prisoner-style style execution or, say, the Holocaust in World War II. You'd consider that not combat. And then there's going to be excess deaths, which are deaths due to the kind of general privation that takes place during a war, but aren't necessarily violent. Then my next question would be, why don't you try to capture more than That why only focus on this one particular type of death within a war? Remember, I said so this is one data set in this suite of data sets. And they all have slightly different purposes.
Skip to 2 minutes and 41 secondsAnd so there's a separate set of people who compile one-sided violence and violence that takes place without state involvement. The focus on battle deaths, or the reason to have one data set that did just that, was driven by an interest in trying to create as long a time series as possible, and knew going into it that would not going to get anything other than battle deaths for conflicts that took place in the '40s and '40s. And since the point was to have a consistent series, we went forward, based on, well, this is what we can observe at T equals 0. So we'll try to record something consistent with that, going through time.
Skip to 3 minutes and 31 secondsSo this particular project is driven by a desire to understand long-term trends in global war violence. Let's change gears a little bit. Could you describe what you went through to actually compile this thing? So it became clear that there was really wide variation in the quality of the data we were getting. So we decided to pull together two things. The data set itself looks like a table, where each row is a year of conflict, and there's an estimate of how many people died in battle. And there's a low estimate, a high estimate, and then where we feel like we have some sense of what the better of those two is or some other, better estimate. There's a best estimate.
Skip to 4 minutes and 25 secondsBut because the quality of the data was so all over the place, we decided to create really very extensive documentation. I'm sorry-- so you're basically, you're looking for every possible source you can find? Right. And you're looking at them, and you're often saying, like, well, these don't agree at all. Or there's no-- I mean, the bigger problem is you've got sources, and there's just no indication of how would the author know this? Where did they get the number? And so the documentation is, I don't know, 400 pages or something.
Skip to 5 minutes and 3 secondsAnd for each conflict year, it's a listing of every estimate we found and where we found it, and then a discussion of why we thought what we thought was most credible. And the aim was that, for any given part of that table, a user could go into the documentation and come to their own conclusion about, so what's the quality of this particular estimate? And even more important, have something that was like a little bibliography of-- so here's what's known about battle deaths at this particular time and place. If there's something you can add or improve upon, this hopefully will serve as a jumping off point for your research.
Skip to 5 minutes and 52 secondsSo I should say that it's really worthwhile to people taking this course to have a look at this documentation document. And I will provide a link for it on the course website. So what you can do-- you can basically go to any war that you're interested in. And then you can see what are all the numbers that have circulated about this war? And then you get a very succinct assessment of the quality of these sources and which ones seem more likely to be close to the truth, and then references, so that you could follow this up more, if you're so inclined, yes. I mean, for someone like me this is something I could spend hours just looking through.
Skip to 6 minutes and 51 secondsI mean, I admit that perhaps I'm not typical in this regard.
Skip to 6 minutes and 57 secondsBut then it's just a monument to the tremendous amount of work that Bethany did on this project. I mean, I wouldn't even venture to guess how many hours this must have taken to go through all this stuff. So it's far more than just a spreadsheet with numbers. There's just a huge amount of information, if you go and you read the documentation. So hopefully, someone taking this course will find it as fascinating as you, or maybe 75% as fascinating. Well, thank you very much for coming, appearing in our MOOC. I'm sure that the students will greatly appreciate this.
Bethany Lacina on the PRIO Battle Deaths Dataset
In this clip I interview Bethany Lacina who is the James P. Wilmot assistant professor of political science at the University of Rochester. We talk about the PRIO Battle Death Dataset which is the product of a major effort, spearheaded by Bethany, to gather together the best possible estimates of battle deaths for every war in the world after World War 2.
Here I underline a few points from this conversation that are particularly relevant to our course.
The project made strenuous efforts to include only violent deaths. In particular, excess deaths were excluded, mainly because reliable excess death information is rarely available.
The project only included a subset of violent deaths which are designated as “battle deaths”. Broadly speaking this means that killings of civilians are only included when these killings occur within the context of battles between organised armed groups. Again, the reason for this restriction is to limit the data to a core of violence forms that have been measured with relative consistency over many decades.
The project evaluates the quality of all battle-death information the researchers can find and then makes judgments on coding decisions. Some of this information comes from casualty recording projects of the sort we covered in Week 1. Some information comes from sample survey and capture-recapture estimates of the sort we covered in Week 2. Other information comes from other procedures such as informed guesstimates of historians or war participants.
Upon my request Bethany provided the following bio:
Bethany Lacina is an assistant professor at the University of Rochester Department of Political Science. She received her PhD in political science from Stanford University. She recently published Rival Claims: Ethnic Violence and Territorial Autonomy under Indian Federalism with the University of Michigan. Her work cataloging the costs of war can be found through the PRIO Battle Deaths Data project.
© Royal Holloway, University of London