Skip to 0 minutes and 14 seconds Welcome back, everybody. I’m fresh from my Swedish lesson and ready to proudly announce the names of my two new guests. They are Therese Pettersson from Uppsala University in Sweden and also Lotta Themnér, also from Uppsala University in Sweden. And they work with a number of databases that are maintained in their department, but, in particular, we’re interested in one database called the GED Database. So could you please explain to me a little bit about this, starting with what GED means. So GED is short for a Georeferenced Event Dataset, and it’s the most disaggregated data that we produce at the UCDP. It’s event level, so down to village and date level for incidents of violence.
Skip to 1 minute and 8 seconds OK, so we’ve been discussing in the course really two different methods for documenting armed conflict activity. And one is sort of listing things death by death, or person by person, and we spent most of our time on that, but also I talked about how sometimes we organise things event by event. So you’re in the event data category. And so what is it that you’re interested in recording specifically? So an event is an incident of organised violence at a specific date and a specific location where at least one person is killed from organised violence between two actors that we can identify.
Skip to 1 minute and 58 seconds And so arranging the data by events would then mean that one attack would sum up all the victims from that particular attack and not listing them death by death. Well, we actually code for three different categories of organised violence that we look at, and those are the state-based armed conflict, which people tend to view as civil wars or conflict between states. It could also be one-sided violence, where we actually have one from the organised act. It could be a state or a rebel group who attacks civilians and kills them. And then non-state conflict, which is basically conflict between non-state actors, be it communal violence or conflict between rebel groups or drug cartels in Mexico.
Skip to 2 minutes and 46 seconds So all of these three are coded to be mutually exclusive, and we can actually add them up together and come up with a total number of fatalities in organised violence. So that’s our definition. We have a cutoff point of 25 fatalities in a year to be actually included in our data. What kinds of sources do you use to find these events? Well, we use a wide range of sources. The first step is doing a search in database for news articles, so it’s always open source, but the first step is news articles. So we read tens of thousands of news articles every year.
Skip to 3 minutes and 24 seconds And then we add information from NGOs, from local sources also, like international organisations, we read books when that is available, but that’s usually a little bit later on, so then we can go back and revise the data, also reports from truth commissions, and so on. But the first step is international media, as well as local sources. And what we do is that we– the first step, like Therese said, that we go through news articles for every country in the world to be able to kind of pick up if something is going on somewhere, and then we can dig deeper in that particular country.
Skip to 4 minutes and 4 seconds But that’s why there’s so extremely many articles to go through, to kind of find things before they start escalating. So what we always do is we always revise the data. We just don’t code one year and leave it, but we know if you look at our version history documents that we have for all the datasets, I mean, there are a long list of changes every single year because we always find new things and there are new sources coming out, so we always want to improve the data continuously.
Skip to 4 minutes and 30 seconds So I’m delighted that you said that, actually, because in one of the lectures I talk about, that’s one of the principles of good practise is that you always consider it a work in progress. And as you find mistakes or just new information, it’s always subject to our vision, which is good. And then on top of it, you actually keep track of the versions, which is even better. To what extent do you think the database as a whole is usable for providing accurate information on global trends in violence over time? I think so, yeah. I mean, we don’t claim to have every second fatality, every record in our data, definitely not.
Skip to 5 minutes and 21 seconds But I think that we can be pretty sure that given the definitions that we have, we have at least like a baseline, at least this number of people were killed, you know, definitely probably more, but at least this number, and I think we can definitely see that we have captured the trends in large. And you were talking about Deep South Watch. Yes, so our main strength is that we can capture the trends globally and across time. So we’re not aiming to be like the perfect case study for each and every conflict in the world, but more like have this global picture and be able to capture the trends.
Skip to 5 minutes and 56 seconds And, for instance, we have been cooperating a little bit with a project in Thailand called Deep South. Yes, and they have obviously much more access to better data since they’re based there, and they have higher fatality numbers than we do, but the trend is the same. So that’s the important thing, that we can sort of feel confident that we’re capturing the trend and that you can compare cases across time and space with our data. OK, great. So it’s been great talking to you. Thank you very much for coming. Thank you. Thank you.
Therese Pettersson and Lotta Themnier on the Georeferenced Event Dataset
In this clip I interview Lotta Themnér and Therése Pettersson, both of the department of peace and conflict research of Uppsala University. We talked about the Georeferenced Event Dataset (GED) produced by the Uppsala Conflict Data Program for which Lotta and Therése are both project leaders.
From the perspective of our course I underline a few points about the GED database.
It is a global database that begins its coverage in 1989 and is updated annually.
It fits squarely within the event-data branch of casualty recording methodology that we studied in week 1. That said, the coding team applies relatively stringent requirements for the inclusion of events in the database, e.g., that the identities of perpetrating groups must be known.
It provides the location of each event. We have not discussed detailed geo-location in the course but this is an important and notable feature of the dataset.
Upon my request Lotta and Therése provided the following bios.
LOTTA THEMNÉR (previously Harbom), b. 1975, MA in Peace and Conflict Research (Uppsala University, 2002); project leader, Uppsala Conflict Data Program, Department of Peace and Conflict Research: articles on conflict data published in the Journal of Peace Research and SIPRI Yearbook since 2005; editor of States in Armed Conflict between 2004 and 2012.
THERÉSE PETTERSSON, b. 1982, MA in Peace and Conflict Research (Uppsala University, 2008); project leader, Uppsala Conflict Data Program, Department of Peace and Conflict Research; articles on conflict data published in the Journal of Peace Research since 2015: editor of States in Armed Conflict between 2010 and 2012.
© Royal Holloway, University of London