longer if they are based in mountainous or forested terrain. Groups received one point if they
were an ethnic group that was dominant in the region where they operated, since that would
suggest that they had higher popular support.
Since “lootable resources” can help groups fund
rebellion, those that had access to such resources got a point.
elections and made a strong showing, they received a point.
Finally, since external support can
enable groups to continue rebellion longer, I gave each group one point if it received substantial
support from external states. This support could not simply be rhetorical, but had to involve
actions such as direct military support, a state allowing the group to base on its territory, or
significant financial support.
The terrain, group dominance, lootable resources, and election performance indicators do
not apply to external states intervening in conflict. To determine the points awarded to external
states, I used the same troop criteria. However, I gave each external state one extra point, based
on the logic that states that can send their armies to another country are generally better
organized relative to domestic insurgent groups. I awarded an additional point if the external
state was a major power, since that is an indicator of how strong the military is.
After completing this coding I had a score for each group (ranging from 0 to 7) that was
designed to indicate how strong they were. For example, the Revolutionary United Front (RUF)
in Sierra Leone in 1991-1996 has a score of four because it had more than 10% the troop strength
of the government (two points), controlled diamonds, and received direct military support from
Liberia. The Alliance for Democratic Forces (ADF) in Uganda has a score of two because it had
1% the troop strength of the government and received support from the Sudan. Force Obote Back
29
The main source I used to code this criterion was Minorities at Risk (MAR), which generally indicated whether
groups were a majority in the area in conflict.
30
In order to code access to resources I started with data from Buhaug, Lujala and Gates 2004 who measure whether
there are drugs, diamonds or gems in the zone of each country in conflict. I then used historical sources to find
references to which groups controlled resources for each conflict that had them.
31
To determine if groups had participated in elections, I used Keesing’s Record of World Events.
32
I coded external support using historical sources and case histories.
27
Convention