29
model, the peace years variable was set to 10, while for the continuation model, war years was
set to 1. Let us begin with the results for conflict onset, for which only civil war in neighboring
countries was found to be significant. In the baseline hypothetical scenario, a country is
expected to have a 1.5% risk of a conflict. A civil war in a neighboring country boosts this
probability to 3%, for a 100% increase in the risk of violence; clearly, a nearby conflict has an
important impact on onset. For conflict continuation, the effect of rivalry and refugees was
analyzed, with the war years variable set to one. How do these variables affect the probability
that a conflict will endure for a second year? The probability of conflict continuation for the
baseline comparison category is 78%, indicating that conflicts are very likely to persist for a
subsequent year. Holding the other variables constant and shifting the value of rivalry from 0 to
1 raises this probability to 83%, or a 6.4% increase in probabilities. For refugees, moving from 0
to 100,000 refugees in neighbors raises the probability of conflict to 86%, or a 10% increase in
probabilities over the baseline.
Extraterritorial Bases and Conflict Duration
In the above analysis, it is assumed that neighboring state weakness, rivalry, and refugees will be
positively associated with conflict because of opportunities for TNR’s to establish extraterritorial
bases. However, it is also possible to directly observe whether or not a rebel organization has an
extraterritorial base once fighting is underway. Thus, an additional hypothesis was developed
regarding the presence of such bases and conflict duration:
H4 (extraterritorial bases): Conflicts will endure longer if the rebels have access to
extraterritorial bases.
Convention