How Civil Wars End: A Rational Choice Approach

The paper develops a rational choice argument to explain under what conditions civil wars end. They draw some hypotheses to test their argument and run a quantitative analysis. 
Their model is quite straight forward. A civil war is fought between a government (G) and a rebel group (R). They argue that negotiated settlement is more likely when the expected utility of a negotiated settlement is higher than the expected utility of continuing the conflict, for both the government and the rebels. They use COW dataset to test their hypotheses. 
Then they identify parameters that are, at least hypothetically, likely to increase the probability of an end to a civil war with a negotiated settlement. 
1. Probability of victory: They build this argument on what William Zartman called “mutually hurting stalemate”. This is a situation where government or the rebels are unable to defeat the enemy but able to prevent the enemy from a victory. For Zartman, this is a preconditon for a negotiated settlement. In order to measure the balance of capacity between parties, they rely on the size of the government army. They hypothesize a negative correlation between the size of the government army and the likelihood of a negotiated settlement. They argue that a large army can easily suppress a rebellion at the early stages. Their tests show that size of the army matter at the level of 20,000 or more. In a civil war between 2 and 12 months, probability of a negotiated settlement with an army of 20,000 is less than 10%. 
2. Cost of continued conflict: They argue that if parties prefer to continue the conflict, they will bear the costs as well. To measure the cost of continuing the war, they use casualty rate. Therefore, higher casualty rates are expected to increase the probability of a negotiated settlement. Their findings do not lend any support for this hypothesis. The results for this variable is not statistically significant. 
3. Duration of the conflict: They assume that longer the conflict has continued, the greater will be both parties’ estimate of the amount of additional time required to achieve victory. Therefore, any factor that increases the actors’ estimate of the amount of time needed for victory will increase the probability of a negotiated settlement. Their findings show that this variable is statistically significant. 
4. Third party intervention: In this part they clarify the term intervention. In some cases presence of a third-party actor is needed because an agreement or a treaty should be policed and monitored. But in some cases, interventions favor one side over the other. Mason and Fett expect a negative correlation between an intervention by other nations and the likelihood of a negotiated settlement. Their findings do not support this hypothesis. 
5. Finally, they focus on what is at stake in a civil war. They argue that what the stakes are in a civil war and whether they are indivisible or not matter because negotiated settlements are impossible if what is at stake is indivisible. They drawe two hypotheses from this argument. In a revolution, they argue, negotiated settlement is unlikely because both parties fight for the same thing: control of the government. Therefore, they expect that
a) Separatist conflicts are more likely to end with negotiated settlements than revolutions. 
If what is at stake is about ethnic identity, negotiated settlements are less likely. Therefore, they expect that
b) In a civil war that is ethnically based is less likely to end with a negotiated settlement than a civil war that is not. 
Results do not support both hypotheses. 
In conclusion, their main finding is that longer civil wars are more likely to consider a negotiated settlement. This confirms Zartman’s argument that “mutually hurting stalemate” is necessary for negotiated settlements. Their findings also show that size of the government army matter because larger armies are more likely to suppress the rebellion at the early stages. If governmens fail to end conflict at the early stages, conflicst are likely to extend more.