International cooperation: Although sanctions were less successful than policy-
makers had anticipated, the number of sanction episodes did not decline in the 1990s.
However, unilateral sanctions are imposed less frequently then than in the 1970s or
1980s (Elliott and Hufbauer 1999). One reason for this change is the di¢ culty of en-
suring the cooperation of other countries in enforcing sanctions imposed unilaterally.
Not only does international support reduce sanctions’ cost to the sender, but it also
makes the target more likely to satisfy the sender’s demands. Thus, it is plausible to
argue that the outcome of sanctions depends on the response of third countries that
can either cooperate with the sender or exploit economic opportunities generated by
sanctions.
Threat stage: Sanctions are unsuccessful because most successes are accomplished
at a threat stage. Before engaging in sanctions, the sender is likely to achieve compli-
ance by pre-communications with the target. Niou and Lacy (2004) explain formally
that the omission of the threat stage can result in selection bias where the failed
outcomes are overestimated and the successful outcomes are underestimated in most
empirical studies.
3
The Model
I develop a simple crisis bargaining game with two-sided incomplete information and
two unitary actors, a sender and a target.
6
The sequence, outcomes, and payo¤s of
the game is displayed in Figure 2.
7
The structure of the game is as follows:
1. The sender chooses either to stay at the status quo or to impose sanctions. If she
selects sanctions, there are n-levels of possible sanctions and the sender chooses one of
them.
2. Having observed the sender’s decision, the target decides either to comply or to
resist. If she complies, the game ends peacefully (compliance).
3. If the target resists, the sender decides either to back down from the initial sanctions
(back-down) or to continue them (deadlock).
6
A simpler version of the model was developed by Lewis and Schultz (2003). A di¤erence lies in
the fact that I allow the sender’s …rst choice to be multiple, not binary, so that the sender can adopt
an optimal intensity of sanctions.
7
At each …nal node, I write the outcomes and the corresponding payo¤s of the sender and target
in parentheses, i.e. (sender’s payo¤, target’s payo¤).
7