mixed regimes is that positive unexpected events do not make them more likely to exit a war.
This is inconsistent with both the unexpected events hypothesis and Goemans’ regime type
hypothesis.
Even given these difficulties, the unexpected events hypothesis does reasonably well. If
regime type is ignored, unexpected events are clearly correlated with when states choose to exit
wars. When regime type is included, the results are still strong for dictatorships, though far
weaker for democracies and mixed regimes. Goemans’ regime type hypothesis is consistent with
the behavior of dictatorships and to some extent mixed regimes, but is clearly inconsistent with
the behavior of democracies.
It is possible the difficulties with some of the results for the regime type variables is that
they fail to take into account the regime type of the belligerents’ opponents. Naturally, states
cannot unilaterally end a war, though the sometimes can unilaterally exit from a multilateral
conflict.
Perhaps future studies will correct this methodological flaw and in doing so explain the
lack or reaction of mixed regimes to positive unexpected events. It seems likely these hypotheses
are on the right track, but as of yet are incomplete. Only future studies can determine this for
certain.
Appendix: Coding Rules for the Unexpected Events and Season Variables
As was mentioned in the main text, the descriptions and coding of the unexpected events
variables need to be covered in detail. This appendix will describe the variables and discuss the
rules and sources used for their coding.
Five variables indicating whether unexpected events happened during a particular war-
month and the nature of those events are included in the dataset. They are Unexp, Mil, Pol,
Impmil and Imppol. Unexp is a dichotomous variable (0, 1) which records if an unexpected event
34 Such an exit does require the tacit agreement of the opponents of the state exiting the conflict.
19
Convention