There are three fundamental duration dynamics of civil conflicts: time until conflict onset, conflict duration, and time until conflict recurrence. Theoretical and empirical models of war usually focus on one or at most two aspects of these three important duration dynamics. We present a new split-population seemingly unrelated duration estimator that treats pre-conflict duration, conflict duration, and post-conflict duration as interdependent processes thus permitting improved predictions about the onset, duration, and recurrence of civil conflict. Our findings provide support for the more fundamental idea that prediction is dependent on a good approximation of the theoretically implied underlying data-generating process. In addition, we account for the fact that some countries might never experience these duration dynamics or become immune after experiencing them in the past.