Every Story Has a Beginning, Middle, and an End (But Not Always in That Order): Predicting Duration Dynamics in a Unified Framework

Abstract

There are three fundamental dynamics of civil conflicts: Onset, duration, and cessation. Theoretical and empirical models of war usually focus on one or at most two aspects of these three important dynamics. We argue that a better understanding of conflict needs to incorporate all three conflict dynamics as belligerents' choices to fight in the first place will depend on their expectation of fighting duration and the risk of recurrence once the fighting stops. We introduce a theoretical framework that treats onset, duration, and recurrence as interdependent processes. We also present a new duration-duration-duration estimator that treats pre-conflict duration, conflict duration, and post-conflict duration as interdependent processes thus permitting improved predictions about the onset/recurrence, duration, and cessation of conflict.

A theoretical and empirical comparison of the temporal exponential random graph model and the stochastic actor-oriented model

“…Prior applications of out-of-sample prediction to SAOMs include Desmarais and Cranmer (2012a), Kinne (2013), Koskinen and Edling (2012), and Warren (2016). There is also a substantial literature looking at out-of-sample prediction with other network models (e.g., Chiba et al, 2015;Montgomery et al, 2015;Ward et al, 2013).…”

Military Coalitions and Crisis Duration

“…The first component of (3), the joint probability that a coalition is formed and the crisisphase lasted for duration t i is obtained by applying Bayes' rule and taking the derivatives 28 For applications of copula functions in political science, see Chiba, Martin, and Stevenson (2015), Chiba, Metternich, and Ward (2015), and Fukumoto (2015).…”

Flexible Causal Inference for Political Science

“…For example, to model covariate-response relationships in a more flexible way Chib and Greenberg (2007) and Marra and Radice (2011) introduced theoretically founded Bayesian and likelihood estimation approaches based on penalized regression splines, thereby allowing for a number of different flexible covariate-response structures. Examples in political science include Chiba, Metternich and Ward (2015) and Fukumoto (2015). To deal with the problem of nonlinear dependence between outcomes, Winkelmann (2012) discussed a modification of the recursive bivariate probit which introduces non-normal dependence between the marginal distributions of the two equations using the Frank and Clayton copulas.…”