Rare Events in International Relations: Modeling Heterogeneity and Interdependence with Sparse Data

摘要: The interdependence of international events is obvious to even casual observers global politics. History replete with examples repeating within states and/or being influenced by outcomes in other states. Despite this, much the current literature International Relations either mishandles or outright neglects this dependence, thereby threatening credibility our inferences. In large part, stems from difficulty modeling such dependence when one's data are binary and rare, as they often for many most widely-studied phenomena IR (e.g., violent conflict, economic crises, etc\ldots). For type, commonly-used strategies capture frequently ill-suited and, such, new approaches required. Therefore, thesis aims clarify empirical challenges which arise these data, detail problems existing approaches, offer alternatives should be preferred. The focus principally on two potential (and related) sources bias may time-series cross-sectional (b-TSCS) data: true (inter-)dependence unit heterogeneity. first, outcomes, actions, choices some unit-times depend directly those unit-times. To model both spatial serial a spatiotemporal-lag probit estimated using maximum-simulated-likelihood recursive-importance-sampling (MSL-by-RIS) presented. This allows us lagged-latent shown have several advantages over models observed indicator consistency, effects estimation, predictive accuracy). second main threat unobserved heterogeneity, that is, time-invariant unit-characteristics influence outcome, action, choice, but go unmodeled. While fixed estimators traditionally solution issue, received heavy criticism political science applications b-TSCS data. light criticisms, penalized-maximum-likelihood fixed-effects (PML-FE) proposed suffers few drawbacks permits estimation novel unit-specific substantive effects. addition, original analyses into intrastate conflict financial crises offered highlight value testing existing, motivating new, theories behavior.