The most commonly used statistical models of civil war onset fail to correctly predict most occurrences of this rare event in out-of-sample data. Statistical methods for the analysis of binary data, such as logistic regression, even in their rare event and regularized forms, perform poorly at prediction. We compare the performance of Random Forests with three versions of logistic regression (classic logistic regression, Firth rare events logistic regression, andL1-regularized logistic regression), and find that the algorithmic approach provides significantly more accurate predictions of civil war onset in out-of-sample data than any of the logistic regression models. The article discusses these results and the ways in which algorithmic statistical methods like Random Forests can be useful to more accurately predict rare events in conflict data.
Although some rebel groups work hard to foster collaborative ties with civilians, others engage in egregious abuses and war crimes. We argue that foreign state funding for rebel organizations greatly reduces incentives to "win the hearts and minds" of civilians because it diminishes the need to collect resources from the population. However, unlike other lucrative resources, foreign funding of rebel groups must be understood in principal-agent terms. Some external principals-namely, democracies and states with strong human rights lobbies-are more concerned with atrocities in the conflict zone than others. Multiple state principals also lead to abuse because no single state can effectively restrain the organization. We test these conjectures with new data on foreign support for rebel groups and data on one-sided violence against civilians. Most notably, we find strong evidence that principal characteristics help influence agent actions.Wartime atrocities are often portrayed as irrational, depraved, and senseless. However, recent research in this area has increasingly viewed violence as strategic, adopting Arendt's argument that violence is rational to the extent it assists actors in attaining their goals. 2 Scholars have suggested that actors resort to victimizing civilians to improve their bargaining position over their adversary's, expedite war termination, and generate resources. 3 This vein of research also claims that attacks on civilians are intended to alter the behavior of the targeted groups 4 and to eliminate disloyal or threatening populations. 5 We focus on the manner in which rebel groups obtain resources and how this influences their behavior toward civilians. All rebel organizations, regardless of their ideology and putative grievances, need to secure resources to finance their operations. 6 Guns, uniforms, and other war supplies are expensive to procure. However, as Weinstein has shown, rebel organizations differ greatly with respect to the resource environments in which they operate, and this often shapes their behavior toward civilian populations. 7 Resource-poor rebels without easy access to commodities such as drugs and gemstones must rely on the goodwill of the population. By fostering deep local ties and protecting the interests of ordinary people, rebel organizations can both secure goods and ensure "moral commitments and emotional engagements" to their cause. 8 By contrast, resource-rich rebels are less dependent on civilians for their needs and their survival-a condition that may lead to abuse and mistreatment of civilians.
This article addresses current methodological research on nonparametric Random Forests. It provides a brief intellectual history of Random Forests that covers CART, boosting and bagging methods. It then introduces the primary methods by which researchers can visualize results, the relationships between covariates and responses, and the out-of-bag test set error. In addition, the article considers current research on universal consistency and importance tests in Random Forests. Finally, several uses for Random Forests are discussed, and available software is identified.
Case studies suggest that ethnic groups with autonomous institutional arrangements are more prone to secede, but other evidence indicates that autonomy reduces the likelihood of secession. To address this debate, we disaggregate their autonomy status into three categoriescurrently autonomous, never autonomous, and lost autonomy-and then unpack how each shapes the logic of collective action. We argue groups that were never autonomous are unlikely to mobilize due to a lack of collective action capacity, whereas currently autonomous groups may have the capacity but often lack the motivation. Most important, groups that have lost autonomy often possess both strong incentives and the capacity to pursue secession, which facilitates collective action. Moreover, autonomy retraction weakens the government's ability to make future credible commitments to redress grievances. We test these conjectures with data on the autonomous status and separatist behavior of 324 groups in more than 100 countries from 1960 to 2000. Our analysis shows clear empirical results regarding the relationship between autonomy status and separatism. Most notably, we find that formerly autonomous groups are the most likely to secede, and that both currently autonomous and never autonomous groups are much less likely. Keywords nationalism, separatism, decentralization, autonomy Does autonomy satisfy the demand for more self-determination, or instead foster the capacity and whet the appetite for independence? Is political 2 Comparative Political Studies XX(X) decentralization a viable solution for multiethnic states, or a slippery slope? Some scholars and policymakers see autonomy as the main mechanism to resolve tensions and redistributive issues between the central government and spatially concentrated, culturally distinct groups (Bermeo, 2002;Bermeo & Amoretti, 2003;L. Diamond, 1999;Stepan, 1999). Others studies, however, show that autonomy can actually exacerbate relations between the state and ethnic groups, for it cultivates the capacity for collective action and self-rule without significantly reducing desire for more of it (Brancati, 2009;Bunce, 1999;Coppieters, 2001;Cornell, 2002;Roeder, 1991).This article posits that the theoretical and empirical disagreement over the effect of autonomy is partly due to conflating two distinct situations in the implicit reference category-groups that are not autonomous. This binary classification compares groups that are currently autonomous to a baseline that is comprised of both groups that have never been autonomous, and those that once had autonomy, but lost it. Groups that have never been autonomous and those that have lost autonomy are distinct in their motivation and their capacity for collective action. Whereas never autonomous groups may be unlikely to mobilize due to a lack of collective action capacity, currently autonomous groups may have such capacity, but should find the cost of attempting to alter the status quo too high compared with the benefits they currently enjoy from being auton...
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.