Research

Publications

The underlying risk factors associated with the duration and termination of biological, sociological, economic, or political processes often exhibit spatial clustering. However, existing nonspatial survival models, including those that account for “immune” and “at-risk” subpopulations, assume that these baseline risks are spatially independent, leading to inaccurate inferences in split-population survival settings. In this paper, we develop a Bayesian spatial split-population survival model that addresses these methodological challenges by accounting for spatial autocorrelation among units in terms of their probability of becoming immune and their survival rates. Monte Carlo experiments demonstrate that, unlike nonspatial models, this spatial model provides accurate parameter estimates in the presence of spatial autocorrelation. Applying our spatial model to data from published studies on authoritarian reversals and civil war recurrence reveals that accounting for spatial autocorrelation in split-population models leads to new empirical insights, reflecting the need to theoretically and statistically account for space and non-failure inflation in applied research. 


Survival data often include a fraction of units that are susceptible to an event of interest as well as a fraction of “immune” units. In many applications, spatial clustering in unobserved risk factors across nearby units can also affect their survival rates and odds of becoming immune. To address these methodological challenges, this article introduces our BayesSPsurv R-package, which fits parametric Bayesian Spatial split-population survival (cure) models that can account for spatial autocorrelation in both subpopulations of the user’s time-to-event data. Spatial autocorrelation is modeled with spatially weighted frailties, which are estimated using a conditionally autoregressive prior. The user can also fit parametric cure models with or without non-spatial i.i.d. frailties, and each model can incorporate time-varying covariates. BayesSPsurv also includes various functions to conduct pre-estimation spatial autocorrelation tests, visualize results, and assess model performance, all of which are illustrated using data on post-civil war peace survival. 


Policymakers and peacebuilding research often focus on rebel groups when studying demobilization and integration processes, but post-war governments must also manage the non-state militias that helped them gain or maintain power. Why do some post-war governments disintegrate their militia allies, while others integrate them into the military? We argue that when a salient ethnic difference exists between the (new) ruling elite and an allied militia, a process of mutual uncertainty in the post-war period will incentivize governments to disintegrate the group. However, governments will be most likely to integrate their militias when the military has sufficient coercive capabilities but few organizational hindrances to re-organizing. Using new data on the post-war fates of victorious militias across all civil conflicts from 1989 to 2014, we find robust support for these claims. The results suggest that a government’s optimal militia management strategy is shaped by both social and organizational constraints during the post-war period. 


In most contemporary civil wars, governments collude with non-state militias as part of their counterinsurgent strategy. However, governments also restrict the capabilities of their militia allies despite the adverse consequences this may have on their overall counterinsurgent capabilities. Why do governments contain their militia allies while also fighting a rebellion? I argue that variation in militia containment during a civil war is the outcome of a bargaining process over future bargaining power between security or profit-seeking militias and states with time-inconsistent preferences. Strong states and states facing weak rebellions cannot credibly commit to not suppressing their militias, and militias with sufficient capabilities to act independently cannot credibly commit to not betraying the state. States with limited political reach and those facing strong rebellions, however, must retain militia support, which opens a “window of opportunity” for militias to augment their independent capabilities and future bargaining power. Using new data on pro-government militia containment and case illustrations of the Janjaweed in Sudan and Civil Defense Patrols in Guatemala, I find evidence consistent with these claims. Future work must continue to incorporate the agency of militias when studying armed politics, since these bargaining interactions constitute a fundamental yet undertheorized characteristic of war-torn states. 

R Software

Working Papers (Drafts Available Upon Request)

The global rise of right-wing populist (RWP) leaders has raised concerns about the threat they pose to a cooperative international order, but there is little systematic evidence linking RWP leaders to military aggression. Are RWP leaders more prone to initiating international disputes? We argue that a RWP leader’s hyper-nationalist rhetoric can galvanize popular support for militant internationalism and confidence in the leader’s ability to use force to achieve national goals, but this only affects dispute initiation in participatory democracies. Using survey experiments fielded in India and Japan and machine learning methods for causal inference, we find strong support for our claims about the effects of RWP rhetoric on civilian attitudes. Statistical results from original data on populist leaders worldwide (1886-2014) then show that RWP leaders in participatory democracies are more likely to initiate militarized disputes. Our results are troubling given the recent increase in RWP leaders elected in participatory democracies.


Warring actors often employ non-state militia groups during civil conflicts, but once the conflict ends, ruling elites must manage these unregulated forces. Auxiliary militias can serve repressive or coup-proofing purposes, but they also may undermine the legitimacy and long-term viability of the post-war regime. Why do some victorious militias survive longer than others in the post-war period? We argue that incumbent and rebel victors face differing security environments following a civil conflict, which, in combination with different political institutions, shapes their incentives to retain or eliminate the non-state forces that supported them during the war. More specifically, incumbent autocrats that remained in power and rebel democrats that seize power tend to retain auxiliary militias in the post-war period, whereas incumbent democrats and rebel autocrats quickly eliminate them after the conflict has ended. Analyses using data on the post-war survival of 169 victorious militias from 1989-2014 are consistent with our expectations. 


Civil wars involving more than one rebel group comprise more than half of all internal conflicts over the past 75 years. Although previous work has shown that the distribution of power, third-party sponsors, and ethnic or ideological homophily affects the tendency of groups to cooperate in these multiparty wars, qualitative variation in the type of rebel alliances is not well-understood. For instance, while some rebel groups conduct occasional joint attacks, others form joint congresses or military commands with elected representatives. Why do some groups cooperate incidentally, while others publicly commit to formal cooperation in the future and still others create joint power-sharing institutions to coordinate their war efforts? In this paper, I propose a new conceptual framework for understanding rebel alliances and their "contract provisions"  based on the level of public commitment made by rebel leaders and whether they agree to  share power over decision-making with the leaders of other groups. I then develop a new theory of intra-organizational bargaining, in which the distribution of power between central leaders and subcommanders affects the group’s credibility in making cooperative commitments to other rebel organizations.  Using an original dataset on inter-rebel cooperation agreements and their features, encompassing over 1,170 rebel-dyad years and 233 distinct rebel organizations in multiparty conflicts around the globe, I find strong empirical support for my theory. The conceptual, theoretical, and empirical contributions of this analysis suggest a need to develop a better understanding of inter-rebel relations, including the causes and consequences of cooperative arrangements between armed organizations in war.  


Multiparty civil wars are longer and more likely to recur than wars between a government and only one rebel group because divergent preferences across rebel organizations hinder the prospects for a successful negotiated settlement. Previous research on these patterns, however, largely ignores the relationships between the rebel groups themselves when they come to the negotiating table. In over half of multiparty civil conflicts, rebel organizations form alliances with other groups in order to improve their military capabilities and bargaining leverage relative to the government. These alliances vary extensively in their form, including whether rebels commit to cooperating politically (e.g., through a joint political platform or in negotiations) or militarily (e.g., in battlefield operations). Can rebel alliances improve the prospects of a multilateral negotiated settlement in multiparty wars, or are negotiations between governments and non-allied rebels just as likely to succeed or fail? I argue that political alliances between rebels signal to governments the opening of a bargaining space, thereby increasing the probability of multilateral negotiations and peace settlements. However, military alliance provisions affect the credibility with which rebels can return to war as well as the costs for each party to continue fighting. These alliances therefore put pressure on governments to negotiate, but also make rebels more likely to pull out of negotiations before signing a peace agreement. Using original, global rebel alliances drawn from both primary texts and a variety of secondary sources and a polytomous, two-stage research design to model both negotiations and settlements, I find evidence consistent with my argument. The results have important implications for understanding multiparty conflict processes and resolving these particularly complex civil wars.


The prevailing scholarly wisdom is that rapid democratization in post-conflict settings increases the probability of conflict recurrence and slows economic recovery. These conclusions paint a grim picture for post-civil war societies because political exclusion and economic grievances could cause conflict recurrence. I challenge this conventional wisdom conceptually and empirically. Since civil war violence affects certain subnational locations more than others, aggregate measures of economic reconstruction (e.g. GDP growth) distort the reality of reconstruction efforts actually being implemented in previously violent locations (PVLs). I argue that when ideologically motivated rebel groups are allowed to participate in post-war elections, the combination of the mobilization efforts required to transform from a militant organization to a political party and the credible political competition this creates for existing parties increases the incentives for the central government to provide public goods to PVLs to expedite reconstruction. Using geo-referenced conflict event data as well as data on night-time light emissions collected from satellite imagery, I find strong support for my hypotheses. I also find that ideological rebel participation in elections is negatively associated with post-war growth in non-PVLs. The results highlight some important trade-offs for designers of post-war institutions as well as the perils of the ecological inference fallacy in the context of economic reconstruction after civil war.


The literature on violence against civilians in armed conflict has focused primarily on the decision for actors to target civilians, separating conflict actors into those groups who do and those who do not engage in the behavior. Further research has disaggregated this violence by tactic focusing, for example, on those groups which employ sexual violence. Research on civilian victimization conceptualizes civilian violations and the presence or absence of a given form of behavior however, there has been less attention to the scope and scale of victimization. In this article we demonstrate the importance of theorizing civilian victimization events, particularly those mass atrocity events which involve high numbers of civilian casualties. We demonstrate that these mass atrocity events are conceptually distinct from events which kill a lower number of civilians and likely follow a different logic from lower-scale events. Our findings suggest a new category of victimization advancing our study of violence against civilians, its causes, and its prevention.


The uniqueness of the global jihadist network, and the competition for allies between Al-Qaeda and ISIS, begs several important yet empirically understudied questions. Why have affiliations and alliances formed between some terrorist groups and not others, and how have these processes changed over time? Are there important differences between the Al-Qaeda and ISIS networks that help explain the structural trajectory of the global Islamist network, and how does the placement of a group within the network affect its behavior? To answer these questions, we study the ISIS and Al-Qaeda alliance networks using temporal exponential graph models (TERGMs) as well as study whether nodal placement in different network substructures affects the lethality of each group. We find strong evidence that the global Islamist network should not be treated as a monolith, as the structure and formative processes of the Al-Qaeda and ISIS networks are empirically distinct. We also find that organizations in triadic relationships within these networks commit more attacks and are more lethal than groups that are not. Interestingly, we also find no evidence that the number of alliances or centrality of groups explains patterns of terrorist attack behavior. Our results have important implications for understanding the development of global terrorist networks.

Research in Progress