How do states implement political power-sharing settlements following civil war?  Prior scholars emphasize agreements that establish permanent consociational institutions, preventing renewed violence by dividing control of state power.  Using a corpus of 316 peace agreements since 1975, I identify a separate form of power-sharing: temporary unity governments appointed by leaders of the former armed groups.  Despite a lack of scholarly attention, over 80% of all power-sharing agreements are temporary, rather than institutional.  I illustrate the importance of explicit theorizing about the effects of temporary power-sharing using a novel theory of how rebel leaders use their appointment power.  I argue that these leaders promise followers post-war government positions to enforce group cohesion and loyalty during war-ending negotiations. I use panel causal inference techniques to show that rebel groups are less likely to fragment if offered a temporary unity government.

Recent research shows that the predominant form of power-sharing implemented after civil war is short-term unity governments where appointments are divided between former belligerents. These governments are designed to implement a post-war settlement before giving way to elected leaders. How do these temporary power-sharing governments affect subsequent electoral transitions? Post-war unity governments have significant discretion to determine the timing of the first post-war election, and often delay these due to ostensible concerns over electoral violence or incomplete electoral reforms. However, I argue that transitional leaders also have incentives to prevent elections to retain power. Using new data on the forms of post-war power-sharing, coupled with data on post-war election timing and  results, I show that temporary unity governments are less likely to hold subsequent elections than other postwar regimes. I argue that this throws into question the impact of power-sharing on a state’s long-term democratic development.

Spatially and temporally disaggregated political violence event data enable micro-level research on core peace studies topics such as civil war, the repression-dissent nexus, peacekeeping effectiveness, and terrorism. These event data sets are drawn largely or exclusively from news media sources, and, as a result, suffer from a range of biases.  Events are less likely to be recorded if they are less violent, or occur in non-Western locations, rural areas, or states with media censorship regimes.  Proposed solutions, such as supplementing event data with media image databases, surveying local journalists or civil society members, or relying on high quality data from actors such as UN Missions, either suffer the same biases or are impossible to scale.  Instead, we pilot the use of remote sensing data to overcome the inherent biases in news-based event data in the ongoing war in Gaza.  First, we train computer vision models on high-quality, but infrequent, building damage annotations from UNOSAT.  Second, we use these models to predict damage from new, unlabeled, imagery.  Finally, using spatial econometrics, we assess the relationship between event data and our damage predictions.  We find that building damage better captures the geographic dimension of political violence.

Ratio variables, such as defense burden, relative power, or per capita wealth and spending, are frequently used in international relations research. Existing work uses restricted ratio regression specifications that constrain certain coefficients to zero and yield biased estimates when these assumptions fail.  We provide general ratio independent and dependent variable specifications and argue that authors should test their assumptions using these models. Ratio independent variables are an interaction between the numerator and the reciprocal of the denominator. These constituent terms must be included to produce unbiased estimates unless they have no effect in the absence of the other term. Further, ratio independent variables are used to estimate conditional relationships that can be obtained using linear interactions in the absence of reasons to assume non-linearity of the marginal effects.  Second, unbiased estimation with a ratio dependent variable requires all right hand side terms, including the intercept, be divided by the same denominator.  We highlight the pitfalls of common restricted specifications using Monte Carlo simulations. We replicate two recent publications on power parity and defense burden that use restricted ratio models and show that the substantive conclusions change when the specifications are corrected.

Security interest similarity is an important concept in international relations. The most common measures of this concept focus on the similarity of alliance portfolios. Extant alliance portfolio similarity measures are useful but also have some limitations. First, interest similarity may be signaled without alliances. Indeed, alliances may be most likely when there is uncertainty about overlapping interests. Second, there are significant differences within alliance types based on degree of institutionalization, differences unaccounted for by extant similarity measures. Perhaps because of these conceptual limitations, extant alliance portfolio similarity measures fail a basic validity test: they suggest that rivals have closer security interests than non-rivals. We propose a new measure of security interest similarity based on arms transfers. Arms are the foundation of defense. Allies without arms are of little security help, but arms without allies are often beneficial. We show that our measure of security interest similarity, based on arms transfers, has more surface, content, and context validity than existing measures based on alliances. Arms portfolio similarity shows that rivals have less security interest similarity than non-rivals. Arms portfolio similarity also substantially outperforms existing measures in explaining militarized conflict and war joining.