"Bridging Levels of Analysis: Selection as a Multi-level Process"
(with Mark Nieman and Doug Gibler). Download
Abstract: Many dyadic processes are nested within broader system-level structures, yet such dependency is often ignored in empirical research. As a result, empirical estimates of observable events data are prone to inferential problems associated with non-random sample selection. We demonstrate that this type of multi-level data generating process (DGP) is equivalent to a selection DGP and can be modeled using known selection models. Using Monte Carlo simulations from a multi-level DGP, we show that a selection model outperforms other commonly employed estimators, including random effects models; the multi-level structure of the data also helps meet the exclusion restriction. To further illustrate the importance of the proposed modeling approach, we replicate two prominent empirical studies of government-opposition behavior---a model of civilian protest outcomes and estimates of civilian killing by insurgent groups---and demonstrate that structural selection affects many of the inferences we draw from the observable data.
"Network Analysis Using a Local Structure Graph Model: Application to Alliance Formation"
(with Mark S. Kaiser). Download
Abstract: We introduce a local structure graph model (LSGM)---a class of statistical estimators that allows for modeling a number of theoretical processes among network edges. LSGMs are especially relevant to political science, as a large number of political networks are outcomes of edge- rather than node-level political processes. Formation of political coalitions and voting blocks, balancing and bandwagoning, policy learning, imitation, diffusion, and tipping-point dynamics and cascade effects are all processes, theoretical understanding of which require focusing on relationships among edges within networks, rather than nodes. Central theories of coalition formation, for example, emphasize that edges form in response to formation of other edges, both within ideological proximity (bandwagoning) and at the opposite end of the ideological spectrum (balancing). Such processes, however, are not easily modeled using the traditional network approach of treating actors (legislators or international states) as network nodes and alliances among them as edges. Any network processes, in which network edges form in response to formation or characteristics of other edges are best modeled using a ``second-degree'' network, in which an edge (e.g., an alliance) is thought of as a node, and a relationship among edges (e.g., belonging to the same neighborhood) are treated as edges. We demonstrate the properties of LSGMs using Monte Carlo simulations and an empirical application to the international alliance network.
"Bankrolling Repression? Modeling Third-Party Influence on Protests and Repression." (with Elena Labzina). Download
Abstract: Game theoretic models of protester--government interactions suggest that governments should repress weak protesters, while negotiating with the strong. In reality, however, we often observe seemingly off-equilibrium outcomes, such as a government's failure to make concessions, despite its inability to successfully repress the protests. In 2013-2014 Ukrainian crisis, for example, President Yanukovich failed to reach an acceptable settlement with the protesters, opting for repression and, when unsuccessful, fleeing the country. We explore the conditions under which a government opts for repression, even despite a high risk losing office. This becomes an equilibrium outcome for governments that rely on support of outside third parties that have a stake in the crisis outcome. We model the interaction between the government, the protesters, and the third-party as a three-player signaling game with incomplete information. In contrast to the bulk of game-theoretic literature that treats violence between the government and the protesters as an off-equilibrium outcome, our model identifies reasonable conditions, under which protests, repression, and even leader removal are part of the game equilibrium. This equilibrium's existence, in particular, depends on the third party's concern with immediate power gains vs. long-term legitimacy. We further demonstrate that our model provides a unifying theoretical framework for explaining a number of known empirical regularities and provides novel testable predictions on leader-specific outcomes, the link between repression and domestic institutions, and violent vs. peaceful protests.
"Applying Network Theory to International Processes: Application to Indirect Trade" Download
Abstract: The study of international relations (IR) has derived great benefits from recent advances in network analysis. Fast-paced progress in statistical modeling of networks within the study of IR, however, has not been matched by equal advances in theoretical understanding in outcomes related to such international networks as the trade, sanctions, or alliance networks. This is especially true for networks characterized by higher-order network relationships (e.g., triads, 2-stars, 4-cycles). This paper takes a first step towards a richer theoretical understanding of such complexities, by focusing on one of the more commonly used measures of dependencies---transitivity, or actors' tendency to form triangles. The paper explores the possible causal processes behind the formation of triangles in the international trade network and assesses the measurement validity of the commonly used transitivity measure as a proxy for the posited causal mechanisms. I do so by comparing the estimates associated with the transitivity measure of indirect trade flows between states to a corresponding instrumental variable measure.
"Autocratic Regimes and Escalation in International Crises: A Strategic Model." Download
Abstract: What is the relationship between authoritarian domestic institutions and escalation of threat in military disputes? Contrary to the conventional view of personalist regimes as polar opposites of democracies, I explain and demonstrate that personalist challengers may act similar to democracies, albeit for different reasons. Personalist and oligarchic targets, in the meantime, act both distinct from democracies and each other. These theoretical and empirical insights become apparent as a result of unpacking the concept of audience costs into two related but distinct categories: inconsistency costs and belligerence costs. The theoretical predictions are derived from a modified version of the bargaining model of war that explicitly accounts for these two types of costs. These predictions are tested with a strategic logit estimator using newly available incident-level dispute data that allow for a more precise measure of escalation. The results help enhance our understanding of crises escalation strategies of different types of authoritarian regimes.
"Autocratic Regimes and Diversionary Uses of Force." (with Brian Lai and Sara McLaughlin Mitchell). Download
Abstract: This paper analyzes the variance in autocratic regimes’ propensity to use force for diversionary purposes while controlling for states’ opportunities to use force. Theoretically, we expect military regimes to use diversionary force more frequently than party regimes and for strongman regimes that combine characteristics of military control and personalist leaders to be most belligerent in the face of domestic turmoil. Empirical analyses of politically relevant directed dyads from 1960-2001 provide strong support to the theory, showing that military regimes and strongman regimes are much more likely than other autocratic regimes to initiate the use of force when inflation is high. However, these effects are conditioned by opportunities for states to use force, as diversionary motives manifest themselves most clearly in the context of interstate rivalry.
"Pinning One’s Hopes on a Flag? The Effect of Patriotic Symbols in Positive and Negative Candidate Assessments." (with Kyle Mattes). Download
Abstract. This paper aims at establishing a clearer connection between the use of national symbols, voters’ reflexive judgments of political candidates, and election results. We presented research participants with briefly shown images of unfamiliar political candidates who ran against each other in real elections, varying which of the two candidates was wearing a flag lapel pin. We asked participants to make trait judgments based solely on viewing the photographs. We found that competence, threat, simulated vote, and patriotism judgments in the laboratory correlated to real world election outcomes. We also found that adding a flag lapel pin to a losing candidate’s picture made the candidate actually look more threating. Furthermore, candidates with the lapel pin appeared more competent and more electable whenever the lapel pin made them seem more patriotic. Our findings show that it may be difficult for otherwise undesirable candidates to change voters’ first impressions of them by juxtaposing their images with patriotic symbols.