- Major Powers and Militarized Conflict
Forthcoming at the Journal of Conflict Resolution
(with William Reed and Carla Martinez Machain)
This paper attempts to answer the question of why major powers engage in more active foreign policy behaviors than minor powers. It does so by comparing two explanations for the increased conflict propensity of major powers. The first explanation focuses on major powers' observable capabilities, while the second stresses their different behavior. We incorporate both into an ultimatum model of conflict in which a state's cost of conflict consists of both observable and behavioral components. Using data from the period from 1870 to 2001, we empirically illustrate the observable and behavioral differences between major and minor powers. We then utilize a decomposition model to assess the relative significance of the two explanations. The results suggest that most of the difference in conflict propensity between major and minor powers can be attributed to observable differences.
- Manuscript in PDF
- Replication materials (IQSS Dataverse)
- Decomposing the Relationship Between Contiguity and Militarized Conflict
American Journal of Political Science Vol. 54, No. 1 (Jan., 2010) (with William Reed)
It is well known that the majority of militarized conflicts and wars have been fought by neighbors. Yet, much remains to be learned about the relationship between shared borders and militarized conflict. This paper decomposes the effects of territorial contiguity into ex ante "observable" and "behavioral" effects. It provides powerful empirical evidence for the claim that although neighbors are more likely to experience conflict because of ex ante differences in observable variables such as economic interdependence, alliance membership, joint democracy, and the balance of military capabilities, most conflicts between neighbors occur because of differences in how neighbors and nonneighbors respond to the observable variables.
- Paper (journal subscription required)
- Replication materials (IQSS Dataverse)
- Procuring Peace after Prolonging War:
International Organizations and the Durations of International Conflict and Post-conflict Peace
- Dissertation Chapter [PDF]
How do international organizations (IOs) influence states' conflict behavior in the absence of centralized enforcement? This study develops and tests an argument about how IO membership affects the enforcement problems states face in the aftermath of militarized conflict. It maintains that joint membership in IOs that explicitly promote peaceful settlement of disputes improves enforcement conditions by increasing the costs of cease-fire violation in the long run. As a result, these IOs make a cease-fire more durable once the disputants agree to stop fighting. However, precisely because they expect longer peace after conflict, the member states have incentive to adopt tougher bargaining positions during conflict, causing a delay in reaching a cease-fire. A survival analysis that recognizes the interdependence between the durations of conflict and subsequent peace demonstrates that IO membership lengthens the durations of conflict and subsequent peace.
- Won the 2012 Dina Zinnes Award for the best graduate student paper presented at 2011 ISA Annual Convention.
- The Strength of Cease-fire Agreements and the Duration of Postwar Peace [Show details]
- Dissertation Chapter [PDF]
Do stronger cease-fire agreements keep peace longer after war? Although there are theoretical reasons to expect that stronger agreements promote durable peace, the extant empirical research provides mixed support for this expectation. This paper reexamines this argument empirically, addressing two inferential problems overlooked in the past studies. First, since the strength of cease-fire agreements is endogenous to the baseline prospect for peace, I employ a copula-based two-stage estimation that explains agreement strength and peace duration jointly. Second, I allow the effect of agreement strength to vary over time. This is important because agreement strength matters little right after the war, for there exists a rough consensus among the ex-belligerents about the likely outcome of a next war. As time passes, however, the effect of agreement strength will start to show because there will be a greater chance that some exogenous shocks distort this consensus. Analyzing the duration of postwar peace from 1914 to 2001, I demonstrate that stronger cease-fire agreements indeed promote more durable peace.
- Estimating Interdependent Durations in Militarized Conflict and Beyond [Show details]
- Dissertation Chapter
This paper introduces a simple method to analyze interdependent duration processes that typically arise in strategic interactions involving two political actors. In analyzing the duration of intra- and interstate conflict, for example, the time until each side decides to terminate violence is strategically dependent on one another. I conduct a series of Monte Carlo analyses that estimate common duration models as well as the proposed strategic duration model. These simulations show that ignoring strategic interdependence can lead to biased parameter estimates, including the inability to recover the true form of duration dependence. The proposed approach is easy to implement, and applicable to many situations that are of interest to political scientists. I provide an empirical application of the method by studying the effect of outside intervention on the duration of civil wars. Contrary to the conventional wisdom that outside interventions prolong the time until a negotiated settlement is reached, my findings suggest that balanced interventions can facilitate a rapid settlement.
- Institutional Opposition, Regime Accountability, and International Conflict
Revise and Resubmit at the Journal of Politics
(with Songying Fang)
- Manuscript in PDF
Can international organizations constrain a leader's behavior during an ongoing military conflict? While existing studies have shown that joint membership in international organizations reduces the likelihood of dispute initiation, the question of whether institutional opposition can exert influence on disputes that have been escalated to military confrontations has yet to be addressed. We develop and test a theory of how domestic politics provides a mechanism through which international organizations can reverse the course of a military crisis. The argument leads to the hypothesis that more responsive regimes are more likely to deescalate military crises when an international organization opposes their actions. We test the hypothesis with an analysis of territorial disputes from 1946 to 1995. We find that while neither institutional opposition nor the degree of democracy independently reduces the tendency for a country to escalate a conflict, the joint effect of the two does.
- Commitment Problems, Territorial Claims, and Third-Party Conflict
(with William Reed)
- Down for revision
Typically, bargaining theories attribute the outbreak of war to dynamic commitment problems or asymmetric information. While numerous studies have investigated the informational explanation for war, less attention has been paid to the commitment problem explanation. This article develops a model of third-party conflict caused by the interaction between asymmetric information and a commitment problem. Hypotheses from the model are empirically assessed with regards to conflict outbreak. The results show that active territorial claims within a dyad form expectations that strategic territory will be transferred, either by force or by peaceful negotiation. These expectations about territorial change can result in a large and rapid shift in power distribution between each of the disputing parties and third parties, which can cause a dynamic commitment problem. Support for this argument is found using data on territorial claims between 1919 and 1995.
- A Solution to the Selection Problem in Models of Government Survival
(with Lanny Martin and Randy Stevenson)
- Down for revision.
Recent theoretical work on coalition politics suggests that government formation and survival are jointly determined outcomes. These arguments imply that the sample of observed governments are nonrandomly selected from the population of potential governments. In this paper, we use a copula-based framework to account for the dependence between these processes. Our results suggest that, by ignoring the selection problem, current approaches to examining government survival substantially overstate the substantive importance of several covariates commonly included in empirical models.
- LATEXresource for Rice students & scholars
- I created some Beamer style files and presentation examples (slide and poster) for Rice students and scholars.
- Link to the RiceBeamer page
- Presenting statistical results
- I ran a workshop for grad students and faculties in the political science department about how to present statistical results using R and Stata in combination. You can download this archived directory (zip, 45 MB) that contains the handout, an accompanying R code, Stata do-file, data, etc., necessary to replicate all the figures and tables I talked about in the workshop. Topics I covered include:
- Producing publication-quality regression tables efficiently (Stata)
- Calculating and plotting substantive effects (R) ··· Example 1, Example 2
- Plotting descriptive statistics (R) ··· Example 3
- Plotting regression coefficients (R) ··· Example 4
- For more of these graphs, go to Using Graphs Instead of Tables website by Jonathan Kastellec and Eduardo Leoni.
- R utilities for clustered standard error
- As a quantitative IR researcher, I have seen many examples utilizing the following two estimation options of Stata -- , robust and , cluster(dyad). These frequently-used options are supposed to give us the "corrected" standard error estimates when observations are suspected to be interdependent within analysis unit (e.g., dyad) and/or distribution is not identical (i.e., errors are heteroskedastic). The good news is that the Stata implementation of these techniques is quite easy. The bad news is that this is a very handy tool for "star-gazing" -- an act of searching for statistical significance (stars!) by trying a bunch of different specifications and statistical patches without paying careful attention to the causal mechanism behind the data-generating process. Robust and/or clustered standard errors sometimes make your p-values much smaller, causing otherwise statistically insignificant coefficients to be very significant. Well, there is nothing inherently wrong about reporting robust standard errors instead of conventional ones. But, you should not use them just because this is the only way to have stars for your favorite variables. And you should not use them without understanding what they really mean.
- So..., what exactly do we mean by standard error being "robust"? In what sense do we say that the estimation uncertainties are "adjusted for possible non-independence within dyad?" And, more (or less) importantly, how do we execute them on R? I suppose some step-by-step examples might help us gain intuition behind these methods, and will get us ready to implement the techniques on our own.
- logit..., cluster(dyad) ··· Example R code to calculate clustered s.e. for a typical logit model of interstate disputes.
- logit..., robust ··· Example R code to calculate robust s.e. for a typical logit model of conflict.
- Sweave output (pdf) of the above two. This is generated by running the Sweave source (.Rnw) on R and then typesetting the tex output (.tex) on LATEX.
- R utilities for selection model estimation
- The R code below estimates the Heckman's probit model in R.
- R utilities for copula estimation
- The R code below contains functions that transform the dependency parameter (θ) of the Frank's bivariate copula into some common measures of correlation. The function frank.rho() outputs the Spearman's ρ and the function frank.tau() outputs the Kendall's τ. The calculation involves numerical integration as a part of the Debye function.