Research

Publications

1. “Winners and Losers in U.S.-China Trade Disputes: A Dynamic Compositional Analysis of Multinational Corporations’ Market Penetration” (with Yoo Sun Jung). 2024. Online first at Social Science Quarterly. [Ungated] [Publisher site] [Appendix] [Replication]

2. “Remittances, Terrorism, and Democracy” (with Casey Crisman-Cox). 2023. Online first at Conflict Management and Peace Science. [Ungated] [Publisher site] [Appendix] [Replication]

3. “Signaling Restraints: International Engagement and Rebel Groups’ Commitment to International Law” (with Hyeran Jo, Joshua Alley, and Soren Jordan). 2021. International Interactions 47(5): 928-954. [Ungated] [Publisher site] [Appendix] [Replication]

Working Papers

1. Appeal to the Public: Why Leaders Make Public Threats. [Draft]
   • [Abstract]

     Why do leaders go public? During an armed crisis, leaders have strong concerns about the public's evaluation of their conflict behavior. I argue that leaders issue public threats, especially ambiguous ones, to address these concerns. Using public statements, leaders can provide domestic audiences with information on their progress in the crisis, reframe the issues at stake, and make a justification for the decisions they have made or will make further. Moreover, leaders strategically embrace ambiguity within their public statements to satisfy broader audiences with heterogeneous preferences over how to handle the crisis. I test my argument with a novel measure of the US leaders' perceived public concerns during the Vietnam War. Applying supervised learning methods to the declassified White House documents from 1961 to 1976, I measure the US decision-makers' time-varying concerns about the public's evaluation of their foreign policy. The analysis of the US foreign policy documents finds that leaders are more likely to issue public threats, and they make these threats more ambiguous as they perceive the public to be more concerned about leaders' policy competence in Vietnam. These findings imply that the presence of domestic audiences could undermine the credibility of a state's public threats.
2. Power structure, domestic constraints, and selective targeting.
   • [Abstract]

     Previous studies of interstate conflict suggest that leaders who are accountable to domestic audiences have an incentive to target weaker countries in militarized disputes due to the fear of post-defeat domestic punishment. I argue that the effect of domestic audiences on leaders' decisions vary with the international power structure. Under a bipolar structure, minor powers face great uncertainty about dispute outcomes because fierce superpower competition undermines the importance of the balance of power between the disputants. Moreover, the shared concerns about the total war between the two superpowers lead the minor powers to be ambiguous about the superpowers' commitments about support. The dispute outcome uncertainty discourages the leaders with domestic audiences from making risky decisions, whereas it encourages the relatively unconstrained leaders to adopt reckless policies. The analyses of the initiation of militarized disputes between 1946 and 2010 show that regime types with domestic audiences are less likely than personalist regimes to initiate militarized conflicts against stronger opponents in the Cold War period. However, the difference between each non-personalist regime and personalist regime decreases following the end of superpower competition. This finding implies the diminished role of domestic audiences in constraining state leaders during the post-Cold War era.
3. Does the Sector Matter? Deteriorating Political Relations and Foreign Direct Investment. (with Yoo Sun Jung)
   • [Abstract]

     Do political tensions between states disrupt investment flows? Deteriorating political relations may cause negative attitudes towards foreign firms and their products and even retaliatory economic measures. With a hostile political environment, investors expect firms' performance and potential profits to decrease; thus, they perceive the host country riskier. We argue that rising political risk increases as a function of industry fixed asset intensity associated with the irreversibility of investment. We expect that political tensions deter investment for fixed asset-intensive industries because of a substantial increase in investors' perceived risk, but have no such effect for low fixed asset industries, as the increase is not large enough to alter investors' decisions. We test our argument using data on greenfield FDI projects in 126 developing countries during 2003 - 2019. Our analysis finds that political tensions reduce investment in most industries, but not in low fixed asset industries.

Selected Works in Progress

1. Multinational Firms and the Impact of Trade Disputes on Investment Decisions. (with Erica Owen and Yoo Sun Jung )
   • [Abstract]

     The rise of global value chains (GVCs) is reshaping the political economy of trade in several ways, including the politics of trade disputes. Trade disputes affect access to markets and suppliers in ways that are likely to influence investment decisions. Indeed, recent work examines how multinationals influence the initiation and duration of disputes. Yet we know little about how multinationals respond to trade frictions in a world of GVCs. We offer a theory of international trade and investment that interrelates trade, FDI, global production, and GVC participation. We argue that trade disputes can actually lead to greater inward FDI in respondent countries because a (resolved) trade dispute signals lower barriers to trade. This hurts domestic producers in the respondent and creates opportunities for multinationals in the complainant country. However, we expect that the effect of trade disputes will depend on how and to what extent the industry integrates into GVCs. In particular, we expect the impact of a dispute to be greater where there are more backward linkages, that is, in industries that rely on imported inputs. We use data on dyad-industry level greenfield FDI from FDI markets between 2003 and 2015 to test our hypotheses.
2. Saving Lives: Variety of international efforts to reduce militant violence. (with Hyeran Jo and Lisa Hultman)
   • [Abstract]

     What saves lives in civil conflicts around the world? International actors have tried a variety of measures such as mediation, sanctions, and peacekeeping. What measure works better to reduce violence against civilians by insurgent militant forces and under what circumstances? We hypothesize that the relative efficacy of measures hinges on the balance of power between the government forces and insurgent militant forces. When rebels are militarily weak, sanctions might work better reducing the militant violence against civilians. For medium-strength rebels, peacekeeping might work better. When rebels are strong, mediation might work better in saving civilian lives. We present our theory and test the balance-of-power approach with the monthly data of militant violence of three decades between 1990 and 2020. Our work contributes to the literature on global security governance, providing an integrative view of international measures, beyond the analysis of each measure separately.
3. Intended and Unintended Consequences of International Interventions: Patterns of Militant Violence in the Democratic Republic of Congo. (with Hyeran Jo, Niels Appeldorn, and Yewon Kwon)
   • [Abstract]

     Mediation, peacekeeping, and sanctions – the international community has tried various intervention methods to reduce conflicts around the world. This paper examines the humanitarian consequences of international efforts with a focus on the fluctuation of militant violence against civilians. We argue that external interventions in internal conflicts alter political, military, and economic balance among militant groups, creating both intended and unintended consequences depending on the militant characteristics such as adaptability, co-optation, and rivalry. By altering political balance among militant groups, non-inclusive mediation increases the violence of excluded militant groups, while decreasing the violence of included militant group. By altering military balance among militant groups and creating a security vacuum in one area, forceful peacekeeping might reduce the violence of the targeted group, but at the same time, might increase the violence of the rival group that has been co-opted with host government forces. By altering economic balance and hurting one group more than others economically, sanctions might decrease the violence of the non-adaptable militant group that fails to adjust, but inadvertently increase the violence of adaptable groups that can easily shift their resource bases to other lucrative sources. We test these arguments using the interrupted time-series intervention analysis in the context of the conflict in the Democratic Republic of Congo, one of the longest running civil wars featuring multiple militant groups. Our findings of the differential impacts of intervention measures on a diverse set of militant groups have implications for external intervention in internal conflicts, highlighting the limits and opportunities for global security governance. Outside interventions in internal conflicts alter the political, military, and economic balance, with heavy consequences on the civilian lives in conflict zones.
4. Optimizing Thresholds in Machine Learning for Improved Regression Analysis. (with Jong Woo Jeong)
   • [Abstract]

     The growing trend among political scientists to apply machine learning algorithms for measuring theoretical concepts, and the subsequent use of these measures in regression analysis, highlights the importance of the algorithms' performance in determining measurement accuracy and regression estimator robustness. However, there has been a lack of comprehensive discussion among political scientists on the impact of imperfect machine learning predictions on regression estimators. This article addresses this gap by providing a practical guide for achieving the best measurement quality when applying machine learning algorithms to measuring latent variables. We particularly focus on identifying the most suitable threshold values for classifying the concept of interest. In binary predictions, where machine learning is widely used, such threshold settings are key in determining measurement accuracy and avoiding attenuation biases in regression estimates, especially in the context of complicated time-series cross-sectional data. Our simulations and empirical analyses show that the proposed method significantly reduces biases by adjusting threshold values for classification.
5. A Doubly Robust Difference-in-Differences Estimator for Causal Inference with Time-Series Cross-Sectional Data. (with Thomas Chadefaux)
   • [Abstract]

     This paper addresses the challenge of deriving robust causal effects from time-series cross-sectional (TSCS) data. The task is especially complex with multiple treatment status changes, heterogeneous treatment effects, and unobserved time-varying confounders, leading to increased bias and reduced efficiency. Here, we introduce a novel difference-in-differences (DID) estimator to assess the average treatment effect on the treated (ATT), building upon the principles of doubly robust DID estimation. Our approach involves creating matched sets by pairing each treated observation with control observations from different groups that share an identical treatment history. We employ a combination of propensity score and outcome regression methods, incorporating machine learning algorithms with cross-validation, to calculate both immediate and long-term ATTs. Our simulation and empirical analyses demonstrate the estimator's semi-parametric efficiency and resilience to incorrect model specifications. We also introduce an open-source software package for these methods' implementation.