Research
Peer-Reviewed Publications
Racial Tropes in the Foreign Policy Bureaucracy: A Computational Text Analysis (with Austin Carson and Maya Van Nuys)
Forthcoming, International Organization. Replication materials here.
Abstract: How do racial stereotypes affect perceptions in foreign policy? Race and racism as topics have long been marginalized in the study of International Relations (IR) but are receiving renewed attention. This article assesses the role of implicit racial bias in internal, originally-classified assessments by the United States foreign policy bureaucracy during the Cold War. We use a combination of dictionary-based and supervised machine learning techniques to identify the presence of four racial tropes in a unique corpus of intelligence documents: almost 5,000 President's Daily Briefs (PDBs) given to Kennedy, Johnson, Nixon, and Ford. We argue and find that entries about countries that the US deemed as “racialized Others” -- specifically, countries in the Global South, newly independent states, and some specific regional groupings -- feature an especially high number of racial tropes. Entries about foreign developments in these places are more likely to feature interpretations which infantilize, invoke animal-based analogies, and imply irrationality and belligerence. This association holds even when accounting for the presence of conflict, the regime type of countries being analyzed, the invocation of leaders, and the topics being discussed. The article makes two primary contributions. First, it adds to the revival of attention to race but gives special emphasis to implicit racialized thinking and its appearance in bureaucratic settings. Second, we show the promise of new methodological tools for identifying racial and other forms of implicit bias in foreign policy texts.
Advisers and Aggregation in Foreign Policy Decision-Making (with Tyler Jost, Joshua D. Kertzer, and Robert Schub)
In International Organization. Article and link to replication materials here.
Abstract: Do advisers affect foreign policy and, if so, how? Recent scholarship on elite decision-making prioritizes leaders and the institutions that surround them, rather than the dispositions of advisers themselves. We argue that despite the hierarchical nature of foreign policy decision-making, advisers’ predispositions towards the use of force shape state behavior through the counsel advisers provide in deliberations. We test our argument by introducing three original datasets, including an original dataset of 2,685 foreign policy deliberations between US presidents and their advisers from 1947 to 1988. Applying a novel machine learning approach to estimate the hawkishness of 1,134 Cold War-era foreign policy decision-makers, we show that adviser-level hawkishness affects both the counsel advisers provide in deliberations, and the decisions the leader makes: conflictual policy choices grow more likely as hawks increasingly dominate the debate, even when accounting for leader dispositions. These results enrich our understanding of international conflict by demonstrating how advisers' dispositions, which aggregate via deliberation, systematically shape the choices leaders make.
Painful Words: The Effect of Battlefield Activity on Conflict Negotiation Behavior
In Journal of Conflict Research. Article and link to replication materials here.
Abstract: How does battlefield activity affect belligerents' behavior during wartime negotiations? While scholars have studied when and why warring parties choose to negotiate, few insights explain what negotiators do once seated at the table. I argue that actors engage in obstinate negotiation behavior to signal resolve when undergoing contentious and indeterminate hostilities. I explore this claim by analyzing all negotiation transcripts and associated daily military operations reports from the Korean War. Using text-based, machine learning, and statistical methods, I show that high levels of movement or casualties in isolation produce clear information on future trends, thus yielding more substantive negotiations, while more turbulent activity featuring high movement and casualties in tandem produces cynical negotiations. Moving past contemporary literature, this study explores micro-level dynamics of conflict and diplomacy, builds a theoretical bridge between two perennial views of negotiation, and provides a framework for studying war by applying computational methods to archival documents.
Speaking with One Voice: Coalitions and Wartime Diplomacy
In Journal of Strategic Studies. Article and link to replication materials here.
Abstract: When and why do countries in a wartime coalition engage in diplomacy during hostilities? This paper establishes a theoretical framework of coalitional diplomacy that highlights each member’s private costs and benefits to fighting or seeking a negotiated exit. I argue that the propensity for coalition members to engage in negotiations is a function of the coalition’s balance of military contributions, as well as the coalition’s battlefield successes and failures. Evidence supporting these claims stem from a large-scale quantitative analysis of two centuries of interstate wars, as well as a close study of the Allies in the Crimean War.
The Credibility of Public and Private Threats: A Document-Based Approach (with Azusa Katagiri)
In American Political Science Review. Article and link to replication materials here.)
Contemporary studies of conflict have adopted approaches that minimize the importance of negotiation during war or treat it as a constant and mechanical activity. This is tightly related to the lack of systematic data that tracks and illustrates the complex nature of wartime diplomacy. I address these issues by creating and exploring a new daily-level dataset of negotiations in all interstate wars from 1816 to present. I find strong indications that post-1945 wars feature more frequent negotiations and that these negotiations are far less predictive of war termination. Evidence suggests that increased international pressures for peace and stability after World War II, especially emanating from nuclear weapons and international alliances, account for this trend. These original data and insights establish a dynamic research agenda that enables a more policy-relevant study of conflict management, highlights a historical angle to conflict resolution, and speaks to the utility of viewing diplomacy an essential dimension to understanding war.
Talking While Fighting: Understanding the Role of Wartime Negotiation
In International Organization. Link to article and replication materials here.
Contemporary studies of conflict have adopted approaches that minimize the importance of negotiation during war or treat it as a constant and mechanical activity. This is tightly related to the lack of systematic data that tracks and illustrates the complex nature of wartime diplomacy. I address these issues by creating and exploring a new daily-level dataset of negotiations in all interstate wars from 1816 to present. I find strong indications that post-1945 wars feature more frequent negotiations and that these negotiations are far less predictive of war termination. Evidence suggests that increased international pressures for peace and stability after World War II, especially emanating from nuclear weapons and international alliances, account for this trend. These original data and insights establish a dynamic research agenda that enables a more policy-relevant study of conflict management, highlights a historical angle to conflict resolution, and speaks to the utility of viewing diplomacy an essential dimension to understanding war.
Interstate War Battle Dataset (1823-2003)
In Journal of Peace Research. Link to article here. Link to dataset here.
Extant scholarship on interstate war and conflict resolution predominantly utilizes formal models, case studies, and statistical models with wars as the unit of analysis to assess the impact of battlefield activity on war duration and termination. As such, longstanding views of war have not been tested systematically using intra-conflict measures, and deeper studies of war dynamics have also been hampered. I address these gaps by creating and introducing the Interstate War Battle (IWB) dataset, which captures the outcomes and dates of 1,708 battles across 97 interstate wars since 1823. This paper describes the sources used to create these data, provides definitions, and presents descriptive statistics for the basic battle data and several daily-level measures constructed from them. I then use the data to test the implications of two major theoretical perspectives on conflict termination: the informational view, which emphasizes convergence in beliefs through battlefield activity; and Zartman's ripeness theory, which highlights costly stalemates in fighting. I find suggestive evidence for informational views and little support for ripeness theory: New battlefield outcomes promote negotiated settlements, while battlefield stagnation undermines them. The IWB dataset has significant implications, highlights future research topics, and motivates a renewed research agenda on the empirical study of conflict. Abstract
Working Papers
Racial Inequality in War (with Connor Huff and Robert Schub)
Latest draft here.
Abstract: How does racial inequality shape who dies in war? Focusing on the era of US military segregation, we argue that discriminatory societal institutions and prejudicial attitudes combined to reduce commanders’ beliefs about Black soldiers’ combat effectiveness. These biased assessments decreased the likelihood that Black soldiers were assigned combat occupational specialties, and that Black combat units received key frontline assignments. However, commanders’ biases also created a desire to preserve White lives. Accordingly, we expect Black soldiers received worse support. These choices shaped soldiers’ risk of death. Analyzing the case of WWI, we leverage data on over 40,000 infantry fatalities and show that White units incurred four times as many combat fatalities as comparable Black units. However, holding fixed exposure to combat, Black units suffered higher levels of non-combat deaths. Commanders thus deemed Black soldiers insufficiently qualified to fight as equals, but sufficiently expendable to die in war's least consequential conditions.
Language and the Emergence of Bureaucracy in Foreign Policy (with Austin Carson)
Draft available upon request.
Abstract: This article uses novel computational techniques to assess how the diversity and complexity of language evolves in classified settings within the foreign policy bureaucracy. Analyzing over 4,900 declassified intelligence summaries prepared for the U.S. president, we find that over time, foreign policy events were described with increasingly homogeneous and technical language. Notably, such patterns do not appear in contemporaneous presidential speeches, New York Times articles, or declassified State Department materials. We argue that this change reflects a distinctive moment of organizational specialization and centralization in the American intelligence bureaucracy during the 1960s and 1970s. We test this explanation with a quantitative measure of CIA organizational change and qualitative evidence on evolving CIA writing guidance. Even after controlling for presidents, external events, and other factors, we find that broader bureaucratic changes forged a more standardized and technical writing style, altering the language the bureaucracy used to make sense of foreign developments.
Does Intelligence Pay? Assessing Information Advantages in Declassified Daily Intelligence Briefings (with Austin Carson)
Draft available upon request.
Abstract: Does intelligence reporting provide novel and more useful information to leaders than what is available in mainstream news reports? We assess this question through a same-day comparative analysis of almost 5,000 President’s Daily Briefs (PDBs) and almost 2 million full-text articles on foreign affairs in the New York Times. We use a novel approach for systematically comparing an originally highly-classified knowledge source with a public knowledge source to measure informational differences and informational advantages for leaders. We find find considerable and interesting divergence in the slate of countries are covered in each source. To assess informational advantage, we analyze relative success in anticipating coups. Intelligence material, we show, is more likely to mention unrest in a country as well as key facts related to an upcoming coup more quickly than traditional news reporting. The article contributes a novel methodological approach to measuring private information. It also offers a rare test of whether intelligence pays--that is, the value added of massive investments in intelligence bureaucracies made by modern states.
Bureaucratic Protection and Foreign Policy Information Provision in China (with Tyler Jost)
Draft available upon request.
Abstract: When do foreign policy bureaucracies in authoritarian regimes provide professional information and when do they withhold it? We argue that bureaucratic protection from arbitrary punishment shapes information provision. Bureaucrats are more likely to provide substantive and original information pertaining to their functional position, what we term professional information, when leaders use power sharing institutions to tie their hands from punishing bureaucrats if the leader arbitrarily dislikes their information. Applying a range of supervised and automated text analysis methods to an original set of over 6,000 archival documents from the People's Republic of China during the Cold War, including nearly 2,000 cables from the Ministry of Foreign Affairs, we show that foreign policy reporting exhibited more substantive focus on international cooperation and greater originality when Chinese diplomats enjoyed higher levels of protection.
Looking for Trouble: Analyzing Search Engine Data during International Crises (with Lizhi Liu)
Latest draft here.
Abstract: Many theories of crisis bargaining and conflict are predicated on the idea that actors constantly observe and process information obtained from words and deeds. However, very few studies have explicitly evaluated whether individuals inform themselves about developments in the midst of crisis, and little thought has been put into the specific mechanisms that would motivate such behavior. This paper uses search engine data from Baidu (China's largest service) to address these issues through the lens of the current South China Sea dispute. Baidu search data provides unique access to daily, province-level, and unbiased statistics on behavior in an illiberal and autocratic country. We use this resource to show that citizens indeed seek information in response to crisis events, and that these patterns of information seeking are best explained as attempts to resolve uncertainty and anxiety that the South China Sea dispute and its related events imply.
Flipping Channels: Diplomatic Signals in Front of Multiple Audiences (with Azusa Katagiri)
Latest draft here.
Abstract: How do states choose to use public or private signals to communicate their intentions? Formal literature, particularly involving notions of costly signaling, has focused on differences in these two channels primarily in terms of costs to reneging on any statements made. Moreover, a great deal of work has only considered signaling interactions between two actors in a vacuum. We contend that the presence of third parties that are involved in disputes with a specific sender will affect the sender’s decision to direct more of their signals of resolve through public or private channels. To explore this claim, we conduct a computational analysis of 19,000 diplomatic statements from the Berlin Crisis of 1958 to 1963, which allows unprecedented access to public and private signals exchanged between the United States and Soviet Union. Through the use of structural topic models and text similarity measures, we find that the Soviets send more signals, which are both more aggressive and focused in their content, via private channels to the US when embroiled in direct disputes with the US. However, these signals shift over to public channels when Soviets are involved in multiple disputes with states around the world. Our contribution is two-fold. Substantively, we outline new mechanisms to understand how different forms of diplomatic communication differ beyond their costliness. Methodologically, we provide a unique example of how statistical learning methods can be applied to study the strategic logic of diplomacy.
Taking Responsibility: When and Why Terrorists Claim Credit for Attacks
Latest draft here.
Abstract: Despite general academic consensus on terrorism being a strategic and communicative act, most extant work has only analyzed these features in relation to the actual attack itself. However, much of the communicative and strategic weight may come from a public claim of responsibility for the attack (or lack thereof). A terrorist attack is fundamentally an intimidating costly signal that foments uncertainty; claims of credit can naturally be viewed as a manner in which to augment the costliness and uncertainty of this base signal. Using this insight, I develop a series of hypotheses to test several well-established and costly qualitative theoretical arguments in the literature. Focus is placed on Kydd and Walter’s (2006) important work on extremists’ strategies. Using econometric analysis on the Global Terrorism Database and case studies of three extremist organizations in Pakistan, I find that credit-claiming is indeed consistent with these strategies: While a dominant strategy of intimidation keeps claim rates low, extremists are more likely to claim responsibility for attacks that involve high costs (suicide and casualties), institutionally constrained states (democracy), and competitive environments. Attempts to sabotage political moderation serve to suppress claims. Moreover, groups with limited and specific aims, such as separatist organizations, are likely to have a higher claim rate than organizations with more sweeping and amorphous objectives. These findings not only tap into a largely overlooked dimension of violent political communication, but function as a useful test of extant theories in the terrorism literature that have thus far been evaluated using very small-N analysis.
Explaining Military Relapse (with Jonathan Chu)
Latest draft here. Please do not cite without authors’ permission.
Abstract: Renewed interest in authoritarian rule has sparked a groundswell of literature on regime change and consolidation. Most of this work does not explicitly mention or address the tendency for military regimes to transition to multi-party civilian rule, only to quickly revert back to military control. We define this phenomenon in which a military regime transfers power to civilian leaders only to retake power \military relapse." By integrating insights from contemporary work on autocracies with older literature on the military and civil-military relations, we develop a theory to explain military relapses. We find that characteristics unique to the military motivate their exit from power. In establishing the civilian regime to follow, the balance between civilian and military prerogatives is determined by the relative levels of unity within each party, which also influences their mutual levels of trust. These factors, combined with the potential costs for the military of reclaiming power, explain the likelihood of and trends in military relapse. Case studies of multiple civilian-military interactions in Myanmar and Peru support this theory.
Additional Research
Quantitative Archival Analysis in Strategic Studies: The World War II Pacific Submarine Campaign (with Stephen D. Biddle and Laura Resnick Samotin)
Empirical Studies of Conflict Working Paper #38. Link to paper here.
Abstract: Nonmaterial variables are important in war, but can be difficult to study quantitatively given their absence from standard datasets. We present a method for overcoming this challenge by using optical character recognition (OCR) software and a process of crowdsourced data conversion to code narrative archival records at scale, producing large-n data on nonmaterial variables such as force employment, combat motivation, leadership, culture, initiative, organizational structure, or military doctrine. We illustrate this method by coding a new dataset on the determinants of success and failure in submarine warfare, and demonstrate its utility with a preliminary analysis of the resulting data. Tentative findings suggest that nonmaterial variables are likely to matter for naval combat in ways that would render analyses based on materiel alone suspect, but that not all nonmaterial traits matter equally. Many believe that initiative and leadership, in particular, are important determinants of success in combat – and indeed they may well be. But our initial analysis suggests that for submarine combat in World War II they may have been less important than the intellectual and personal qualities that produced strong performance at the Naval Academy.
Understanding Risk and Resilience to Violent Conflicts (with Manu Singh, Jacob N. Shapiro, and Benjamin Crisman)
Report, Empirical Studies of Conflict Project, Princeton University. 2017. Link to report here.
First paragraph: This research aims to establish potential opportunities for policy intervention by asking the following question: if we could predict conflict 5-years out, what would separate the predictable failures from the unexpected successes (i.e. places where conflict was expected but did not happen) and what would separate the predictable successes from the unexpected failures (i.e. places where conflict was not expected that experience it). The idea is to move beyond standard conflict prediction variables to identify previously-unrecognized opportunities for preventive action within a time-frame sufficiently long for significant action by the UN, World Bank, and other international organizations. Put differently, our goal is to provide policy recommendations by examining what led to unexpected resilience to conflict in some countries and what happened in places with low risk based on observable characteristics.
Autocracies of the World, 1950-2012 (with Beatriz Magaloni and Jonathan Chu)
Link to data and codebook here.
A regime dataset that classifies all country-years from 1950 to 2012 by different forms of autocratic rule. Introduces a novel form of measuring personalist rule across all autocracies and offers several corrections to extant regime classification data.