Colaresi, Michael & Mahmood, Zuhaib. (2017). Do the robot: Lessons from machine learning to improve conflict forecasting. Journal of Peace Research. 54(2): 193-214. Link.
- Winner: 2017 best visualization award, Journal of Peace Research
"UN Security Council action in interstate crises: A reassessment and a path forward"
- Working paper (comments welcome)
- Abstract: In this research note, I examine the determinants of UN Security Council (UNSC) involvement in interstate crises. Previous research has focused on two sets of features: one where its behavior is driven by the interests of the 5 major powers, and one where its behavior is driven by the institution’s mandate and the norms it seeks to protect. This paper will build upon an analysis by Beardsley and Schmidt (2012) to evaluate the predictive power of the two models out-of-sample. The results are consistent with the finding that the UN’s organizational mission generally predicts UNSC actions better than major power interests. However, these results are driven by the highest levels of action at the UNSC, such as peacekeeping. Performance is substantially weaker when predicting diplomatic action, such as statements of condemnation. Finally, variables such as requests for assistance add substantial leverage to predicting these middle categories. These results reveal a deficit in current empirical knowledge of the UN’s role as a vehicle for diplomacy and communication, relative to that of costlier actions. This suggests a path forward for scholars to understand how and why these less costly, diplomatic behaviors occur at the UN–and their place in international politics more broadly.
"Babbling or communicating? A research note assessing the informative value of speeches delivered at the UN General Assembly via voting patterns on resolutions"
- Paper: APSA (2018). Do not cite; comments welcome.
- Abstract: The conventional wisdom in international relations holds that foreign policy speech is largely cheap talk, and by extension cannot be trusted to contain useful information about the underlying interests of states. Yet little empirical research to date has directly scrutinized this claim, largely due to a dearth of large-scale empirical data on speech and state interests. In this paper, I propose a theory of foreign policy speech at the United Nations General Assembly, and test this “cheap talk” hypothesis using new data on the lexical content of foreign policy speech at the UNGA to predict voting behavior on UNGA resolutions. The results of this analysis do not support the “babbling” hypothesis, suggesting that there may be systematic information about voting behavior in speech. Additional analysis shows that the informative value of speech increases as the length of speeches increases, and that the use of speech to compensate for positions is most likely when speaking on less controversial issues, whereas the use of speech to reinforce positions is most likely when speaking on the most controversial issues. This is consistent with the proposed theory that the content of speech is strategically contingent on the risks of aligning with opposing countries. This paper has implications for how scholars understand the informative value of speech, its use to communicate foreign policy alignment patterns, and what this means for its role in diplomacy and cooperation more broadly.
- Presentations: ISA (2016, JSS)
(With Michael Colaresi). "Tracking state preferences in high resolution with UN Security Council Speeches"
- Presentations: ISA (2016); Peace Science (2016, poster)
- Poster presented at PSS 2016
Ongoing Data project
(In conjunction with the Social Science Data Analytics initiative at Michigan State University)
- Indexing, parsing, and analysis of the complete record of UNSC (1989-2015) and UNGA (1950-2015) speeches;
- Indexing and parsing of the complete record of UNSC institutional documents (1994-2015) and resolutions (1945-2015).