Learning about Political Violence through Observational Data

Session Submission Type: Full Paper Panel

Session Description

Although political violence occurs in myriad contexts and forms, empirically studying its causes and consequences remains enormously difficult. For ethical and practical reasons, researchers are unable to randomly-assign many pertinent causal variables of interest. This panel brings together studies implementing a diverse set of causally-identified research designs in observational settings, rigorously interrogating questions of interest to scholars of political violence. The studies included in this panel investigate violent conflict in diverse settings: India, Iraq, and Ukraine.

Each paper in this panel applies a distinct causally-identified research design to examine underlying causes and consequences of violence. Revkin et al. leverage a natural experiment in Mosul, Iraq to study how civilian support for insurgents changed across otherwise similar neighborhoods exposed to different levels of collateral damage during the battle to retake Mosul from the Islamic State (IS). Milliff et al. implement a difference-in-differences research design to examine spatial and temporal variation in anti-minority lynching violence in India. Their analysis demonstrates that the strength of majoritarian institutions—rather than state weakness—increased the incidence of lynching. Talibova et al. examine how Ukrainian elites’ social media postings changed in response to the 2022 Russian invasion, combining network analysis and natural language processing. They find that the invasion increased Ukrainian elite cohesion and intensified their links with various international constituencies. Stommes implements a regression discontinuity design to examine whether incumbent representatives’ partisan alignment with a government’s ruling party affected armed political party violence in West Bengal, India. His study draws upon a novel dataset of armed party violence in West Bengal and finds that incumbent alignment with the ruling party significantly increased pre- and post-election violence.

Two discussants will offer comments: Professors Gareth Nellis and Darin Christensen have made important contributions to our understanding of political violence by implementing robust observational research designs, including regression discontinuity designs and other natural experiments. Professor Laia Balcells, whose cutting-edge research on political violence spans across multiple settings and types of violence, will serve as Chair. This panel generates valuable substantive insights and also offers important practical lessons for scholars wishing to study challenging research questions related to political violence. By leveraging natural experiments and other identified designs, this panel provides empirically-robust findings regarding causes and consequences of violent conflict.

Sub Unit

• Division 21: Conflict Processes

Individual Presentations

• Understanding Norm-Based Vigilante Violence - Aidan Milliff, Florida State University; Blair Read

• A Natural Experiment on Civilian Harm in Iraq - Mara Revkin, Duke University; Benjamin Curtiss Krick, Duke University; Jonathan Petkun, Yale Law School

• Armed Political Parties and Their Violence: Evidence from West Bengal, India - Drew Stommes, Yale University

• Social Media Dimension of Political Violence during Russia's Invasion of Ukraine - Roya Talibova, Harvard University; Erik M. Wibbels, University of Pennsylvania; Ernesto F. Calvo, University of Maryland

Chair

• Laia Balcells Georgetown University

Discussants

• Gareth Nellis University of California, San Diego

• Darin Christensen UCLA