Modeling Dynamic Violence: Integrating Events, Data Analysis and Agent-Based Modeling

Abstract

Which actions by governments stoke or pacify an insurgency? Scholarly research on the topic has often been relegated to the study of this question at the country level, comparing across large units and rarely looking inside the state. Our research focuses on the primary actors in a contest for authority within a state: the government, dissidents, and the population. In contrast to much previous work, we tackle the difficult question of how population dynamics affect the rise and fall of insurgency. We investigate the question in the context of India and its states and territories. India is particularly well suited to this research as it presently experiences terrorism, insurgency, ethnic conflict, riots and other actions that threaten the stability of the state. Using an agent-based model (ABM), geographic information systems (GIS), data on public sentiment, and events data, we address this question from a multidisciplinary approach. The agent-based model formalizes the interactions of states, dissidents, and the population, the GIS framework allows for actual demographic and geographic information to influence this interaction, and the events data and sentiment data allow us to test empirical implications from the model directly. As the NSF-Minerva grant is midstream, this paper reports initial work to couple the ABM and empirical analysis. We expect the results to have important implications for the study of political violence, order, and state-building. The approach is policy relevant, furthermore, and can be adapted to other regions and countries.

Keywords: terrorism, violence

JEL Classification: D74, H56