Strategic Reporting: A Formal Model of Biases in Conflict Events

• In Event: Domestic Origins of Conflict

Abstract

Conflict events data represent a significant share of datasets used in research on civil war, insurgency, terrorism, and peace studies. These data are often sourced from either the warring parties themselves or investigating NGOs. Potential biases in these data may arise for two reasons: first, the parties may not want to disclose their uses of violence when doing so decreases their popular support and second, NGOs incur investigative costs. In this paper, we use a game-theoretic disclosure model to elucidate how these strategic incentives affect the reporting of conflict events and how they create or undermine reporting biases. In the model, governments have private information and their reports are initially unverifiable but can be verified via NGO investigations. We rely on two disparate, illustrative cases to demonstrate plausibility of the assumptions and mechanisms of the model. Counter to intuition, we find that as NGOs becomes less effective in investigating conflict events, NGO event-provided data becomes less biased relative to government-provided data. Our findings have implications for how researchers use conflict events data and, more broadly, our understanding of the micro-foundations of violet conflict and its aftermath.

Authors

• Michael Gibilisco, California Institute of Technology

• Jessica Steinberg, Indiana University- International Studies