How Risky is Nonviolent Dissent? Nonviolent Uprisings and Mass Killings

Abstract

Why do mass killings occur during some popular uprisings but not others? To what degree do characteristics of the uprising itself help to explain (and predict) this variation? In this article, we identify the factors associated with mass killings that occur during and after major episodes of contention (MECs). Drawing on data from 1955-2014, we develop a series of models that (1) identify the structural and campaign-level correlates of mass killings during and after popular uprisings; and (2) validate those models by forecasting out-of-sample atrocities. We find that the structural variables associated with mass killings during national uprisings are generally consistent with those identified by prior literature on atrocities more generally. However, characteristics of the campaigns—such as whether the episode is primarily nonviolent, the elicitation of foreign support for the campaign or the regime, and the behavior of the military during the episode—also play an important role. We find further support for these findings in forecasting out-of-sample mass killing onsets. These findings demonstrate the importance of observing the behavior of various actors in the midst of popular uprisings as a way to better anticipate and prevent atrocities. While campaign-level factors play less of a role in post-conflict mass killings, we find that the duration of the preceding episode continues to influence whether a mass killing takes place after the crisis as well. We conclude by briefly identifying the implications of these findings for scholars, human rights advocates, and dissidents themselves.