What Distinguishes Protests from their Counterfactuals

• In Event: 3451 - Section on Collective Behavior and Social Movements Refereed Roundtables (1 hour)
  In Refereed Roundtable 60min: 9. Participation

Abstract

Social movement scholars often only study occurred protests, given the difficulty of pinpointing potential protests—cases in which movement could happen but do not emerge. This paper follows the recent proposals of McAdam and his colleagues (2012) to study counterfactual cases of protests seriously. We first describe a methodological innovation that use machine learning algorithms to identify protests and potential protests simultaneously, based on both text and images from social media. We applied our algorithms to 9.5 million social media mentions of protest-related words and the associated images from a large Chinese social media platform Weibo. We obtained a dataset with 200,000 protests and over 400,000 potential protests between 2010 and 2017. The potential protest posts expressed grievances and intention to protests, but fail to manifest themselves in real-world actions. Our dataset is the largest dataset that contains both protests and counterfactual cases. Using our unique dataset, we compare the protests and potential protests by time and issues. We find that the rate of mobilization from potential protests to protests increases until the middle of 2015, and declines afterward. We also find that protests by the relatively powerless, such as peasants and migrant workers, have a higher rate of mobilization compared with the city middle class. These descriptive results will provide a sound empirical basis for future theorizing about social movements in general, autocratic regimes, and China.

Author

• Han Zhang, Princeton University