Uncovering Authoritarian Rule: Identifying Collective Action with Social Media Data

• In Event: 2277 - Methodological Advances in Research on Social Movements

Abstract

In authoritarian regimes, collective action--protest or other forms of collective social mobilization by groups outside the government--often represents the most effective form of political participation available to the public. Understanding the prevalence and characteristics of collective action in authoritarian settings would be highly valuable for numerous scientific and public policy purposes, yet independent measures of collective action are largely nonexistent. In this paper, we create a semi-automated system of identifying real-world collective action events by utilizing social media data. We design a two-stage Convolutional Neural Network classifiers that rely on both text and images to identify collective action events from protest mentions on Sina Weibo. The first-stage classifier identifies social media posts that contain mobilization potential. The second-stage classifier then separates protest posts that discussed real-world collective action events, and ``counterfactual'' posts that expressed grievances and intention to protest, but no offline collective action can be inferred from the posts.
We apply our system to China, using data from Sina Weibo, to generate 197,734 unique collective action events from 2010 to 2017, creating one of the largest datasets of collective action events in an authoritarian regime. We demonstrate the usefulness of our new dataset by providing some new findings that contrast the received beliefs about protest activities in China. We show that the number of collective action events in China may be declining after 2013. We also show that what types of protests are more likely to manifest themselves into real-world collective action events.
Our paper provides new approaches to using social media to construct protest datasets and understand protest activities in authoritarian regimes.
We plan to release our datasets and codes to facilitate researchers in the study of social movements and authoritarian politics.

Authors

• Han Zhang, Princeton University

• Jennifer Pan, Stanford University