Unmaking Boundaries: Inter-faction Dynamics during the 2019 Hong Kong Anti-Extradition Law Amendment Bill Protests

Weijun Yuan

2024 Mayer N. Zald Distinguished Contribution to Scholarship Student Paper Award, ASA Collective Behavior and Social Movements Section

Why, when, and how do social actors engage one another across established categorical boundaries? This paper examines boundary-spanning behaviors of moderates and radicals within social movements, analyzing how activist groups endorse one another despite their differences and potential negative repercussions of such endorsements. It argues that boundary-spanning behaviors of movement actors are both adaptive and performative, in response to short-term events such as interactions with various institutional actors, brief windows of political openness, elite endorsements, and international support. These relational events alter political access and audience scope, affecting how movement actors prioritize their images of worthiness, unity, numbers, and commitment, which in turn shapes inter-faction relations. This analysis leverages original inter-organizational endorsements network and political event datasets extracted from over 730,000 Telegram posts during the 2019 Hong Kong Anti-Extradition Amendment Bill protests. Blockmodels and relational event models results show that moderates and radicals are more likely to endorse across factional lines when interactions with institutional actors signal restricted access to the polity and expose movement actors to a broad audience. The findings contribute to broad discussions on boundary-spanning behaviors, inter-organizational coordination, movement dynamics, and contentious strategies.

What Drives the News Coverage of US Social Movements?

What drives the news coverage of social movements in the professional news media? We address this question by elaborating an institutional mediation model arguing that the news values, routines, and characteristics of the news media induce them to pay attention to movements depending on their characteristics and the political contexts in which they engage. The newsmaking characteristics of movements include their disruptive capacities and organizational strength, and the political contexts include a partisan regime in power, benefitting from national policies, and congressional investigations. To appraise these arguments, we analyze approximately 1 million news articles mentioning 29 social movements over the twentieth century, published in four national newspapers. We use negative binomial regression analyses and separate time-series analyses of the labor movement to assess the model’s robustness across different movements, time periods, and news sources. In each analysis, the results support the hypotheses based on the institutional mediation model. More generally, we argue that the inf luence of social movements on institutions depends on the structure and operating procedures of those institutions. This insight has implications for future studies of the inf luence of movements on major social institutions.

Coalitions under Threat: Analyzing the 2019 Hong Kong Anti-Extradition Protests Using Telegram Social Media Data

This paper examines how threats posed by indiscriminate and selective repression affect the shape and structure of interorganizational coalitions during the 2019 Anti-Extradition Law Amendment Bill (Anti-ELAB) protests in Hong Kong. The analysis relies on an original political event dataset and an organization-event network dataset. These datasets were produced utilizing syntactic event coding techniques based on Telegram posts, which Hong Kong protesters used to distribute information, plan future actions, and crowdsource news. Furthermore, Telegram provides detailed information about state activities, event-level coalitions, and violent groups, which is difficult to access from other sources. This study investigates the coalition networks across the movement’s four stages, each of which was marked by a particular type and degree of repression. The findings indicate that indiscriminate and selective repression have varied effects on coalition networks. A wide coalition disintegrates as a result of indiscriminate repression. Selective repression, however, encourages coalitions to form around the groups that are the targets of repression.

Beyond the Protest Paradigm: Four Types of News Coverage and America's Most Prominent Social Movement Organizations

What determines the quality of coverage received by social movement organizations when they appear extensively in the news? Research on the news coverage of social movement organizations is dominated by case studies supporting the “protest paradigm,” which argues that journalists portray movement activists trivially and negatively when covering protest. However, movement organizations often make long-running news for many different reasons, mainly not protest. We argue that some of this extensive news will lead to worse coverage—in terms of substance and sentiment—notably when the main action covered involves violence. Extensive coverage centered on other actions, however, notably politically assertive action, will tend to produce “good news” in these dimensions. We analyze the news of the twentieth century's 100 most-covered U.S. movement organizations in their biggest news year in four national newspapers. Topic models indicate that these organizations were mainly covered for actions other than nonviolent protest, including politically assertive action, strikes, civic action, investigations, trials, and violence. Natural language processing analyses and hand-coding show that their news also varied widely in sentiment and substance. Employing qualitative comparative analyses, we find that the main action behind news strongly influences its quality, and there may be several news paradigms for movement organizations.

How to Analyze the Influence of Social Movements with QCA? Combinational Hypotheses, Venn Diagrams, and Movements Making Big News

Under which conditions do social movements receive extensive attention from the mainstream news media? We develop an institutional mediation model that argues that combinations of the news-heightening characteristics of movements, including their disruptive capacities, organizational resources, and political orientation, and political contexts, including partisan regimes and benefiting from national policies, bring extensive attention to movements. It also holds that investigations will draw extensive media attention to movements, and those that have achieved prominence in the news will remain prominent under specific conditions. We appraise these combinational arguments by examining 29 social movements across 100 years in four national newspapers using qualitative comparative analysis (QCA). Researchers typically use QCA to study the consequences of movements when they hypothesize outcomes to result from multiple combinations of conditions. This raises our second main question: How should scholars best address combinational hypotheses using QCA? Here we employ Venn diagrams to identify and illustrate key analytical issues and anomalies, including constrained diversity in observational data, empirical instances when combinations of conditions do not produce the expected outcome, and instances when unexpected combinations of conditions produce a consistent result. We also demonstrate the value of broad comparisons across movements and over time in these analyses.

Emerging Consensus: How do Activist Groups Converge on their Demands?

Weijun Yuan

How do different activist groups reach a consensus about their demands? A diffusion model would expect organizations with the most resources and connections to be most influential. The emergence of consensus, therefore, is a process in which ideas of large and well-connected organizations are diffused and adopted by other activist groups. An alternative hypothesis, in contrast, highlights the “fringe effects” of radical organizations and the mediation effects of media, where mainstream media pull fringe organizations and their frames to the center of the discursive field by amplifying fear and anger. My paper appraises both hypotheses and moreover, presents an organizational learning model of how activists collectively develop common expressions of their demands. Using text data from 125 Telegram channels over a six-month period, I empirically capture the activists’ cognitive schemas—their framings of their demands—using word-occurrence semantic networks. Furthermore, I use kernelized Principal Component Analysis to capture the convergence of framings created and disseminated by various organizations and activist groups over time. My work reveals that activist groups rarely arrived at any one single common demand. However, some level of consensus emerges as moderate organizations incorporate the frames of radical groups. In this process, activist groups go through a series of “tests and trials,” weighing factors such as media attention and opposition to the movement. This paper is among the few existing attempts to analyze how communication networks and networks of meaning coevolve.

Text Analysis with Generative Large Language Models

Scholars utilizing text data to create datasets have traditionally faced scalability issues, as they rely on trained human coders to code variables that are relevant to their substantive topics. With the advent of automated text analysis methods, the reliability and efficiency of data processing have significantly improved, these methods falter when dealing with complex variables.  Transformer-based Large Language Models (LLMs) are now promising tools that may further enhance our ability to utilize text data and create large-scale datasets. Emerging as powerful tools for textual analysis, LLMs also come with limitations and introduce challenges that demand careful consideration and operations. This paper applies LLMs to content analysis based on newspaper articles and compares the new approaches to existing data coding and sentiment analysis methods. We showcase LLMs’ ability to achieve accuracy comparable to manual coding standards. This paper also provides a framework on how to effectively use LLMs by addressing key considerations such as model selection, prompt engineering, and validations. Finally, we discuss the importance of addressing bias, replicability, and data-sharing concerns.

Weijun Yuan, Edwin Amenta, and Neal Caren

Figure adopted from Attention Is All You Need (2017) by Vaswani, Shazeer, Parmar, Uszkoreit, Jones, Gomez, Kaiser, and Polosukhin, https://arxiv.org/abs/1706.03762

One path or many? Policy Development and Diffusion across Wikipedia Language Editions

Do comparable self-governing collective action institutions converge on comparable policy systems? Do they do so via comparable developmental paths? We test both theories using data on 60 policies shared by 245 Wikipedia language editions which we use as a petri dish for the diffusion of policy in collective action institutions more broadly. We find that policies that are shared tend to be shared widely, that nearly every shared policy can be found in the English edition, and that the clearest predictor of policy adoption order is policy popularity across editions. We also find that policies that have been adopted by the most editions are most likely to be adopted first by a given edition.

Although we do not definitively eliminate the possibility that language editions follow multiple paths in converging on their policy systems (say, by culture or language), the evidence suggests that editions follow a single noisy developmental path, potentially suggesting strong influence across editions and a stronger role of common structural constraints than diverse cultural constraints in determining patterns of policy adoption.