Case Study 35.1: The 2020 BLM Uprisings — Protest Data, Scale, and the Coverage Gap

Overview

The protests following George Floyd's murder on May 25, 2020 constitute the largest recorded protest wave in American history by most quantitative measures. This case study uses the 2020 uprisings as a laboratory for applying protest event analysis methodology, examining coverage bias in real data, and analyzing the relationship between social media mobilization and sustained movement impact.

The Data Landscape

By any measure, the scale of the 2020 protests was historically unprecedented. But different datasets produce substantially different quantitative pictures, and understanding those differences is analytically crucial.

Crowd Counting Consortium data for the period May 26–August 22, 2020 shows approximately 7,750 distinct protest events in the United States, with peak weekly event counts of approximately 900–1,000 events per week in early June. CCC uses trained research assistants reading newspaper coverage, making it one of the higher-quality US protest datasets.

ODA's enhanced dataset for the same period shows 9,847 events — approximately 27 percent more than CCC. The additional events are concentrated in: - Counties with populations under 50,000 (rural/small-town events: ODA captures 2.2x more than CCC in this category) - Events organized by local chapters of national organizations (such as local NAACP branches or community Black Lives Matter chapters) rather than events organized by nationally prominent groups - Events that did not result in arrests or notable incidents (and thus generated less intensive news coverage) - Events documented primarily through social media rather than newspaper archives

GDELT data for the same period shows a dramatically different picture — hundreds of thousands of "events" related to Black Lives Matter and George Floyd, most of which are news articles about the protests rather than distinct protest events. GDELT's automated event-coding inflates counts by treating each news article about protests as a distinct event. Researchers using GDELT for protest analysis must apply aggressive deduplication and filtering, which is technically difficult and produces results that are hard to replicate.

Geographic Analysis: Beyond Urban Centers

A critical finding in Sam Harding's ODA analysis is the geographic breadth of the 2020 protests. Contrary to media narratives that framed the protests as primarily an urban phenomenon:

  • Protest events were recorded in 99.6 percent of all US counties with populations over 25,000
  • 2,143 counties recorded their first-ever documented BLM-related protest event during summer 2020
  • In rural counties (population under 50,000), protest events occurred in communities with demographics as high as 95 percent white — events where local residents organized independently of national BLM organizations

This geographic breadth matters analytically for several reasons. First, it complicates narratives that associate BLM protests exclusively with urban, liberal communities. Second, it suggests the 2020 mobilization reached communities with no prior BLM infrastructure — meaning resources, networks, and organizational capacity created during summer 2020 may persist in communities that had no prior movement presence. Third, it reveals the degree to which viral social media mobilization can overcome the resource constraints that RMT identifies as the key barrier to movement formation in resource-poor environments.

The Coverage Gap in Detail

Sam Harding's comparative analysis of CCC and ODA data reveals the specific dimensions of coverage bias:

Demographic events correlation: ODA's data shows that protest events organized by Black-led organizations in majority-Black neighborhoods receive significantly less newspaper coverage than equivalent events (same size, same day) organized by multiracial coalitions or held in downtown areas. This is not a new finding — research on protest coverage has documented racial coverage disparities for decades — but it is systematically quantified here.

The "newsworthy incident" filter: Approximately 78 percent of events in ODA's dataset that did not appear in CCC's dataset had no arrests, no property damage, and no confrontation between protesters and counter-protesters or police. Events with these "newsworthy incidents" were captured by CCC at high rates regardless of size; events without them were captured only if they were very large.

Rural media desert effect: In counties that had no daily newspaper and limited local news infrastructure, even large protests were undercounted by newspaper-based datasets. The 34 counties in ODA's data where the largest events were documented (500+ participants) and not in CCC are disproportionately rural counties with limited news infrastructure.

Mobilization Mechanisms: Social Media Analysis

ODA's social media data allows analysis of how the 2020 protests spread so rapidly, particularly into communities with no prior BLM presence.

The timeline: Within 48 hours of Floyd's death, Twitter saw the emergence of location-specific hashtags (#[City]Protests, #[State]BLM) that served as coordination mechanisms for locally organized protest. Sam's analysis shows that the median time between the first local #[City]Protests hashtag appearance and the first local protest event was approximately 38 hours — suggesting rapid translation from online to offline mobilization.

Network bridging: Analysis of who initiated local protest organization in communities with no prior BLM infrastructure reveals a consistent pattern: mobilization was led by individuals connected through personal networks to communities with existing movement infrastructure. In many cases, the organizer was a college student who had recently returned home, a person with family connections to a metropolitan area with active BLM chapters, or an individual who had attended a prior protest event in another city. Network bridges carried the movement beyond its organizational base.

The role of Facebook: While Twitter provided the viral spread mechanism and media visibility, Facebook event creation was the primary operational mobilization tool. Local "Protest [City Name] — [Date]" Facebook events were created in thousands of communities, providing the practical information (time, place, logistics) that translated social media attention into physical presence.

Long-Term Movement Impact: What the Data Shows

The 2020 protests generated exceptional media attention, widespread political sympathy, and significant initial policy response (dozens of cities announced police reform measures, several prominent policing policies were changed). By 2022, a more complicated picture emerged:

Policy outcomes: Research by Jessica Trounstine and others finds mixed results on sustained policy change. Some reforms (bans on chokeholds, changes to use-of-force policies, Minneapolis' specific reforms) were implemented and maintained. Others were reversed or never implemented. Defunding or abolishing police departments, the most ambitious demand, was not achieved anywhere.

Electoral effects: Omar Wasow's research on protest and electoral outcomes is directly relevant here. His analysis of the 2020 protests finds that while the protests generated enormous media attention and mobilized new voters, protest-associated violence (whether initiated by protesters, counter-protesters, or police) activated conservative backlash that benefited Republicans in competitive congressional districts. This finding illustrates a fundamental tension in protest analytics: mobilization success (more people protesting) and electoral success (moving votes toward movement allies) are not the same outcome and can even conflict.

Organizational persistence: ODA's 2022 follow-up survey of local BLM-related organizations documents that approximately 65 percent of the local chapters and coalitions formed during summer 2020 remained active in some form two years later — a much higher survival rate than many observers expected. The resources and networks built during summer 2020 did not evaporate with the immediate media cycle.

Discussion Questions

  1. The coverage gap between ODA and CCC shows that rural protest is systematically undercounted by newspaper-based methods. What are the political implications of this undercounting for analysts who advise policy-makers or campaigns on "where the movement is"?

  2. The "newsworthy incident" filter means that dramatic, confrontational protest is over-represented in standard protest databases. If we're measuring "protest activity" primarily through its dramatic and confrontational forms, what are we missing about how movements actually build power?

  3. Wasow's research suggests protest-associated violence hurt Democratic electoral prospects in 2020. Does this finding imply that protest movements should avoid tactics that generate violence (even defensively)? What are the ethical tensions in an analyst advising a movement to moderate tactics for electoral strategic reasons?

  4. The rapid translation from online mobilization to street protest (median 38 hours in 2020) appears faster than what network-building theory would predict. Does this suggest RMT's emphasis on organizational infrastructure is less important than theorists believed, or does it suggest that pre-existing infrastructure (decades of civil rights organizing, Black church networks, HBCU alumni networks) was invisibly present and doing the real work?