Civil resistance statistics: what the datasets show about tactics, transitions, and repression
Civil resistance is often discussed in broad moral terms, but the datasets behind it are much more specific. The numbers show a field that has expanded in scope, become easier to measure, and still carries a persistent risk of repression and escalation.
Table of contents
- Civil resistance in the data
- How the tactic universe has expanded
- What transition datasets capture
- Who organizes and who gets counted
- What violence and repression tables suggest
- How demonstration datasets change the picture
- Why the numbers matter for analysis
Civil resistance in the data
At the broadest level, the statistical picture is simple: civil resistance is not a single tactic, a single movement type, or a single geographic pattern. It is a large family of actions that datasets try to classify in different ways, and those classification choices shape what researchers can and cannot infer.
One useful starting point is the tactic inventory itself. The Nonviolent Tactics Database includes more than 346 methods of civil resistance, while Sharp?s classic catalog contains 198 methods (Civil Resistance Tactics in the 21st Century). That difference matters because it shows the shift from a comparatively compact taxonomy to a more detailed and more contemporary map of tactics.
A bigger tactic catalog does not just add examples. It changes what counts as visible, comparable, and measurable.
The result is a field where the data structure is as important as the headline number. If the taxonomy changes, the story about civil resistance changes with it.
How the tactic universe has expanded
Sharp?s classic typology still organizes tactics into 3 broad categories: protest and persuasion, noncooperation, and intervention (Civil Resistance Tactics in the 21st Century). That structure remains useful because it gives analysts a stable conceptual frame. But the same monograph also groups tactics into 4 media-based buckets: digital technology, material arts, the human body, and language (Civil Resistance Tactics in the 21st Century).
Those two organizing schemes point in different directions. The 3-part typology helps sort tactics by political function. The 4-part media grouping helps sort them by expressive or operational form. Taken together, they imply that civil resistance can be studied either as a political repertoire or as a communication and mobilization system.
A compact comparison helps show the difference:
| Framework | Count | How it organizes tactics | Source |
|---|---|---|---|
| Nonviolent Tactics Database | 346+ methods | Expanded inventory of civil resistance methods | Civil Resistance Tactics in the 21st Century |
| Sharp classic catalog | 198 methods | Earlier comprehensive taxonomy | Civil Resistance Tactics in the 21st Century |
| Sharp broad typology | 3 categories | Protest and persuasion, noncooperation, intervention | Civil Resistance Tactics in the 21st Century |
| Media-based grouping | 4 buckets | Digital technology, material arts, the human body, language | Civil Resistance Tactics in the 21st Century |
That table highlights a basic analytical point: different taxonomies answer different questions. A researcher asking ?how many tactics exist?? gets a different answer from a researcher asking ?how are tactics functionally grouped??
The expansion from 198 to more than 346 methods also suggests that modern civil resistance cannot be treated as a fixed menu. It is a moving repertoire that grows as organizers adapt to technology, policing, and media environments. That is especially relevant when the tactic set is used to code real-world episodes.
What transition datasets capture
The University of South Florida Peace Institute and related USIP work provide one of the most useful windows into civil resistance as a transition process. The CM-CRT dataset covers 83 political transitions initiated through civil resistance from 1945 to the present (USIP CM-CRT dataset). In version 1.0, the dataset records over 1,400 events from 72 transitions (USIP CM-CRT dataset). The same USIP report describes the dataset as containing roughly 1,600 mobilization events in 72 political transitions from 1945 to 2019 (USIP Peaceworks No. 183). Version 1.1 adds 11 more transitions to cover all known nonviolent-action-initiated transitions since World War II (USIP CM-CRT dataset).
Those figures matter because they tell you the unit of analysis is not just the protest event. The dataset is built to capture transition dynamics over time. It is not merely counting street actions; it is measuring political movement inside periods of change.
The event-level structure is also dense. CM-CRT event records include over 100 variables per event, including concepts such as inclusivity, size, repression, and foreign support (USIP CM-CRT dataset). That breadth is useful because civil resistance is rarely explained by a single cause. A dataset with that many fields can capture more of the surrounding political environment.
A few descriptive statistics make the transition pattern clearer:
- Average events peak at 2.5 events in the first month of a transition (USIP Peaceworks No. 183).
- About 20% of CM-CRT events occur in more than one city (USIP Peaceworks No. 183).
- About 80% of CM-CRT events are limited to a single city (USIP Peaceworks No. 183).
- The large majority of single-city events take place in capital cities (USIP Peaceworks No. 183).
Those numbers suggest a pattern of concentrated, early mobilization. The first month is especially active, but the geography is still heavily centralized. That is a useful reminder that civil resistance can be both nationwide in political significance and highly local in where it physically happens.
Who organizes and who gets counted
The organizer tables in CM-CRT make clear that some groups appear more often than others. The most common identifiable organizing group is workers and unions, averaging 3.35 events per transition (USIP Peaceworks No. 183). Students and youth average 1.94 events per transition (USIP Peaceworks No. 183). Political candidates average 1.36 events per transition (USIP Peaceworks No. 183). Religious communities average 0.79 events per transition (USIP Peaceworks No. 183). For the most common nonstate-unspecified organizer category, the average is 8.99 events per transition (USIP Peaceworks No. 183).
That last figure is important. It implies that a large share of activity is recorded without a more specific organizational label. In other words, the data are useful, but they are not perfectly tidy. Civil resistance often becomes legible in aggregate before it becomes legible in detail.
What violence and repression tables suggest
The violence table in CM-CRT is one of the most revealing parts of the dataset because it shows how different organizer categories relate to violent outcomes. Nonstate armed actors account for 22 violent events and are violent in 75.9% of their recorded events (USIP Peaceworks No. 183). The military accounts for 14 violent events and is violent 38.9% of the time (USIP Peaceworks No. 183). Police account for 12 violent events and are violent 38.7% of the time (USIP Peaceworks No. 183). Religious communities account for 19 violent events and are violent 29.7% of the time (USIP Peaceworks No. 183). Students or youth account for 33 violent events and are violent 20.6% of the time (USIP Peaceworks No. 183). Unspecified nonstate actors account for 153 violent events, the largest absolute total in the table, and are violent in 20.1% of their events (USIP Peaceworks No. 183).
These are not interchangeable categories. The violent share varies sharply by actor type, which means the label attached to the organizer tells you something about risk, but not everything. The same dataset also shows a general escalation relationship: when no protests occur in a transition month, the predicted likelihood of violence in the next month is about 6%, while adding a single public demonstration raises that predicted likelihood to nearly 11% (USIP Peaceworks No. 183).
That is a crucial analytic point. Public demonstration activity is associated with a higher predicted violence risk in the next month, but the baseline and the change remain within a narrow range. The rise is meaningful, yet it does not imply that violence is inevitable. It suggests a probabilistic relationship between mobilization and escalation.
The repression data reinforce that interpretation. Roughly one third of protests and sit-ins in the dataset were subjected to government repression (USIP Peaceworks No. 183). Strikes and boycotts were repressed only around 18% of the time (USIP Peaceworks No. 183). That difference hints that authorities do not respond uniformly to every tactic. Some forms of contention are easier to target, monitor, or disrupt than others.
A compact comparison of the CM-CRT descriptive patterns
| Pattern | Statistic | Source |
|---|---|---|
| Transitions covered | 83 | USIP CM-CRT dataset |
| Events in version 1.0 | Over 1,400 | USIP CM-CRT dataset |
| Events described in Peaceworks No. 183 | Roughly 1,600 | USIP Peaceworks No. 183 |
| More than one city | About 20% | USIP Peaceworks No. 183 |
| Single-city events | About 80% | USIP Peaceworks No. 183 |
| Repression of protests and sit-ins | Roughly one third | USIP Peaceworks No. 183 |
| Repression of strikes and boycotts | Around 18% | USIP Peaceworks No. 183 |
A table like this is useful because it shows the dataset is not centered on one spectacular event type. It is a descriptive record of transition-scale behavior.
How demonstration datasets change the picture
The ACLED-based statistics offer a very different perspective because they focus on demonstration events in specific political moments. In the U.S. between 24 May and 22 August 2020, ACLED records more than 10,600 demonstration events (ACLED US Crisis Monitor). Over 10,100 of those demonstrations, or nearly 95%, involved peaceful protesters (ACLED US Crisis Monitor). Fewer than 570, or about 5%, involved demonstrator violence (ACLED US Crisis Monitor). More than 80% of all demonstrations in that window were connected to either BLM or COVID-19, and nearly 73% were associated with the BLM movement (ACLED US Crisis Monitor).
That is a dense concentration of activity. It indicates that the summer 2020 U.S. protest environment was both broad and highly clustered around a few themes.
The same dataset shows that ACLED recorded over 7,750 BLM-associated demonstrations after 26 May 2020, across more than 2,440 locations in all 50 states and Washington, DC (ACLED US Crisis Monitor). California logged 819 BLM-associated events, New York 430, Florida 380, and Illinois 331 (ACLED US Crisis Monitor). Those state totals matter because they show how a national movement can still have strong geographic variation.
The protest-style breakdown is especially important:
- More than 93% of BLM demonstrations involved non-violent demonstrators (ACLED US Crisis Monitor).
- About 9% were met with police or other authority intervention (ACLED US Crisis Monitor).
- Government personnel used force in more than half of those interventions, at least 54% (ACLED US Crisis Monitor).
That combination suggests that peacefulness and intervention are not opposites. A protest can remain largely non-violent while still drawing a measurable state response.
The year-after-Floyd pattern
ACLED?s year-after-Floyd report extends the picture beyond the summer 2020 window. Approximately 94% of all pro-BLM demonstrations were peaceful (ACLED Year of Racial Justice Protests). The same report says 6% involved violence, clashes with police, vandalism, looting, or other destructive activity (ACLED Year of Racial Justice Protests). Demonstrations involving right-wing militias or militant social movements turned violent or destructive nearly 14% of the time (ACLED Year of Racial Justice Protests).
That comparison is useful because it shows very different risk profiles across demonstration types. It also shows that the category of demonstration matters as much as the overall volume. Averages across all protests can hide sharply different patterns within specific coalitions or counter-mobilizations.
The same report says authorities were three times more likely to intervene in pro-BLM demonstrations than in other demonstrations (ACLED Year of Racial Justice Protests). When authorities intervened, they used force 52% of the time against pro-BLM demonstrations versus 26% against all other demonstrations (ACLED Year of Racial Justice Protests). Against peaceful pro-BLM protesters, authorities used force 37% of the time versus under 20% for other peaceful protesters (ACLED Year of Racial Justice Protests).
Those are among the clearest indicators in the dataset that policing behavior varied by protest context. The report also notes at least 38 distinct named far-right groups engaged directly with pro-BLM demonstrators, and about 26% of the demonstrations involving those groups turned violent or destructive (ACLED Year of Racial Justice Protests). There were over 750 counter-demonstrations nationwide in pro-BLM-related activity (ACLED Year of Racial Justice Protests).
Additional statistics deepen the picture:
- Nearly two-thirds of Back the Blue demonstrations were one-sided events with no intervention by authorities (ACLED Year of Racial Justice Protests).
- Only 2% of pro-police demonstrations were met with law-enforcement intervention, compared with over 9% of pro-BLM demonstrations (ACLED Year of Racial Justice Protests).
- Car rammings were eight times more common at BLM-related demonstrations than at other demonstrations (ACLED Year of Racial Justice Protests).
- Car ramming incidents appeared at nearly 1% of all BLM-related events (ACLED Year of Racial Justice Protests).
- Of BLM-related demonstrations that faced car-ramming attacks, 73% were peaceful (ACLED Year of Racial Justice Protests).
Those numbers say two things at once. First, most BLM demonstrations remained peaceful. Second, the risk environment around them included intervention and attack patterns that were not evenly distributed across protest types.
The anti-protest policy environment is also part of the statistical record. More than 100 anti-protest bills were introduced in over 30 states after Floyd?s death (ACLED Year of Racial Justice Protests). In April 2021, over 500 pro-BLM demonstrations were reported around the U.S., more than in any month since September 2020 (ACLED Year of Racial Justice Protests). Since the start of 2021, 95% of pro-BLM demonstrations were non-violent (ACLED Year of Racial Justice Protests).
That is an important corrective to any simplistic escalation narrative. The demonstration volume remained high, but the great majority of events were still non-violent.
What state-level comparisons add
The ACLED report also compares states that introduced or enacted anti-protest bills with states that did not. In 34 states that introduced or enacted anti-protest bills, 43% of all demonstrations since the start of 2020 were related to BLM (ACLED Year of Racial Justice Protests). In 16 states with no such bills, only 37% of demonstrations were BLM-related (ACLED Year of Racial Justice Protests). In the states pursuing anti-protest laws, authorities engaged in at least 6% of demonstrations versus 5% in other states, and authorities used force in 3% of demonstrations versus 2% in other states (ACLED Year of Racial Justice Protests).
The differences are not enormous in raw percentage terms, but they are consistent with a more intervention-heavy environment. That matters because small percentage gaps can still translate into many events when the overall event count is very large.
A final figure from the same report captures the political backdrop. A 2020 poll cited in the ACLED report found that 41% of Americans remained opposed to the BLM movement (ACLED Year of Racial Justice Protests). That helps explain why protest data cannot be interpreted in isolation from broader public opinion and policy response.
Why the numbers matter for analysis
Taken together, these statistics show that civil resistance is best understood as a measured, evolving repertoire with distinct patterns at the level of tactics, transitions, organizers, repression, and policing. The Nonviolent Tactics Database expands the repertoire to more than 346 methods, while Sharp?s classic catalog remains a useful 198-method baseline (Civil Resistance Tactics in the 21st Century). CM-CRT shows how transitions unfold over time, with over 1,400 to roughly 1,600 events across 72 transitions and 83 political transitions covered in later versions (USIP CM-CRT dataset; USIP Peaceworks No. 183). ACLED shows how a large protest wave can remain overwhelmingly peaceful while still producing intervention, violence, and policy backlash patterns that differ by movement and state context (ACLED US Crisis Monitor; ACLED Year of Racial Justice Protests).
The practical lesson is straightforward. Civil resistance statistics are most useful when they are read as patterns, not slogans. The numbers are rich enough to compare tactics, timelines, organizer types, and state responses, but only if the source labels stay attached to the facts.