Do protests change policy?
Verdict
Whether protests work depends heavily on scale, tactics, and duration. Large, sustained, nonviolent, broad-coalition protests with sympathetic media coverage are associated with real, measurable shifts in voting behavior and elite agendas. Violent or highly disruptive protest produces smaller effects and can trigger backlash. A single small, one-off local protest rarely moves policy directly. Grade: moderate, genuine causal evidence exists but is concentrated in a handful of well-designed studies, and the same evidence points clearly to what raises the odds: scale, discipline, and a sequence that channels turnout into organizing and an inside ask.
What the research says
A dataset of 323 resistance campaigns from 1900 to 2006 (Chenoweth & Stephan) found nonviolent campaigns succeeded 53% of the time versus 26% for violent ones, the source of the widely cited "3.5% rule."1 That scope is regime change, foreign occupation, or secession, not routine US domestic policy fights, a limit Chenoweth has repeatedly stated herself.
Two studies using rainfall as an instrument for protest size found real electoral effects. A 2013 study (Madestam et al.) on 2009 Tea Party rallies found bigger rallies, driven by better weather, caused more local organizing, more contributions, higher 2010 Republican vote share, and more conservative voting by affected incumbents.2 A 2025 study (Klein Teeselink & Melios) replicated this design for 2020 BLM protests and found protest intensity caused increased Democratic vote share, with evidence of genuine attitude change, not just turnout mobilization.3
A 2020 study of 1960s civil rights-era protests (Wasow) found nonviolent protest proximity shifted white voters slightly toward civil rights sympathies, while violent protest proximity shifted them right.4 Violent unrest after MLK's 1968 assassination is estimated to have cut Democratic vote share by at least 2 points in affected counties. A 2025 systematic review of 19 studies on "radical flank" effects found positive effects, where radical action makes moderates look reasonable, more common (12 of 19) than backfire (7 of 19), complicating the simple "disruption always backfires" narrative, though the review disclosed a competing interest.5
What it doesn't say
The most famous number, 3.5%, comes from authoritarian-regime-toppling contexts and doesn't transfer cleanly to routine democratic policy fights. Causal identification in the domestic US studies relies on rainfall and geographic-proximity instruments, credible but contested on their identifying assumptions. Effect sizes found are often modest, around 2 points, not landslide shifts. The 1960s media environment, three national broadcast networks, differs enormously from today's fragmented, algorithmic environment, so historical findings may not transfer directly. Real-time data on the October 2025 "No Kings" protest is descriptive and polling data, correlated with declining presidential approval, but no causal study yet connects the two.
What this means for you
Conditions that raise the odds a protest matters: scale, nonviolent discipline (violence risks backlash among persuadable voters, though modest disruption or civil disobedience often doesn't backfire and can help via radical-flank dynamics), sustained repetition rather than a single event, a broad and diverse coalition, and sympathetic media framing. A single small local rally with no electoral or organizing follow-through has little empirical track record of moving policy on its own. The real leverage tends to run through elections and organizing infrastructure built over time, not the protest event itself.
How people actually move this
The protests with a documented track record of moving policy share a structure: broad and diverse participation rather than a narrow base, strict nonviolent discipline, tactical variety across dispersed and concentrated actions, and repetition over months rather than a single day. That's the pattern behind Chenoweth and Stephan's finding that nonviolent campaigns succeeded roughly twice as often as violent ones.1 Scale matters, but the historical record also shows scale without follow-through fails, so the higher-agency version treats the protest as an input to organizing, not the end product: collect contacts at the rally, convert energy into calls and testimony aimed at a specific decision-maker, and pursue an inside ask, a meeting, a drafted amendment, in parallel with the outside pressure. Coordinate timing to a real policy window, a hearing date or a vote, rather than a general grievance. A protest with no next step rarely outlasts the news cycle it made.
Sources
-
Chenoweth, E. & Stephan, M., NAVCO dataset, "Why Civil Resistance Works," 2011. source ↩ ↩2
-
Madestam, A. et al., Quarterly Journal of Economics, "Do Political Protests Matter?," 2013. source ↩
-
Klein Teeselink, B. & Melios, G., Political Behavior, 2025. source ↩
-
Wasow, O., American Political Science Review, "Agenda Seeding," 2020. source ↩