Does protesting work?

There’s been a lot of news about political protests recently in the UK, and some rather unsavoury efforts by the government to stop particular instances of them happening.

Outside of the UK recent years have seen some very high profile such protests. 2020 saw the Black Lives Matter related protests after the killing of George Floyd, the Iranian protests that started in 2022 after the death of Mahsa Amini. The Arab Spring, the Occupy movement and the Yellow Vest protests happened in the decade before. And countless other examples, including a multitude smaller actions that never piqued the interest of the media enough to be widely reported on.

Within the UK, the recent environmental protests by the likes of Extinction Rebellion seem to engender a certain amount of polarisation. The positions at the extreme seem to be something like either:

• That the environmental issue, above any other, needs to protested in any way possible in order to avoid planet-threatening, species-ending catastrophe; versus:
• That the protesters are idiots who do nothing but commit wanton crime whilst frustrating a set of people who can’t do anything about the issue (and of course there’s small number of people who would doubt the manifold evidence that the issue is of import enough to take action on in the first place).

I came to wondering to what extent there was any data-based evidence on the efficacy of protests. Do authoritarians tend to crack down on protests because they work too well – or are they in fact mostly wasted effort that would be far better channelled elsewhere? Are there any characteristics of specific protests or movements that affect the answer to that question, any causes easier to influence via this route than others, and so on.

Of course this is a pretty difficult question to answer. For a start, what even counts as a protest in this context? Some examples are easy to classify – e.g. organised marches that are explicitly designed to affect specific government policy. But how about me deciding to boycott a product? Workers going on strike?

Point of view likely influences people’s view. One person’s solidarity protest is another person’s “hate march“. Were the events of January 6th in the US a protest in the name of free and fair elections, a traitorous insurrection, neither, or both?

Then, what counts as effective? Is it as simple as whether the demands of the protesters are met? But people’s reasons for protesting, for joining a movement, are often varied. This is inevitably the case when you start getting groups of hundreds, thousands or even millions of people together.

And what about less obviously easy-to-measure determinants of success such as increasing awareness, influencing public opinion or bringing a sense of solidarity to those who need it? Given the scope of some issues typically being protested, even the most effective individual effort might in theory be a small yet instrumental stepping stone on a path towards progress that takes years or decades to fully resolve.

Modern day issues that are seen as worth protesting are typically complex and multifaceted. As are the methods of protest, neither of which happen in a vacuum. Whilst lab studies might be able to test a very limited set of hypotheses around the subject, there is no realistic way to run the gold-standard approach to causality – the randomised controlled trial – to protest as a whole at the level of a specific cause or protest. We’d have to find a way to magically create two equivalent and independent worlds where the functionally similar group people are assigned at random to protest or not protest a real-world issue to see what happens at some point in the undefined potentially distant future. That not being a realistic option, I was curious to see which methods researchers had used to address the question.

In my mind, one of the first steps to potentially establishing the general efficacy of protest actions would likely be collecting descriptive data on protests that have happened in the real world. How many are there? What were they about? What did people want? And so on.

Reading around, I found a couple of sources where researchers had gone to the effort of collecting a decently-sized sample.

The first is a book called World Protests – A Study of Key Protest Issues in the 21st Century by Ortiz et al which was published in 2022. It’s freely available online here or you can purchase a paper copy.

In that, they’ve collected details of 2809 protests that happened between 2006 and 2020 within 101 different countries. These were culled from 15 years worth of news reports from online sources in numerous languages, which are also documented and searchable on the World Protests website. As they note, this sourcing necessarily does create some bias. Not every protest makes the news, and what is missing is not random.

They claim that there have indeed been a high number of protests in recent times, comparing it to other times in history where they identify similarly exacerbated demands for change – between 1830 and 1848, 1917-1924 and during the 1960s.

More recently, they identify that there’s been a steady increase in the protests each year from 2006 to 2020 in all world regions. Their commentary talks about inputs including the global financial crisis during 2007-2008, the resulting tendency towards austerity politics and finally general discontent with the democratic process itself and the political and economic systems populations find themselves under.

They categorise the main causes being protested against using a 4-way classification:

• Failure of Political Representation and Political Systems: 1503 protests were related to the lack of real democracy; corruption; failure to receive justice from the legal system; sovereignty and patriotic issues; transparency and accountability; the perceived power of a deep government or oligarchy; anti-war or against the military-industrial complex; the surveillance of citizens; and anti-socialism and anti-communism.
• Economic Justice and Anti-Austerity: 1484 protests on issues related to jobs, wages and/or labor conditions; the reform of public services; corporate influence, deregulation, and privatization; inequality; tax and fiscal justice; low living standards; agrarian/land reform; high fuel and energy prices; pension reform; housing; and high food prices.
• Civil Rights: 1360 protests on ethnic/indigenous/racial rights; right to the commons (digital, land, cultural, atmospheric); freedom of assembly, speech, and press; women’s and girls’ rights; labor rights; LGBT and sexual rights; immigrants’ rights; personal freedoms; prisoners’ rights and religious issues. In this category have been added those radical right protests that sought to deny rights or reject equal rights for a group (e.g., against minorities).
• Global Justice: 897 protests were for environmental and climate justice; against the IMF, the World Bank, and the European Union/European Central Bank; against imperialism (United States, China); against free trade; in defense of the global commons; and against the G20.

Each category includes several subcategories, the volumes of which are shown in their figure 1, reproduced below. Note that very commonly a single protest may have multiple demands or themes, in which case it’d be included in several bars.

The cause of the protests varied over time in the following way, from their figure 2.

The number of protestors involved has been increasing over time too. They estimate that at least 52 events had at least a million protestors, with the largest being a strike in India that involved 250 million participants.

Most protests were towards what might be called “progressive” causes – better employment conditions, public services, climate change, justice for all, that kind of thing. But there were also plenty of protests led by radical right groups, QAnon, those who set themselves against Muslims, refugees and the like.

80% of protests targeted governments – demanding that they “take responsibility for economic, social, and environmental policies so that they benefit all, instead of the few”. Institutions and systems seen as unaccountable were also popular. The most popular targets by protest volume are shown below, their figure 18.

What methods did the protestors use? Well here they had a fairly expansive definition, having identified a whopping 250 different methods of non-violent protest alone. I’m not going to list them all here, but you can see them in the PDF version of their book starting page 122.

Here’s a breakdown of the most common methods that they counted, their figure 17. Again, a single protest might fall into several categories here, hence the figures add up to over 100%.

Some forms of protest included are necessarily recent inventions due to the technology involved – for instance computer hacking or the online leaking of huge quantities of private information from governments or companies.

They put the percentage of protests that involved those media-scare-favoured activities of violence, vandalism or looting at around 20%.

So onto the question that inspired me to start looking into the topic. Were the protests effective? The authors here use a fairly wide definition of success. They talk about “achievements”:

….“achievements” are understood as the set of direct and indirect responses from opponents or by society to a protest episode, responding in some measure to the grievances and demands raised by protestor

So note that this doesn’t mean that the protesters successfully got all their demands met, but rather that they provoked some sort of associated response from either the targets of the protest or more widely in society. They also note that usually this success doesn’t result from one individual act of protest. Rather it can come after several years of protesting against the same issue.

It also, as far as I can tell, looks only at the subject of the protest and compares it to any later changes. It doesn’t directly address causality in the sense that other causes for the potential change of the respective policy are not excluded.

In any case, by this standard, they conclude that 42% of protests led to “some kind of demonstrable achievement, generally a partial success”.

Probably unsurprisingly, protests with more concrete demands – “raise wages!”, “subsidise food costs!” – were more likely to succeed in some way than those with structural goals – “fix inequality!”, “stop imperialism!”.

Likewise protests that opposed a named, nearby opponent e.g. one’s local government were more likely to succeed than those that opposed more distant or structural opponents such as “free trade”. This chart, their figure 20, shows what % of protests achieved something by the type of opponent the protest targeted.

There wasn’t big differences in success rates in terms of the 4 classifications of the topic the protesters were protesting about mentioned above.

Success rates also differed by protest methods. The most successful included:

• Merchant’s strikes: 75%
• Whistleblowing and leaks: 71%
• Hacking: 64%
• Boycotts: 63%

Least successful were:

• General assemblies: 23%.
• Street theatre: 30%
• Noise-making: 31%
• Educational actions: 34%
• Twitter storms: 36%

Protests methods involving violence, vandalism or looting were associated with success 43% of the time.

Although I’d have intuited the percentage of protests being deemed successful in some way would be a bit lower – but would still have been happy to join a protest anyway – the authors are concerned that these success rates might seem offputtingly low to some. They’re at pains to say that they can’t possibly be measuring all relevant consequences.

Numbers may appear pessimistic in terms of the rate of success of protestors. However, these outcomes are not necessarily negative, since many of the protests are engaged with long-term structural issues that may yield results over a long period of time; incremental or short-term achievements may prove to be precursors to more comprehensive change.

The second source I looked at was a chapter called “Why Civil Resistance Works – The Strategic Logic of Nonviolent Conflict” by Maria J. Stephan and Erica Chenoweth. The main goal of the authors appears to be to compare the effectiveness of non-violent protests to violent protests. In doing this they created a model with the intention of explaining some of the factors that make protests more or less likely to get the results that they desire.

The title gives away the conclusion that they come to here. But of course in trying to answer their question they necessarily have to classify protests into those that were “successful” vs those that weren’t, which will be of interest to us here.

The version I was looking at was published in 2008, although it looks like they went on to create a full book with the same name in 2011.

Being written a few years before the study discussed above meant that they had to use an older dataset. The cases they selected here were also chosen on the basis that they represented campaigns seeking to overthrow a government or liberate territory where at least 1,000 participants were observed. These are very specific types of protest campaigns of course – it’s not obvious that results here would necessarily translate to campaigns for the environment or better public services.

They also chose to use information from a much longer span of time, finding details on 323 “violent and non-violent resistance campaigns” that occurred between 1900 and 2006.

Note here the unit is “resistance campaign” rather than individual protests. Their definition of resistance campaign is “a series of observable, continuous tactics in pursuit of a political objective.”

In theory a campaign could be over within a day, or it could stretch out over several years. But typically it will have some kind of leadership, name, as well as start and end dates that encompass any events that occurred within the campaign. Again how they went about selecting the campaigns to focus on necessarily introduces a risk of bias – it was largely a literature review and the leveraging of experts. Footnote 37 of the paper gives the rationale and limitations around analysing of campaigns rather than individual protest events.

Whilst focusing on establishing the differences between violent and non-violent campaigns the paper does acknowledge that classifying campaigns in that way can be difficult, especially when distant in time and space. Some resistance campaigns may encompass both violent and non-violent actions, potentially from different groups of people within the campaign.

They classify each campaign into 3 categories of success:

• Success: the objective of the campaign must have been met within 2 years of the campaign itself ending and the campaign must have had a “discernable effect” on the outcome.
• Limited success: campaigns which are associated with obtaining “significant concessions” for their cause, but whose stated objectives weren’t fully achieved.
• Failure: No significant concessions were obtained.

More precise details as to how it was decided which campaign fit which category aren’t specified that I could see.

In any case, their method finds that major non-violent campaigns achieve success 53% of the time, vs 26% for those involving substantial violence.

Forgive the extreme lack of adherence to data visualisation best practices – I culled this chart from a talk given by one of the authors – but the breakdown of outcome was as follows:

The trends of non-violent campaigns being more successful have been increasing over time:

Breaking down some of the driving factors, here’s their model showing the relative risk ratio coefficients for each type of campaign and each level of success compared to a base of failed campaigns.

Their conclusions include that:

• Campaigns that lead to “security forces and civilian bureaucrats” shifting their loyalty in favour of the protesters are likely to succeed. This behaviour occurs more within non-violent protests.
• Whether violence is committed by the regime against the campaigners or not has little effect on their outcome.
• State and/or international support for a campaign makes it more likely to succeed if the campaign was violent, but has little effect otherwise. They hypothesise that the reason it doesn’t really help non-violent campaigns is that these inputs lead activists to rely too heavily on support from outsiders rather than concentrating on building their own power base internally.
• Campaigns within more democratic regimes are more likely to succeed.
• The longer the campaign goes on the less likely it is to fully succeed, although it may be more likely to get partial success.

A key part of the driving force behind success is the number of people involved. Size matters, and it appears to be much easier to attract fellow activists to non-violent campaigns.

One of the authors, Erica Chenoweth, spoke about this in a TED talk a few years ago:

This is also where the somewhat famous 3.5% figure you may have heard of in the past comes from.

As might fit your intuition, they found that the greater the proportion of population that was involved in campaign the more success it typically had. In fact they found that every single one of the campaigns in their dataset that had the active and sustained participation of at least 3.5% of the population concerned were successful.

And every single one of the campaigns that attracted at least 3.5% of the population was a non-violent one. Non-violent campaigns tended to attract a much larger proportion of the prospective campaigners – in absolute terms this corresponded to 200k people vs 50k people on average. Those folk that are drawn in to non-violent campaigns tend to be a lot more inclusive and representative of their underlying population than those involving violent action.

The idea that non-violent protests are more attractive to potential supporters in terms of their support for the protesters rather than simply their willingness to join the fray has been tested to some extent in the laboratory. Based on surveying people about various contrived scenarios, Simpson et al found that even for causes that are fairly universally seen as “good” in some quarters – here protestors on the side of anti-racism vs those who were fighting for the cause of white nationalism, violence was off-putting.

They found that:

• The use of violence by antiracist protesters against white nationalists decreased support for the antiracist group whilst increasing it for the white nationalist group.
• The use of violence led to the participants feeling like the protesters were more unreasonable, making it less likely for the participant to identify with or support them.
• To some extent the same effect applied in reverse to the scenario where white nationalists were the violent protestors, although there was no equivalent effect on support overall. The hypothesis here is that much of the public thankfully already perceives white nationalist groups as unreasonable and hard to identify with irrespective of how politely they protest.

Note here that the outcomes being studied are around popular support. Increasing support is certainly one desirable outcome for many protestors, but it differs from the studies discussed at more length above which tend to look at to what extent the protestors succeeded at their stated objectives – typically, perhaps for want of a better term, policy objectives – rather than to what extent they increased favourable public opinion. But if, as was suggested above, wider support leads to a greater chance of success then the two are naturally connected.

Relatedly, Wouters and Walgrave published a study that looked at how public opinion on an issue was likely to influence the viewpoints of the set of people who often are the target of protests based on their holding of the power to actually change the things that the protestors care about: politicians.

By manipulating supposed TV footage of asylum policy related protests they found that, whilst their prior beliefs still had a large effect, politicians were actually responsive to their perceptions of the protestors. The dimensions of note when it came to the effect the protestors had on the politicians’ opinions were based on:

• Worthiness – are the protesters “credible” people?
• Unity – do they agree amongst themselves about what they want?
• Numerical strength – are there a lot of them?
• Commitment – do they care strongly enough about the issue not to give up easily?

The most important dimensions according to this study – i.e. those that were most likely to alter politicians’ pre-existing opinions and potentially actions – were numerical strength and unity. So from this point of view, the most effective protests might be those that have a large number of people who are all on the same page when it comes to both the demands they’re making and claims that they use to support those demands.

The crux of staging a protest that successfully affects elected officials’ opinion formation is to come across as united and numerous.

Back outside of the laboratory, the final study I want to mention here is a recent October 2023 entry into the literature by Shuman et al called “When are social protests effective?“. In this one they review some of the studies mentioned above, and many more, in order to try and come up with a general framework that could be used to evaluate the efficacy of protest action both in terms of evidence collected in studies so far as well as to guide the direction of any future such studies.

They observe that across the literature people are coming to differing conclusions as to the effectiveness of protest and suggest that that might be explained by the fact that protests are too varied to analyse as a homogenous whole. Seemingly incompatible findings could be explained by the fact that the protests being analysed are simply different from each other. Just because we call them all “protests” doesn’t mean their potential mechanisms of action are similar.

In the end they come up with a 3-way classification of protest actions. They suggest that protest effectiveness can be defined by:

1. the type of protest.
2. the target audience.
3. the outcome being analysed.

Their figure 2 illustrates this and the categories of interest in each variable more graphically:

Some explanation of the terminology may be helpful here.

Regarding the type of protest:

• Normative nonviolent refers to protests that do not harm people or property, but beyond that also fall within the socially accepted and legal norms of the society they take place in. Examples would include peaceful demonstrations, rallies and petitions. The 2017 Women’s March or 2015 Global Climate March are examples of this.
• Non-normative nonviolent are those that remain non-violent but do not stay within social norms. These include civil disobedience, strikes, sit-ins and roadblocking. The US Civil Rights Movement included several examples of this. Some of the actions that make the UK news from Extinction Rebellion might also feature.
• Non-normative violent protests involve violent action, rioting, property destruction, violence against people, that kind of thing. The riots that started in Los Angeles after the 1992 police beating of Rodney King is an example.
• The radical flank entry is there to acknowledge that social movements can contain multitudes. Actions may consist of a large majority of peaceful, moderate, normative actors but also have a smaller more radical group that is predisposed to break norms or commit violence. Think of the Blank Panthers contingent of the Civil Rights Movement, or any subset of the environmental movement that destroys property as part of their tactics.

It’s acknowledged that often classifying a particular protest even into those categories involves a certain amount of subjectivity at times.

The target audience classification is more self-explanatory. Are the set of people the protesters are targeting with their action – their audience – already likely to be sympathetic to their cause or dead set against it?

The outcome classification refers to what the protesters are trying to achieve (or sometimes perhaps more what the researcher has set out to measure when it comes to the analysis of studies).

• Mobilisation refers to whether social protests can attract non-participants to support their cause or even join their action.
• Policy change is about whether the protest advances the change in policy that the protesters want.

Both of these can be measured in very different ways between between studies. The nature of this review I think means that the authors just have to accept the success vs failure outcome that the underlying studies provide.

So what do they find when they review the findings of other studies using this framework?

Mainly that:

1. Normative non-violent forms of protest are more effective if the goal is mobilise audiences that are already sympathetic to the cause.
2. Non-normative, perhaps even violent, forms of protest are more effective if the goal is to induce policy changes when the audience starts off as resistant.

They offer a few ideas as to why this may be.

For the former scenario: a sympathetic audience would presumably be happy enough if the protest movement succeeded in getting what it wants, so really they just need to be mobilised to join the action. To do this, the protestors need to increase the motivation the target feels to actively support the cause, whilst lowering any cost of actually doing so.

Protests that adhere to social and legal norms, including the norm of non-violence, make it more likely that the average sympathiser will find themselves identifying with the protesters. Research suggests that they see non-violent, norm-following protesters as more reasonable and moral people, people just like themselves – and group identification is a big motivator towards participation.

At the same time there is little “punishment” for supporting the cause. If the protest you join is peaceful and legal, you’re much less likely to be arrested or have your peer group consider you some kind of weird radical activist.

Things are very different though if the people you seek to influence are actively against your aims though. They’re unlikely to ever identify with your protest group, let alone join in on your side. Here the audience’s goal is thus usually all about making the protests stop so normal life can resume, without the change the protestors want.

Given that there’s no positive incentive for the audience to take action, they’ll just ignore the protests if they can. As they basically can’t be converted, the goal for the protestors becomes to motivate the resistant audience to react. And non-normative or violent protests, including those with a radical flank, can be impossible to ignore. “Something must be done”, and that something might include making some kind of concession to the protestors – perhaps a price worth paying to get them to stop.

The obvious problem is that that the “something” might not help the protesters’ cause at all. Instead, a powerful audience might aggressively crack down on the protestors in order to defend the status quo. The authors didn’t find much research on this matter, but suggest that a way to make this undesirable outcome less likely is to ensure a protest always contains some kind of constructive element.

The audience needs to believe that the protestors would stop their actions if their demands were met, that they are people who can be reasoned and negotiated with. Radical flanks may have their uses here. Sure, they might be seen as terrible people who must be stopped, but the bulk of the more norm-complying protesters may appear as much more reasonable prospective partners that if compromised with might end the disruption.

As a final note, not everyone thinks that the likely efficacy of a protest should be the primary motivation to join in. Sometimes it may just be the right thing to do, and a unwillingness to engage on the basis that most likely it “won’t work” firstly becomes an excuse to take no action.

As Vaclev Havel – the first democratically elected president of Czechoslovakia and the Czech Republic – said on the occasion of the 50th anniversary of Amnesty International:

Those that say that individuals are not capable of changing anything are only looking for excuses.

And secondly, not engaging in protest because you think it’s unlikely to work may become something of a self-fulfilling prophecy. However unlikely a given protest is to achieve its specific aims, no-one doing anything at all is surely even less likely to nudge the future of your society in the direction you believe it should go. Simply, if no-one believes things can change, things are less likely to change.

Of course, this kind of critique cannot be made of people whose motivation for not protesting comes from them spending their time working towards the cause in other ways. The sensible prioritisation of these differing efforts is something that one hopes the data and associated literature on these topics can continue to help establish.

An article from the Stanford Social Innovation Review for instance makes an initial attempt to compare the impact of funding an environmental protest movement vs funding an environmental charity or using a carbon offset scheme as measured by the amount of carbon dioxide removed from the atmosphere per dollar spent. They come out strongly in favour of the protesters.

More from Vaclev Havel, this time as found in the 1986 book “Disturbing the Peace” (via Wikiquote):

Hope, in this deep and powerful sense, is not the same as joy that things are going well, or willingness to invest in enterprises that are obviously heading for success, but rather an ability to work for something because it is good.

When a person tries to act in accordance with his conscience, when he tries to speak the truth, when he tries to behave like a citizen, even in conditions where citizenship is degraded, it won’t necessarily lead anywhere, but it might. There’s one thing, however, that will never lead anywhere, and that is speculating that such behavior will lead somewhere.