Mini Post #8: Welcome to the #Resistance
Mini Post #8: Welcome to the #Resistance
Does protesting work?
I’ve decided to try something new: once a week, to write a shorter, narrower post focusing on one specific paper. Last week, I wrote about whether sexist people like feminists movies, and whether an area’s traditions affect its taste in movies. This is the tab with all the previous (and current) posts.
Onto the actual post: does protesting increase support for the protesters’ cause?
There are tons of people protesting around the world as of late: large numbers of people protesting repressive legislation in Georgia (the country, not the US state - shoutout to my friend Sandro Sharashenidze who’s a good source on this), people protesting the government’s fiscal policy in Argentina, farmers protesting in Europe for not being given even more money, pro-Palestine protesters shutting down roads across the US, and I assume many more. Whether you support or oppose each of these movements, the question they pose is broadly similar: does protesting help increase support for the relevant issue?
“Protesting = political support” seems like a no brainer, but not necessarily. First, it could be that the people who attend a protest aren’t random “normies”, but rather, diehard supporters of the cause at hand, and this “show of force” by movement X doesn’t do anything to convince non-believers. Secondly, it could be that by protesting (a disruptive and headline-grabbing technique), the protesters draw more attention to their movement, and thus increase polarization around the issue. Thirdly, it could be that “neutrals” on the topic are actually swayed by seeing how popular a position the protesters have, but their support isn’t persistent or doesn’t impact their broader political attitudes.
This week’s paper is hot off the NBER presses (it literally came out yesterday, April 18th 2024): “Social Movements and Public Opinion in the United States” by Amory Gethin and Vincent Pons. The question tackled is simple: looking at multiple generally progressive mass protest movements in the United States from 2017 to 2022, the authors try to ascertain whether the protests changed broader public opinion, and whether they changed voting behaviors in particular (given the protests were generally associated with left of center causes, then they should have led to greater Democratic Party vote share or turnout in the three subsequent elections).
The protests in question were: four Women’s Marches (2017 to 2020), the 2017 March for Science (pro climate action), the 2018 March for Our Lives (pro gun control), 2018 protests in support of Robert Mueller’s investigation of Donald Trump, the 2019 Climate Strike, protests against Trump’s migration policies in 2017 (“Muslim Ban”), 2018 (anti family separation), and 2019 (“lights for liberty”, broadly pro migrant events), protests in favor of impeaching Donald Trump in 2019, relatively small protests in January 2020 in response to Donald Trump’s Iran policy, and lastly, the Black Lives Matter protests of 2020 after the death of George Floyd.
What data do they use? First, there’s how many people attended each protest in each county, which comes from a proprietary source called the Crowd Counting Consortium (CCC) from 2017 to 2022. The authors classify each large gathering of people based on general cause (gender, migration, “party politics”, environmental, etc.) and ideological lean (liberal/conservative), as well as matching them with a national movement - so, for instance, local climate protests in Ann Arbor, Michigan in 2017 don’t get included in the broader national count.
Secondly, there’s measures of national US interest on a topic in both Twitter and Google. Relying on the Twitte API (calling it Twitter and tweets because the sample runs out in 2022) and on Google Trends data, they look at how intensely users were looking up terms linked to the claims and demands of each movement - so, for example, people who tweeted “gun control now” or googled “mass shooting” in the days before and after the 2018 March for Our Lives. To measure opinions on topics, the authors look at a series of surveys (the The Democracy Fund + UCLA Nationscape survey, the Gallup Poll Social Series, and YouGov’s Cooperative Congressional Election Study) and recode the questions on topics asked to match the protest categories - so, for instance, Nationscape asks 4 questions about gun control, so in 2018, the authors would look at movements in replies to the questions relevant to the March for Our Lives.
The autors have several different empirical strategies: the first is a simple difference in how liberal/conservative people’s views are, and much people are tweeting or googling the cause in question. The simple difference simply measures liberalism/conservatism and tweet intensity before and after the protests. The authors regress the relevant data on the timing of major movements (i.e. whether one was happening or not), controls for the individual’s ideology, employment status, and demographics (race, education, age, gender, religion), and interactions between various terms - for example the March for Our Lives could have increased opposition to gun control among conservatives more than it raised support among liberals. They also use a similar methodology to test for changes in ideology immediately after and before protests and not movements. One last concern is whether support for positions came from what motivated a protest and not the protest itself - for example, it could be that the March for Our Lives caused higher support for gun control, or it could be that the 2018 Parkland high school shooting caused that support. To address this, they also analyze only protests that didn’t respond to specific events, aka “independent protests” such as the March for Science, the Climate Strike, and the four Women’s Marches - the first was on Donald Trump’s inauguration and the other three were on various dates that were generally not related to the issue in particular, but rather to the Marches’ own timing.
The results for the simple difference design are fairly straightforward: first, the number of tweets and Google searches sharply increases immediately after a protest, even when there is some evidence that interest increases before the protests take place. These effects largely decrease over the following ten days, when they return to previous levels. For “independent” protests, these effects are smaller but still significant. Comparing searches about the protests in particular for both independent and regular movements, the authors find a large discrepancy, signaling that independent movements simply receive less attention. For changes in opinion, the evidence is rather weak: there are no significant effects on public opinion for either independent or “regular” protests in any dataset, although it may be possible that the share of people expressing “Not Sure” as an opinion declines. Looking at political behavior (election turnout, Democrats’ vote share, and approval of Donald Trump) the simple differences measurement shows little change in progressive turnout or vote share, and no change in approval - although, for independent movements, there is a small decline in turnout and a small uptick in approval, showing perhaps a backlash.
The second methodology is differences-in-differences. The question isn’t just to compare changes in opinions before and after protests, but also comparing changes in opinions between places that had bigger and smaller protests - so it’s not just whether you had a Women’s March, but whether having a Women’s March in your area caused changes versus places that didn’t. The regression is similar: change in salience or attitude on whether a county had a protest and how big the protest was, as well as controls for time effects, county-level effects (a place where Trump got 80% of the vote would have smaller pro-impeachment protests), demographics for individuals, interactions between a variety of controls. The results for the differences-in-differences approach are similar to previously mentioned ones: online activity and salience grows greatly after the protests and then declines. For political opinions, there is no evidence that two months after a protest, the share of people holding an opinion on the issue increases - meaning that it’s not especially clear that protesting sways “undecideds” one way or another. Looking at the number of people holding liberal views, there is no evidence of medium run change: that is, people don’t get more liberal nor more conservative after major protests. There are also no result on likelihood of voting, Democrat vote share, or (dis)approval of Donald Trump.
The conclusion is that, over the Donald Trump’s (hopefully only) term in office, liberal-leaning protests did not result in any changes to political attitudes, behaviors, or ideologies, and did not increase the notoriety of a political topic more than a few weeks or months. However, this was not true for every movement: some were able to raise the salience or liberalism of voters in certain specific measurements, but not across specifications. The only exception were the 2020 George Floyd-related Black Lives Matter protests, which were successful in shifting political views, attitudes, and behaviors: over the same year, racial issues became much more important, and people became persistently more liberal on race and persistently likelier to support liberal politics. Exactly why, other than 2020 BLM’s enormous salience in social and traditional media, is unclear.
To finish the post, some links
The paper in question
A similar paper from 2022 by Erica Chenoweth et al. examining 2020 protests (both BLM and anti-COVID) to find the determinants of participation and of impact.
A 2021 paper by Magdalena Larreboure and Felipe Gonzalez that finds that the first Women’s March actually managed to increase female election turnout and female support for the Democrats.
A 2013 paper by Stan Madestam et al. that finds that attendance to 2010s Tea Party protests (broadly anti-Obama fiscally conservative rallies) benefitted the Republican Party.
A 2019 paper by Ro'ee Levy and Martin Mattsson finding that the Me Too movement managed to increase reporting and punishment of sexual malfeasance
A 2019 paper by Raquel Fernández, Sahar Parsa, and Martina Viarengo that finds that states with higher AIDS death rates became more supportive of LGBT rights, due to having more active LGBT communities during the epidemic.
A blog post by Matt Yglesias about how abortion rights activists can utilize protests to advance their cause, considering salience vs popularity of the issue.
One thought on the conclusion, that: "racial issues became much more important, and people became persistently more liberal on race and persistently likelier to support liberal politics. Exactly why, other than 2020 BLM’s enormous salience in social and traditional media, is unclear." This seems to line up with the idea that short-term and medium-term political changes come mostly from issue salience. Protests don't change opinions, but if covered in the news, can change what people are thinking about. Hence why anti-Trump protests probably didn't change many minds, because he already dominated the conversation.