Collective Action and Individual Impact, Part II
Cross posted at zachgroff.com
In my last long post, I argued that if we can greatly help a promising collective action, we are obligated to do so. I did not argue that any particular protest has or will be effective, although I offered a few illustrative examples. I thank Carl Shulman and Ben Kuhn for their helpful responses to my original post. In this post, I would like to argue that the habit of supporting social movements is a good one to have and offer a few examples.
To start, let’s consider a few problems where protesting could likely make a difference. It seems to me protests rely on having a problem that is fairly concrete and direct, with an identifiable institution at fault. Poverty in developing countries strikes me as lacking in a clear target for a protest, and existential risk strikes me as too abstract and, again, lacking in a clear target.
These are problems that severely affect large numbers of people and are at least somewhat tractable:
-Mass incarceration
-Climate change
-Global institutional reform relevant to poverty
-Immigration and open borders
-Animal agriculture
Now let’s look at the costs and benefits of participating in a protest. Carl Shulman writes, “But to be convincing you have to actually show your work that the returns on protesting really are competitive with the opportunity cost of time (e.g. earning and donating some money, studying to advance your career and ability to get things done, doing malaria research, resting up after all of the above).”
Though time is approximately fungible, it seems likely that people sort their time, like money, into budget categories. Given that, we must ask which category protesting is likely to come out of. It seems unlikely it would come out of one’s job and more likely to come out of either resting up or self-improvement.
A basic economic approach would value leisure time at the amount one could earn in that time, but going to a protest doesn’t obliterate leisure time—it simply replaces the alternative activity with going to a protest. This will be costlier for some people than for others. If marching is a chore, this cost could approach or even surpass the forgone earnings, but if marching is energizing, or even just benign, the cost could be fairly low. The cost will also, of course, depend on how important your being well-rested is—getting less sleep to march in a protest would be quite costly for a life-saving surgeon.
To get more specific, again quoting Carl Shulman, “If the tens of millions of person-hours spent on the protests were spent at minimum wage jobs and the proceeds donated, then tens of thousands of African lives would have been saved; if one can earn higher wages or contribute in other scarce ways, the opportunity cost will be larger still.”
I assume he is going with GiveWell’s estimate that $3,340 given to the Against Malaria Foundation can, on expectation, save one life. 300,000 people at, say, the People’s Climate March, had they spent five hours at a minimum wage job instead of marching would have earned enough to save about 3,000 lives. As stated above, I don’t think that working is the relevant alternative—protesting would most likely detract from mentally allotted leisure time. Still, even if leisure time is worth 1/10th the amount of time that work time is (and the people marching work minimum wage jobs on average), that march had an opportunity cost of 300 lives that could have been saved, a pretty hefty opportunity cost.
How unfavorable is this comparison? Well let’s turn to the benefits.
The benefits of any one action are far more diffuse and less measurable than the sort of interventions effectiveness-minded people like to favor. Clearly, the benefits will vary greatly by protest. Each year there are many, many protests with little effect. As someone who grew up outside of Washington, D.C., I can report that every time I visited the White House, I saw a protest outside for a cause that would likely be forgotten.
Yet every so often there is a protest with a profound effect. I cited the People’s Climate March in my original post, and that protest, which appeared on the front page of the New York Times, seems to have catalyzed one of the liveliest student protest movements in years, which has kept climate change in headlines across the country and perhaps led to Senate Democrats’ scuttling of a bill to approve the Keystone XL pipeline.
Meanwhile, the Black Lives Matter marches have made racism in the criminal justice system one of the principal political issues of 2015, and while police in the U.S. kill 1,000 people each year, the U.S. prison population numbers above 2,000,000, which is significant considering the potential tractability of the problem and the misery of being imprisoned, is a worthwhile problem.
But beyond this anecdotal evidence, there is a fairly strong scholarly consensus that protests work, spanning political science, history, and economics. Specifically, political scientists’ estimates of the effectiveness of nonviolent protest movements range from 37% to 75% of such movements achieving their goals.
Of course, that protest movements work does not tell us how well they work, which is what we need to compare participation in them to the alternative. A highly rigorous study of the Tea Party movement estimated that an additional protester at a Tea Party rally produced 12 extra votes for the Republican party (the question of whether voting matters may be the subject of a later blog post). There aren’t many other estimates like this in the literature.
There are many reasons the Tea Party estimate might be an over-estimate for a standard protest, even more so a large one. For one, the Tea Party protests were fairly small, so an additional person likely yielded larger returns. For two, there’s a case to be made that the Tea Party protests happened at a favorable time politically with the pendulum swinging back from extremely large expectations for the incoming President.
So let’s take that estimate and divide it by 10. Suppose each person at the People’s Climate March produced an extra 1.2 votes in favor of climate action. That’s 360,000 extra votes for climate action. Some very rudimentary back-of-the-envelope math using estimates by Nate Silver of the chance that a vote swings an election yields an estimate of a .6% chance of swinging a Presidential election. If swinging the election improves the chance of climate action by 50%, then this march would increase the chance of climate action by .3%. Now if we use as a ballpark estimate the EPA’s estimated net global benefits of climate regulations ($67 billion), this march would, on expectation, yield an expected $201 million in benefits—enough to save 60,000 lives.
Many of my assumptions may have inflated this estimate—chiefly, the assumption that the likelihood of each vote swinging the election is independent. I also based each person’s wage on the minimum wage—clearly many people will earn several times the minimum wage. On the other hand, one could argue that I made a conservative estimate of individual impact, and that I assigned to each marcher the average likelihood of swinging an election, which should underestimate the impact since marchers who are from swing states have a higher chance of swinging the election by an order of magnitude. If the march had 1/200th the impact, it would still outweigh the opportunity cost.
There are some further problems, though. We just looked at aggregate or average impact, rather than marginal impact. It is likely that the 300,000th marcher does not have a marginal impact equal to the average impact. The relationship between marginal impact and number of marchers is quite unclear. It seems most likely to me that marginal impact would have an S-shape: protesting by yourself or with a few of your buddies likely has very little impact because of the strength (or weakness) in numbers. An additional person would probably do much more good for a small- or medium-sized march.
Given this, it seems to me to be roughly a toss up that the average person in the People’s Climate March would have much of an expected impact, let alone a high-impact person with potential for higher wages and other factors. But this march was likely not all that high on the effectiveness scale. For a medium-sized march around a potentially higher-impact issue like, say, open borders or animal agriculture, it seems plausible to me that attendance at a march would be high-impact. Given that we are habit-forming creatures, I think an effective person ought to make a habit of showing support for small- to medium-sized protests around highly important causes.
The case is stronger as one moves higher up on the collective action ladder. Organizing a protest, or helping to organize one, can mean getting many more people out. The opportunity cost is of course larger as well, but turning out a large enough number of people with a fraction of your effectiveness can create a multiplier effect. People within the effective altruism movement have pushed for more movement-building and advocacy for their multiplier effects. There are possible (and promising) alternative proposals to political protests—pledge parties to pledge to donate 10% of one’s income with Giving What We Can, for instance—but these seem better for consolidating and utilizing a community than building one. There is something about the raw emotional effect of being part of a crowd united around a cause that can change people in a way more coolheaded activities cannot. Scholars who have studied protest movements speculate that this psychological effect is what helps generate and sustain activists.Again, whether I can have a big impact by organizing collective action depends quite a bit on my own attributes. But given that one of the most vital ingredients in a protest is the structure that transforms it from an event into a movement, if I have this ability, I could be doing significant damage by sitting out.
At the end of the day, the question of collective action and its importance comes down to the question of emergence. Many world-changing events have been caused by masses acting together. Few world-changing events have been caused by individuals acting alone. We are individuals, and we cannot directly control anyone but ourselves. But when we look only at the visible results of individual actions, we miss the more diffuse effects our actions have on those around us—the way our actions spread and shape social norms. When you look at the effects of people acting together, though, mass political action may be more effective from a rationalist perspective than it seems at first.
Thanks for putting in this effort, and upvoted for taking up the empirical task, Zach. My main thoughts, adding on to those from my comment in the last thread:
These examples are mixing cause selection (malaria bednets vs climate) and the tactic of protesting vs donation/lobbying/research/writing and other alternative things to spend time on
Estimates of the effectiveness of attending protests derived from the Tea Party rain study include follow-on effort in donations, volunteering, attending further protests, etc; so you can’t substitute the cost of attending one protest for an hour (including travel time) for all the many hours and dollars a Tea Party activist might have contributed
I remain open to the idea that political action does better than AMF; indeed, I think it likely that the best political interventions for improved immigration and foreign aid will do better than AMF in terms of QALYs
Even for areas where I suspect politics is the best approach, that doesn’t mean that protesting is competitive with donation: funding think tanks like the Center for Global Development, lobbyists, public interest law firms, and social scientists, or even paying for protests are all competing with spending time protesting, and division of labor tends to pay off
It might help to do a head-to-head comparison within the same cause, e.g. your DxE protests vs a donation of the cash value of a typical EA DxE protester’s time to one of ACE’s top charities (or ACE), or an hour of the time of someone working for one of those charities; or contrast the climate march with donations to thinks tanks and lobby groups working on climate
Responses to particular points in the piece are below:
I was.
The analysis I did assumed complete success in eliminating all police shootings without doing any harm. Accounting for the chance of failure, or converting to a concrete goal, e.g. mandatory bodycams for police, would reduce estimated cost-effectiveness.
There is a switch here from the marginal person at a Tea Party rally to your attendance at a single rally. Presumably much of the impact comes from people who attended the Tea Party rally on a non-rainy day following up by volunteering, donating, etc. Those are substantial additional costs. But then if you spend the day protesting but don’t do all those follow-up activities, you won’t get the same benefits, and so the cost-effectiveness will fall, perhaps greatly if the impact came mostly from some attendees doing vigorous activity later.
This looks off by orders of magnitude.
Most of this economic value is health benefits to richer people, and won’t be spent on malaria bednets.
The EPA says the plan would involve a cut of 730 million metric tons of CO2, and various benefits to Americans from reducing local emissions of particulate matter, soot, and smog from coal plants. The latter would save some thousands of American lives (from the whole program, not the part attributed to the protest), and save hundreds of thousands of work-days otherwise lost from asthma or other health problems. The EPA values American lives at $7.4 million, but the cost of saving a life in Africa is thousands of times less than that.
For the 730 million tons of carbon, again the social cost of carbon is based on economic output and WTP rather than DALYs, so losses to rich countries play a disproportionate role. Giving What We Can thinks that the direct DALY effects of buying carbon credits are orders of magnitude worse than their recommended charities. Using those estimates of DALYs per ton of carbon, we would again get lives saved orders of magnitude smaller.
That was my point, that the money donated to the protests and the value of person-hours spent on them, applied to cost-effective foreign aid charity, would save more lives than eliminating all those police shootings in the United States for decades, which would be a surprisingly large effect on shootings.
Capturing media and political attention for one issue to some extent pulls it away from other issues. Progressives might instead have been more focused on climate or mass incarceration or immigration or education.
Mass incarceration, and the mandatory minimum laws and long sentences that increase prison time per crime, involve many more DALYs than police shootings and have simpler mechanistic responses available (pulling back on minimum sentences, as in the recent California initiative). I think that’s one reason why, for instance, GiveWell has focused on mass incarceration rather than police killings. But the protests have been focused on police killings rather than on prison.
The aggregate effects of systematic harassment, stop-and-frisk, disrespect of the citizenry, assault, and other police misconduct are plausibly larger than shootings, and the remedy of police bodycams would help to relieve those as well as illicit shootings (which are a subset, to eliminate all shootings would also require things like addressing the availability of guns). In Ferguson the Department of Justice found that the shooter should not have been indicted, but that excessive fines and other systemic problems with the Ferguson Police Department’s policies and police behavior demanded its intervention.
I could be convinced that protests about police shootings will have a large impact on sentencing or other harms that fall under criminal justice broadly rather than police shooting specifically, but I’d want to see more evidence showing how resources mustered for the one are transferring to the others.
I’d then want to characterize those benefits more realistically, rather than using bounds like “all police shootings cease for decades after 1 year of protests,” and consider costs like focusing the capturing the public and progressive agenda on this issue rather than others, or the billions of dollars of economic damage affecting poor neighborhoods from rioters exploiting the non-violent protests as cover.
Looking at just the evidence you’ve presented it still looks like it would be better to donate to AMF than to the protests, and that a modest opportunity cost of time or the price of public transportation or a Lyft fare would produce more QALYs/avert more DALYs applied to AMF.
I still don’t see it. If the numbers in your examples were more favorable they would work fine without worries about emergence from a consequentialist perspective. When one takes into account diminishing marginal returns and things like the Gelman et al. paper on probability of tied elections that you cite above, expected value analysis on your individual action works fairly well.
Thanks for your comments, again. Very helpful, and good to have this discussion. Regarding your bullet points:
1) I thought I delineated this pretty clearly, but the effectiveness of protesting definitely depends on the cause. 2) This is correct—but I’m not sure this is all that crucial, since presumably the same increase in commitment to the cause would happen to someone else. An effective altruist would hopefully direct this in even more productive ways. The counter-argument is that EAs may already be motivated enough that this doesn’t matter. I’m unsure on that question. 3-4) Agreed, and you may be right. I would like to see that comparison done. 5) This might have been a better route to go, although I suspect there is room for protesting and social movement organizing around open borders, mass incarceration, and other important causes, which was why I made it more general. I will probably give a more specific treatment at some point.
Regarding much of your post, I agree that the crux of the issue is whether protesting around these shootings affects mass incarceration. I would think this moves the debate forward on mass incarceration fairly strongly, but I could be wrong. I didn’t provide much evidence on this question since it wasn’t the primary focus of the piece.
Yes, I don’t mean to say you didn’t acknowledge things differ by cause, just that without a benchmark of non-protest activities in the cause it’s hard to show the advantage of protesting as a tactic, and that I think the mixed examples haven’t shown an existence proof because of cause selection issues.
Is the idea here that one goes to the protest and then donates to climate action instead of AMF in the future, possibly donating more in total? I.e. it’s like volunteering ineffectively for a charity to build affiliation with it? Or that if you attend you will cause others to commit and so will get almost the same gains even if you don’t put in further resources?
I imagine a protest where 1000 people attend and one where 1001 people attend. The extra impact could a) come mostly from the chance that the extra person will become entangled with the movement and invest more time, money, and social capital later; b) mostly come from a diffuse effect where each person who attends is more likely to become committed the more people who attend; c) some other effect like media coverage.
B) seems a bit strange because each attendee can only interact with so many other attendees, and it would imply more superlinear returns than the Tea Party study suggests. However, I could see that the impact of attendance might be as high as an exogenous attendee, even after subtracting out the average later donations and commitment (which would make the cost-benefit for you much worse), if you were much more likely to cause other attendees to commit their time and money. Basically, marketing to the enriched audience of protest attendees.
Sure.
Oh, sorry—the first quote was very unclear. I meant that in EA attending a protest would presumably also become more invested in it like people in the Tea Party study. That could be a point against climate action, as you say, but if I already felt, say, open borders was the most important cause and I went to a protest for that, it would deepen my investment.
OK, I understand now. I think that there is a huge difference between:
1) Spending 4 person-hours on activity X will change the outside world to produce Y QALYs.
2) Spending 4 person-hours on activity X will lead you to spend 40 person-hours to produce Y QALYs
If I don’t think Y QALYs/40 hours is cost-effective enough to want to pursue, then I don’t want to do 2), even if Y/4 hours would be enough.
It’s great to have this debate explicitly, so thanks for producing this!
That said, I think your estimates err significantly on the optimistic side. A couple of specific points:
This is quite extreme—your subsequent calculation shows that’s an absolute rather than relative 50%. This would be thinking that a single election could move the global likelihood of climate action (assuming full action with success, and zero otherwise) from 25% to 75%, according to who wins. It looks more than an order of magnitude too optimistic to me at this step.
Careful. An important fact is that money goes differently far in different contexts, and the figure that you are using for “enough to save 60,000 lives” represents an extremely good use of money. The benefits under discussion will not be distributed so as to all go on such cases. In fact many of them are health benefits which have been converted into a dollar value (I couldn’t find the conversion rate on a skim read, but I can guarantee that it will be a lot more expensive than $3,500 per life—probably between 1 and 3 orders of magnitude more, depending on the country they benchmark from).
Thanks for pointing this out. I agree it is good to have this debate explicitly.
Regarding point one, it doesn’t seem all that extreme to me, but maybe I’m wrong. Republican administrations have rolled back regulations on climate and Democratic administrations have expanded them, and it doesn’t seem like there’s all that much variation in this.
The second one is a good point, though it’s tough to figure out the right approach to do this quantification, since it is the global cost of carbon, so it’s not only American lives involved in the quantification (http://reep.oxfordjournals.org/content/7/1/23.full.pdf). In addition, at the risk of getting into another thorny debate, it leaves out the costs for wild animals, which I would guess would be at least an order of magnitude worse than those for humans. I’m going to make another post to address this. On balance, this seems to me to move participation in this march from ‘toss up’ to ineffective, but I don’t think that generalizes to causes that are potentially easier to impact or smaller marches.
Good spot. A single election result should produce <1% change of global climate change action, right? Changing 10% of USA senators to democrat could arguably cause like 1% change in that probability....
The CBA was for the US EPA actions announced by President Obama (although future Presidents will be able to undo them, so it’s questionable to count it as ‘done’ by a single President).
If it were intense long-term global climate action to keep below 2 degrees Celsius of warming, the estimated benefits would be measured in trillions, not billions.
If someone earning to give made 10x the minimum wage, and didn’t enjoy protesting (so it came out of their work time, not leisure time), then your estimate of the benefits of non-participation is already off by a factor of 100.
Add in that the case for the climate march is way less robust than the case for e.g. donations to AMF (problems: do the estimates for the Tea Party generalize to estimates for Democrats? Do protests about climate change, rather than partisan protests, actually affect the Presidential election?), and the case for non-participation seems much stronger for a large group of your audience.
Regarding the first, that’s why I included the caveat here: “Given this, it seems to me to be roughly a toss up that the average person in the People’s Climate March would have much of an expected impact, let alone a high-impact person with potential for higher wages and other factors.”
I think on other issues (e.g. animal agriculture) this may still be overwhelmed. Also, thanks for your comments, as usual.
Yeah, you said it was a toss-up after making a bunch of extremely favorable assumptions (other folks in this thread have pointed out other corrections, the most important probably being that dollars of environmental action != dollars to GW top charities). My point is that if you relax the extremely favorable assumptions to anything more realistic, the case for the climate march doesn’t seem strong at all. Not just “a toss-up” but a lot weaker.
Maybe the case is stronger for other causes, but you didn’t really talk concretely about those, so I can’t comment.
Interesting article (and comments)!
If someone was an organiser of such a rally, it would be quite feasible that each hour they invest into the rally would cause 10 hours (or, in some circumstances, several magnitudes more) worth of other people marching. The organiser makes a much bigger difference with their time.
So in my opinion, even more interesting than the question “Should we attend rallies?” is the question “Should we organise rallies?”
A very interesting post (as was the previous one). On this point:
I expect in fact it’s a large march where they make the most difference.
I model my guess at the likely curve shape and resulting benefit in a recent post, Rationality of demonstrating & voting.
I wish posts like this came with a complementary spreadsheet for the empirics.
Do you mean “ought” in the same sense of “morally ought” that you meant in your previous post about collective action? If so, I think you need to rise to a much higher standard of proof to be convincing about this. So far, all you’ve shown is that a highly-speculative quantitative analysis of a single protest shows that it might plausibly be competitive with not protesting, setting aside any regression from the fact that the estimate has probably four orders of magnitude of variance. You don’t provide very much evidence for the assumption of an S-curve of marginal effectiveness (or any arguments about where the inflection points are); or any argument that the highly important causes are equally easy to influence (for instance, swinging a presidential election probably has much smaller implications for factory farming than for climate action).
That’s how I meant ought, given a consequentialist view along the lines of Singer’s Famine Affluence, and Morality or Shelly Kagan’s The Limits of Morality. I’m not sure the uncertainty problem is unique to this form of action—if the evidence for this action is speculative or uncertain, then that makes the opportunity cost for the alternatives uncertain. Uncertainty is infectious, and I don’t think that undoes our moral commitments.
I think you interpreted my parent comment as saying:
That’s not what I meant; sorry if I wasn’t clear. I meant that you haven’t provided enough evidence to move my prior on participating in protests from “not very effective” to “very effective”.