Will protests lead to thousands of coronavirus deaths?

post by Larks · 2020-06-03T19:08:10.413Z · EA · GW · 59 comments

Prior to the recent protests, the US seemed to be making some progress on controlling coronavirus, with an r0 of probably around 0.9, a slowly declining number of cases, and many states starting to ease lockdown. In particular, restrictions on large gatherings helped significantly slow the spread, because they reduce both the number of infected people who can spread it and the number of new people who can become infected. One (BERI funded!) study suggested that banning large gatherings reduced r0 by around 28%.

Unfortunately, protests seem in many ways ideal for spreading the disease. They involve a large number of people in a relatively small area for an extended period of time. Even protests which were advertised as being socially distanced often do not end up that way. While many people wear masks, photos of protests make clear that many do not, and those that are are often using cloth masks that are significantly less effective than surgical or n95s in the face of repeated exposure. Additionally, protests often involve people shouting or chanting, which cause infectious droplets to be released from people's mouths. Exposure to tear gas can apparently also increase susceptibility, as well as cramped indoor conditions for those arrested.

It's hard to estimate how many new cases will be caused by the protests, because there doesn’t seem to be good statistics on the number of people at protests, so we can't model the physical dynamics easily. A simple method would be to assume we have lost the benefits of the ban on large gatherings over the last week or so. On the one hand, this may be an over-estimate, because fortunately most people continue to socially distance, and protests take place mainly outside. On the other hand, protesters are actively seeking out (encouraging others to seek out) boisterous large gatherings in a way they were not pre-March, which could make things even worse. On net I suspect it may under-estimate the incremental spread, but given the paucity of other statistics we will use it as our central scenario.

If the r0 was around 0.9 before, this suggests the protests might have temporarily increased it to around 1.25, and hopefully it will quickly return to 0.9 after the protests end. Even if we assume no chain infections during the protest - so no-one who has been infected at a protest goes on to infect another protester - this means the next step in disease prevalence would be a 25% increase instead of a 10% decrease. Unfortunately the exponential nature of infection means this will have a large impact. If you assume around 1% of the US was infected previously, had we stayed on the previous r0=0.9 we would end up with around 9% more of the population infected from here on before the disease was fully suppressed. In contrast, with this one-time step-up in r0, we will see around 12.5% of the population infected from here - an additional 3.5% of the population.

Assuming an IFR of around 0.66%, that's a change from around 190,000 deaths to more like 265,000. Protesters skew younger than average, suggesting that this IFR may be an over-estimate, but on the other hand, they are also disproportionately African American, who seem to be more susceptible to the disease, and the people they go on to infect will include older people.

So it seems quite plausible to me that the protests might cause around 75,000 thousand additional Americans to die of covid, including probably over 10,000 African Americans. Additionally, despite the massive reduction in international travel, there will probably be some spread to other countries, especially if they have copy-cat protests.

There is clearly a lot of uncertainty about this number - my guess is that the actual r0 impact may be higher and longer lasting, but the IFR may be lower. If r0 is sufficiently low the disease will be suppressed more quickly afterwards, potentially making these numbers significantly too high (if the r0 were as low as 0.7, the protests would only cause an incremental 20,000 deaths). If r0 is a little higher then these numbers will be a substantial under-estimate. On the other hand, if the r0 was sufficiently high then containment will inevitably fail, and almost everyone in the US will catch it regardless of protests!

However, it seems hard to avoid the conclusion that the incremental number of deaths is quite likely many thousands. (In contrast, the Washington Post estimated that US police killed 41 unarmed people in 2019, of which 10 were black.)

My guess is that the indirect and statistical nature of these deaths makes people less sensitive to them. Probably if they were more emotionally salient, many protesters would not be willing to so endanger their lives.

As a result it seems to me that, even ignoring injuries and fatalities directly incurred during the protests, and the damage caused by associated looting, that the protests are quite bad, and it would be good if instead people stayed home to save lives and protect their local health systems.


I have not edited any of the prior sections

So how close were the predictions in this post? It's very hard to judge, because the main predictions were about the delta between protests and no protests, rather than an absolute forecast. But we can say a few things based on what has happened since the 3rd of June.

I modelled US cases as basically on a declining trend ( r0=0.9 ), which would then see a short one-time increase in r0 to 1.25, and then return to a decline with r0=0.9 . Overall the US numbers would remain on an exponential decline with a one-time step up from the protests.

What actually happened? At time of publication, the 7-day average of daily new cases was 22,000. A month later it had risen to 60,000, and would continue to rise until peaking at almost 70,000 in July. This is a much larger increase in cases than I was expecting, and it coincided almost exactly with the start of the protests.

How much of this was a direct result of the protests? Unfortunately this is hard to tell. My best guess is 'some but not most'.

At the same time as the protests a lot of states were undoing their lockdowns - in my opinion prematurely - which would naturally cause an increase in disease transmission. It is possible that protests increased this effect, by undermining the credibility of pro-lockdown experts for reasons I outlines here [EA(p) · GW(p)]. In some cases the government's support for these protests forced them to relax other covid restrictions, as they were found guilty of discriminating in favour of the protests. However I suspect a significant degree of liberalisation would have occurred even without the protests.

One strong piece of evidence would be contract tracing; unfortunately in some cases contract tracers were told by pro-protest politicians not to ask people if they had been to a protest. Public health officials from some cities said that protests caused covid spread; officials from other cities deny it.

The average age of infected people has fallen significantly. This is the age profile for people who go to protests, but also the profile for those who go to bars and clubs. It has also significantly reduced the realised IFR.

A NBER paper from Dave et. al. argues that places with protests saw large offsetting reductions in social activity: the more people protesting/rioting, the less other people wanted to go outside. I didn't consider this sort of behavioural response in the article. To the extent it is true, it means that while protesters did impose costs on others, these costs were borne in the form of more time staying indoors rather than covid transmission.

This study uses newly collected data on protests in 315 of the largest U.S. cities to estimate the impacts of mass protests on social distancing, COVID-19 case growth, and COVID-19-related deaths. Event-study analyses provide strong evidence that net stay-at-home behavior increased following protest onset, consistent with the hypothesis that non-protesters’ behavior was substantially affected by urban protests. This effect was not fully explained by the imposition of city curfews. Estimated effects were generally larger for persistent protests and those accompanied by media reports of violence.


Comments sorted by top scores.

comment by Robert_Wiblin · 2020-06-04T11:01:53.448Z · EA(p) · GW(p)

I think a crux for some protesters will be how much total damage they think bad policing is doing in the USA.

While police killings or murders draw the most attention, much more damage is probably done in other ways, such as through over-incarceration, petty harassment, framing innocent people, bankrupting folks through unnecessary fines, enforcing bad laws such a drug prohibition, assaults, and so on. And that total damage accumulates year after year.

On top of this we could add the burden of crime itself that results from poor policing practices, including a lack of community trust in police due to their oppressive behaviour and lack of accountability.

Regardless of where a consequentialist analysis would come down, it is a tragedy that people feel they need to choose between missing an opportunity to fix a horrible system of state violence, and not spreading a dangerous pandemic.

Replies from: Larks, willbradshaw, Halstead
comment by Larks · 2020-06-06T03:52:44.128Z · EA(p) · GW(p)

I think this is the wrong question.

The point of lockdown is that for many people it is individually rational to break the lockdown - you can see your family, go to work, or have a small wedding ceremony with little risk and large benefits - but this imposes external costs on other people. As more and more people break lockdown, these costs get higher and higher, so we need a way to persuade people to stay inside - to make them consider not only the risks to themselves, but also the risks they are imposing on other people. We solve this with a combination of social stigma and legal sanctions.

The issue is exactly the same with ideologies. To environmentalists, preventing climate change is more important than covid. To pro-life people, preventing over half a million innocent deaths every year is more important than covid. To animal rights activists, ending factory farming is more important than covid. To anti-lockdown activists, preventing mass business failure and a depression is more important than covid. But collectively we are all better off if everyone stops holding protests for now.

The correct question is "is it good if I, and everyone else who thinks their reason is as good as I think this one is, breaks the lockdown?" Failure to consider this, as it appears most people have, is to grossly privilege this one cause over others and defect in this iterated prisoners dilemma - and the tragic consequence will be many deaths.

Replies from: Rook, Halstead, matthewp
comment by Rook · 2020-06-06T16:41:53.506Z · EA(p) · GW(p)

This was a very clear and valuable comment for me. Strongly upvoted.

However, you could argue, given the current political momentum of the BLM protests, there's a unique reason to support those now over protests that support other causes. BLM protests today may be able to encourage criminal justice reforms that (1) could last for a very long time (2) wouldn't be possible in the future (or would be significantly more difficult), when there's less political momentum behind criminal justice reform.

comment by Halstead · 2020-06-08T12:05:56.470Z · EA(p) · GW(p)

I think this argument conflates the fact that p, and people's belief that p. Consider these two principles

1. If people correctly believe that going on the protests produces more good than harm, then they should go on the protests.

2. If people believe that going on the protests produces more good than harm, then they should go on the protests.

Principle 1 seems to me clearly correct from a utilitarian point of view. Principle 2 is absurd - people can have mad and false beliefs. If someone believes that going on a neo-nazi rally is going to produce greater marginal benefits than staying at home, that doesn't mean that they should in fact break the lockdown. The proposition "The BLM protests will produce more good than harm" doesn't entail principle 2.

(I'm not saying that the protests do in fact produce more good than harm, I'm just criticising Larks' counter-argument in the above comment)

comment by matthewp · 2020-06-09T20:12:56.041Z · EA(p) · GW(p)
But collectively we are all better off if everyone stops holding protests for now.

Who is the 'we' here and by whose yardstick the benefit measured?

Animal rights activists are not turning out in large numbers to get tear gassed and beaten for the cause. This is pretty good evidence that they are not in the set of 'everyone else who thinks their reason is as good as I think this one is'.

As usual, there are better alternatives being neglected here. Those who want more lockdown have, in this situation, two options to get it: more violence or more concessions.

Negotiation is certainly possible. So, a consequentialist might lay additional covid deaths at the step of a government which failed to negotiate.

Add to this the obvious virtue of the demand to end police brutality and recognize that black lives matter. That being an option now, it seems particularly bizarre, and wrong, to delay granting the wish.

Replies from: Dale, aarongertler
comment by Dale · 2020-06-10T15:42:10.798Z · EA(p) · GW(p)
Who is the 'we' here and by whose yardstick the benefit measured?

Investigations into police brutality that follow viral footage have historically been quite harmful for all involved. The upside is a small reduction in police brutality. The downside is a massive increase in non-police brutality, as found in this recent paper:

all investigations that were preceded by "viral" incidents of deadly force have led to a large and statistically significant increase in homicides and total crime. We estimate that these investigations caused almost 900 excess homicides and almost 34,000 excess felonies. The leading hypothesis for why these investigations increase homicides and total crime is an abrupt change in the quantity of policing activity. In Chicago, the number of police-civilian interactions decreased by almost 90% in the month after the investigation was announced. In Riverside CA, interactions decreased 54%. In St. Louis, self-initiated police activities declined by 46%. Other theories we test such as changes in community trust or the aggressiveness of consent decrees associated with investigations -- all contradict the data in important ways.

Indeed the harm done by one day of reduced policing in Chicago may have already rendered the protests a net negative, even ignoring spreading Coronavirus:

From 7 p.m. Friday, May 29, through 11 p.m. Sunday, May 31, 25 people were killed in the city, with another 85 wounded by gunfire, according to data maintained by the Chicago Sun-Times.
In a city with an international reputation for crime — where 900 murders per year were common in the early 1990s — it was the most violent weekend in Chicago’s modern history, stretching police resources that were already thin because of protests and looting.
The Rev. Michael Pfleger, a longtime crusader against gun violence who leads St. Sabina Church in Auburn Gresham, said it was “open season” last weekend in his neighborhood and others on the South and West sides.

I also think you misunderstand your fellow EAs:

Animal rights activists are not turning out in large numbers to get tear gassed and beaten for the cause. This is pretty good evidence that they are not in the set of 'everyone else who thinks their reason is as good as I think this one is'.

Many animal rights activists believe that the status quo is far far worse than the holocaust. There are billions of animals being farmed for meat today, generally treated very cruelly. Whatever you think of the state of US race relations, it is clear that, if animals matter, they are much worse off - both much more numerous and treated much much worse!

I think what you are missing is that there are factors other than believed importance of cause that determine one's actions. For example, animal rights activists might care about suppressing the pandemic! Or they might think getting tear gassed was counter-productive!

You suggest that concessions will help reduce the scale of the protests, but my impression is that the literature suggests that actually repression is effective. For example, this study on the 2011 London Riots, where first-time looters were punished relatively harshly, found it was successful in reducing crime:

The criminal justice response was to make sentencing for rioters much more severe. We show a significant drop in riot crimes across London in the six months after the riots, consistent with a deterrence effect from the tougher sentencing. Moreover, we find that non-riot crimes actually went in the opposite direction, suggesting a response from criminals who look to have substituted away from the types of crimes that received tougher sentences. We find little evidence that spatial displacement or extra police presence on the streets of London in the wake of the riots accounts for these patterns of change. More evidence of general deterrence comes from the observation that crime also fell in the post-riot aftermath in areas where rioting did not take place.

Similarly, this study on Israeli counter-terrorism police:

An increase in repressive actions leads to a reduction in terrorist attacks. ... An increase in conciliatory actions has no effect on terrorism.

Finally my guess is that this is sort of irrelevant anyway because OP is probably not a senior government official; she may be able to persuade some friends not to go protest, but probably can't change US policy.

Replies from: matthewp
comment by matthewp · 2020-06-10T22:55:05.750Z · EA(p) · GW(p)
You suggest that concessions will help reduce the scale of the protests, but my impression is that the literature suggests that actually repression is effective.

Presented with options to get largely non-violent protestors-for-justice to go home quickly:

a) Justice

b) Repression

Your response is that b) is a tried and tested intervention. Seriously?

That is not the path to human flourishing.

Replies from: aarongertler
comment by Aaron Gertler (aarongertler) · 2020-06-11T08:49:00.034Z · EA(p) · GW(p)

I don't think this is an accurate portrayal of what Dale was trying to say.

I don't see them actively recommending a particular policy in the post -- just noting that some studies of repressive behavior find that it may lead to a certain outcome. It can be true that repression sometimes quells riots while also being true that it has many other negative outcomes and should clearly be avoided. (Though I didn't see Dale say that, either, and I don't want to put words in their mouth.)

Of course, the vague term "repression" and the differing social context of the examples Dale cited mean that blanket statements like "literature suggests that repression is effective" aren't very useful, and I wish they'd acknowledged that more clearly in their post, especially given the awful consequences of policies like "harsher prison sentences for a lot of people."


As for the claim that "justice" will clear up protests quickly; leaving aside the question of which specific demands will have a positive impact on their own merit (likely many), have we seen enough demands granted so far to have a sense of what usually happens after vis-a-vis public protest? Especially in cases where actually following through on promises of change will take a long time?

The clearest example of responsiveness to protest I can recall (haven't been following the topic too closely) was action taken by the Minneapolis City Council to ban certain restraint practices and explore "dismantling" the police department. Did either action lead directly to a reduction in public protest?

comment by Aaron Gertler (aarongertler) · 2020-06-09T21:30:52.153Z · EA(p) · GW(p)

Negotiation is certainly possible. So, one might lay additional covid deaths at the step of a government which failed to negotiate.

Even if it isn't difficult to cast blame at one's government, this doesn't mean much for the people who have died. It also seems unlikely that governments are going to feel much additional pressure from deaths for which they bear only indirect responsibility.

I don't have any developed opinion on the original post, but I did want to take mild issue with the idea of thinking about deaths as a bargaining tool. (I'm sure you meant for this to be a neutral/factual point about negotiating, but it's hard for me to shake off the devastating impact of additional deaths.)

comment by willbradshaw · 2020-06-04T12:09:47.944Z · EA(p) · GW(p)

I suspect that a lot of protesters would be very angry we're even raising these kinds of issues, but...

If we're being consequentialist about this, then the impact of the protests is not the difference between fixing these injustices, and the status quo continuing forever. It's the difference between a chance of fixing these injustices now, and a chance of fixing them next time a protest-worthy incident comes around.

Sadly, opportunities for these kinds of protests seem to come around fairly regularly in the US. So I expect these protests are probably only reducing future injustices by a few years in expectation. Add to that the decent chance that the protests don't achieve very much[1], and it might be even less.

Normally, of course, it would be well worth it. But if it's true that mass protests during a pandemic will cause many thousands of deaths, then the above reasoning becomes pretty important.

Regardless of where a consequentialist analysis would come down, it is a tragedy that people feel they need to choose between missing an opportunity to fix a horrible system of state violence, and not spreading a dangerous pandemic.

This is certainly true.

  1. In particular, it's important to ask how a pandemic affects the chances of success. If it decreases them (say, because people are unusually unsympathetic to people seen as irresponsibly crowding together) then the expected value of these protests (relative to waiting) falls. If it increases them (say, because politicians and public authorities are unusually keen to resolve the crisis and get people off the streets) then that would be a counterargument to my claims here. ↩︎

Replies from: alexrjl, Robert_Wiblin
comment by alexrjl · 2020-06-04T13:48:35.109Z · EA(p) · GW(p)

A couple more weak arguments pointing towards protesting now:

  • if protest success depends on number of people, the fact that many people are currently out of work may make current protests more likely to succeed.

  • if protests are about demonstrating strength of feeling, then saying (with one's actions) 'I care enough about this to risk a deadly disease' arguably does that pretty effectively.

  • There's also the question of whether protesting meaningfully affects the election, if so the effect size of that could dwarf everything else, but I don't have any idea which way it cuts and can see good arguments in both directions.

Replies from: willbradshaw
comment by willbradshaw · 2020-06-04T13:53:55.877Z · EA(p) · GW(p)

Thanks, this is exactly the kind of response I'd like to see.

I agree that the first point points in a pro-protesting direction. The second might also but I am uneasy about it (for a young person, most of the impact of their getting sick is infecting others, so the actual message is "I care enough about this to risk giving others a deadly disease", which is somewhat less attractive). I agree that the third could go either way.

(Notably, the third point makes rioting an even more terrible idea than I already thought it was.)

comment by Robert_Wiblin · 2020-06-04T14:15:09.774Z · EA(p) · GW(p)

I didn't mean to imply that the protests would fix the whole problem, obviously they won't.

As you say you'd need to multiply through by a distribution for 'likelihood of success' and 'how much of the problems solved'.

Replies from: willbradshaw
comment by willbradshaw · 2020-06-04T14:51:38.755Z · EA(p) · GW(p)

Sure, I didn't think you were saying that the protests would be a panacea. My main point was less about probability/degree of success and more about counterfactual impact.

comment by Halstead · 2020-06-08T10:54:47.036Z · EA(p) · GW(p)

These points don't apply to the UK and elsewhere to anywhere near the same extent, so the post does at least seem like a good argument against the protests in the UK and elsewhere.

Replies from: Sean_o_h, HaukeHillebrandt
comment by Sean_o_h · 2020-06-09T07:54:02.849Z · EA(p) · GW(p)

There are other factors (relating to points made in the post) to suggest the protests in UK and EU may carry less risk comparatively. Police tactics at protests in different countries may be a relevant consideration - e.g. the heavy use of tear gas in the US (bad for spreading covid, as larks notes) isn't happening in the UK. R0 also a relevant consideration - likely much lower in many european countries now than in many parts of the US.

Replies from: Halstead
comment by Halstead · 2020-06-09T10:12:09.284Z · EA(p) · GW(p)

Is the R0 lower in Europe? The 3 day average deaths per million is similar in the US, UK, Sweden, and (eg) Italy. Fatal shootings by police officers of citizens per capita are about 170x lower in the UK than the US. Imprisonment as % of the population is about 20x higher in US than US. Prison conditions seem far worse in the US than UK

Replies from: Sean_o_h
comment by Sean_o_h · 2020-06-09T10:32:12.060Z · EA(p) · GW(p)

I would expect deaths to be on a lag (it takes a few weeks on average for people to get sick enough to die). At a quick glance, France, Spain, Italy and Germany are reporting an average of well under 1k new cases a day for the last 7 days, compared to 19-25K/day for the US (obviously necessary to correct for the USA having a 5-6x larger population than these countries).


Edit: this site estimates R0 as being 1.02 in the US overall, and <1 in all western and northern European countries (although >1 in several eastern european countries)


comment by HaukeHillebrandt · 2020-06-08T15:53:12.809Z · EA(p) · GW(p)

Contra: https://www.theguardian.com/commentisfree/2020/jun/07/britain-is-not-america-but-we-too-are-disfigured-by-deep-and-pervasive-racism

Interesting stats on police violence in the UK:


Might suggest that the benefits of protesting in the UK and elsewhere outweigh the costs of virus spread especially given the differential state of the pandemic.

Replies from: Halstead, Dale
comment by Halstead · 2020-06-08T19:09:29.797Z · EA(p) · GW(p)

I don't think this is a counterpoint to my claim which was that the problem of state mistreatment of black people is considerably lower in the UK vs the US. I didn't claim that there wasn't unfair mistreatment of black people in the UK.

comment by Dale · 2020-06-08T18:13:54.927Z · EA(p) · GW(p)

These articles do not appear to contradict what Halstead said at all.

The first link appears to be an opinion piece rather than a serious piece of analysis - for example it does not include any comparison of the rates of Police killing between the UK and the US. It complains that UK police haven't been found guilty of murdering black men for a long time, but does not compare this to the number of unarmed black men shot by cops in the UK - a number which is approximately zero most years! It mentions that black men are imprisoned at higher rates than white men in the UK, but does not compare this to the rate at which they commit crimes, which is also significantly higher. Indeed, the only time it actually makes a direct comparison between the US and UK it actually (begrudgingly) agrees with Halstead:

Few people would deny that in many respects life is better for non-white people in the UK than in the US.

Overall I would not consider that article to be a particularly serious analysis of the issue.

Your second link (which I see you found by following a link in the Guardian article) is significantly more data-orientated, but again the only time it directly touches on the issue at hand it seems to agree with Halstead:

14% of deaths in police custody or otherwise following contact with the police since 1990 were BAME. This is proportionate to the population as at the 2011 census.

Finally, neither article contains any comparisons to the pandemic.

comment by AGB · 2020-07-08T18:05:55.898Z · EA(p) · GW(p)

For posterity, the only data I've seen on this question suggests that this has not played out the way the OP and many others (myself included) might have expected. The economist ran an article* which links to this paper**. In short, cities with protests did not record discernible COVID case growth, at least as of a few weeks later. Moreover, quoting the paper (italics in original):

"Second, where there are social distancing effects, they only appear to materialize after the onset of the protests. Specifically, after the outbreak of an urban protest, we find, on average, an increase in stay-at-home behaviors in the primary county encompassing the city. That overall social distancing behavior increases after the mass protests is notable, as this finding contrasts with the general secular decline in sheltering-at-home taking place across the sample period (see Appendix Figure 6). Our findings suggest that any direct decrease in social distancing among the subset of the population participating in the protests is more than offset by increasing social distancing behavior among others who may choose to shelter-at-home and circumvent public places while the protests are underway. "

In other words, it seems that protestors being outside was more than offset by other people avoiding the protests and staying home.

* https://www.economist.com/graphic-detail/2020/06/30/black-lives-matter-protests-did-not-cause-an-uptick-in-covid-19-cases

** https://www.nber.org/papers/w27408

comment by Timothy_Liptrot · 2020-06-03T20:53:41.960Z · EA(p) · GW(p)

I agree with your main claim that the protests will cause more deaths than they save, but I disagree with your estimate of caused deaths by at least an order of magnitude.'

The estimated change in reproduction number is not compelling. Firstly, the largest protest in American history had attendance of 3*10^6, or one in a hundred Americans. So assuming this protest is tied for most attended in US history, and other Americans behave the same, the protestors must each have an R of 30 to bring our average R from .9 to 1.25. That is 10 times the pre-SD reproduction number. Assuming the protestors have an R of 3 or 4 seems reasonable to me. Now that we know much transmission is from speaking, 10 would be the upper limit of my 95% CI. Assuming an R of the 3, reduce the estimate to 10%

Also the protests are short relative to the time a person is contagious. I doubt the protests will maintain attendance in the millions for more than two days, which is less than people are typically contagious. Therefore their influence on our effective reproduction number will be less than you estimate. Seems like people are contagious for at least 4 days, so call that a reduction by half. So reduce the estimate by a further 50%.

So I would reduce your estimate from 75,000 to 4,000. Which is still a lot. More than my expected benefit from the protests by movement building.

Replies from: Sean_o_h, Larks
comment by Sean_o_h · 2020-06-03T21:08:08.672Z · EA(p) · GW(p)

I am also sceptical about the central scenario. Protests have disproportionately high visibility relative to the numbers participating. If the 28% reduction Brauner et al estimate was the result of banning protests, parades, sporting events, concerts, political rallies, beach gatherings, lectures, house parties, crowded gyms and restaurants, yoga classes, and many other categories of >10 people gatherings, plus the indirect effects (e.g. on public transport etc) of these measures, then it seems unlikely to me that the recent protests could have an equivalent reverse impact, despite how relatively widespread they have been.

Replies from: Larks
comment by Larks · 2020-06-04T02:35:18.705Z · EA(p) · GW(p)

I think many of those examples would fall under their other categories like "Most Businesses Suspended" or "School Closure". Things like 'beach closures' do not, but population density on beaches tends to be much lower than at protests (at least of the beaches I have been to).

Additionally, I worry that the protests might reduce other forms of social distancing. Imagine you are a moderate conservative, who had to cancel your son's graduation and your daughter's wedding, and hasn't been able to go to confession for months. You wanted to go back to work, but all the experts told you that it was too dangerous, even though you knew you'd be careful. They even stopped you playing golf - you weren't even allowed to do a couple of rounds by yourself, standing by yourself in the middle of the green! Now all of a sudden these so-called experts are joining the hippies in a chaotic screaming looting protest, with nary a six foot gap to be seen. How likely is it that you will trust them again?

comment by Larks · 2020-06-03T21:30:41.283Z · EA(p) · GW(p)

Thanks, good comment!

I spent a while trying to estimate the r0 in this way, thinking about the dynamics of protests. In the end I couldn't really come up with much confidence as any level of at-protest-r0; in particular it wasn't obvious why it couldn't be much higher, given the loud close contact between a very large number of people. It certainly seems plausible to me that an infected person could easily pass within one meter of a very large number of people. When I try to visualise the number of talking people I spent time close to pre-Covid, vs the number I would be close to at a busy protest, it doesn't seem implausible to me that the latter could be orders of magnitude higher. But I don't have any data on this so it is rather speculative!

Replies from: Timothy_Liptrot, Bluefalcon
comment by Timothy_Liptrot · 2020-06-04T00:04:31.278Z · EA(p) · GW(p)

This is an interesting point. The protests are happening largely outside so there is a further reduction, possibly itself an order of magnitude. So you really need two orders of magnitude to get one hundredth of Americans to contribute .3 to the reproduction number.

Imagine putting people into a room, and measuring the number of possible transmission paths. At one person in the room there are 0 possible transmissions. At two people, there is one possible transmission. 3 people, three transmissions. 4 people, 6 transmissions. The scaling is n! where n is the number of people in the room. So if protests involve a larger average number of people in breathing contact, then protesters might have an effective R ten times higher than our regular lives.

But that scaling only lasts out as far as the droplets spread. Do we model droplets as rays, equally likely to move in any direction? Then the droplet spread drops off by r^2 as the distance to another person. Then the connection factor stops scaling quickly because as people are packed in they become further away. I do not know when the n! scaling stops, but I imagine the number of protestors within 6 feet of one another is a good metric. The last protest I was at there were like 10 people within 6 feet of me. 10! is...

Oh my. Perhaps the R0 jump will be noticable.

Replies from: BenMillwood, Larks
comment by BenMillwood · 2020-06-06T10:59:38.231Z · EA(p) · GW(p)

The number of possible pairs of people in a room of n people is about n^2/2, not n factorial. 10^2 is many orders of magnitude smaller than 10! :)

(I think you are making the mistake of multiplying together the contacts from each individual, rather than adding them together)

Replies from: Linch
comment by Linch · 2020-06-06T18:33:52.609Z · EA(p) · GW(p)

lol I thought that 10! was a surprise, rather than a factorial...

comment by Larks · 2020-06-04T02:20:43.383Z · EA(p) · GW(p)

I was thinking about the outside issue. It seems in general this is quite protective, presumably because the wind blows the droplets away, rather than their being recycled in a largely air-tight room. But for a sufficiently large protest, presumably the wind is blowing them away... onto another part of the protest! So I worry that this factor will be less protective here.

Great explanation of the scaling issues, good way of thinking about it.

A protest near me had six foot markings on the ground to give each individual protester their own box... which was then ignored in practice.

comment by Bluefalcon · 2020-06-05T08:35:29.226Z · EA(p) · GW(p)

Look at other superspreader events, like large church choirs. Those are indoors, so probably worse than protests, but you can adjust for that.

Replies from: Larks
comment by Larks · 2020-06-05T12:11:52.728Z · EA(p) · GW(p)

Hey, thanks for this. Do you have any good data on the super-spreader events, and how to adjust for inside/outside? I agree that 'you' in a general sense can, but unfortunately this doesn't mean that 'I' specifically can!

Replies from: willbradshaw
comment by willbradshaw · 2020-06-05T13:26:41.428Z · EA(p) · GW(p)

Upvoted for this line, which made me laugh:

I agree that 'you' in a general sense can, but unfortunately this doesn't mean that 'I' specifically can!

comment by Juan Cambeiro · 2020-06-07T12:09:27.277Z · EA(p) · GW(p)

Trevor Bedford's back-of-the-envelope calculation estimates 200-1100 deaths per day of protests. Note that his analysis assumes an R0 of 0.95, which is definitely not a safe assumption. https://twitter.com/trvrb/status/1269533303536664576

Replies from: Sean_o_h
comment by Sean_o_h · 2020-06-09T10:51:08.384Z · EA(p) · GW(p)

He did another analysis assuming R0 of 0.9, obtaining estimate of 150-600 downstream deaths per day of protests. https://twitter.com/trvrb/status/1269715787482148864

Replies from: Juan Cambeiro
comment by Juan Cambeiro · 2020-06-09T16:28:52.302Z · EA(p) · GW(p)

I don't know why the discussion is being limited to the R0 being 0.90 or 0.95. Is it not plausible — indeed, likely — that the R0 might be 1, 1.1, 1.2 or even much higher? Anything above R0=1 would imply a cascade of transmission events that would result in far more deaths than any figures these analyses are showing, and it seems reasonable to be very concerned about that even if these transmission chains are stopped or burn out at some point in the near future.

Replies from: Sean_o_h
comment by Sean_o_h · 2020-06-09T17:04:41.291Z · EA(p) · GW(p)

R0 could certainly be much higher in principle, though if it is, it doesn't seem to be reflected in the number of positive cases being recorded in the US - which has been holding steady or slightly declining for the past month - or the number of deaths (declining, although there would be a lag there). These indicators could be misleading of course - the US, like the UK, is nearly certainly undertesting and undercounting cases. However, the number of tests has been going up, and if the number of cases being 'caught' isn't increasing this is some indication that R0 is somewhere a little below 1. So I would tentatively agree with the OP's suggestion here.


One reason looking at different assessments based on R0=0.9 is informative is that it illustrates just how high the uncertainty and room for error is in these sorts of analysis. The OP predicts 75,000 deaths as a plausible possibility; Bedford predicts 150-600/day, so 2,100 -8,400 over the course of 2 weeks of protests (assuming each day carries the same impact on R0, which is probably wrong); both using R0=0.9 as a central assumption, and recognising that the present R0 is a key factor. Because present value of R0 is such a critical factor, comparing different estimates at the same R0 makes it easier to compare/contrast.

One of my personal concerns is that the BLM protests may end up unduly scapegoated (in terms of their role being overestimated) for any increase in cases and deaths; the US administration has done quite a bit of scapegoating already in my view, and there are many ways in which its own response has been far from adequate. My intuition is that other aspects of states reopening prematurely are likely to play a bigger impact in a possible second wave. If there were a significant overestimation of the impact of the BLM protests for example, this would be bad not only for the BLM movement and antiracism in the US; it would also be bad in terms of understanding the other causes of increase of R0 and putting in the appropriate planning overall for preventing and responding to future waves. (Likewise however, if those of us who think the impact of the protests is lower than in reality are wrong, it would be good to update).

Replies from: Juan Cambeiro
comment by Juan Cambeiro · 2020-06-09T18:17:57.899Z · EA(p) · GW(p)

I'm not convinced that we would have already seen a significant uptick in reported/confirmed case numbers quite yet (weren't the largest protests this past Saturday?). The median incubation period is ~5 days, most people don't get tested at the time of symptom onset, and the PCR test turnaround time still seems to still be at least a day or two. Perhaps most importantly, most of the protestors seem to be relatively young and so many may be asymptomatic or may have mild cold/flu-like symptoms. I'm more interested in (and concerned about) any secondary transmission events that may involve older family members that protestors might live with/come into contact with. Many of these older folks would presumably have more serious symptoms and so would be more likely to show up in confirmed cases/hospitalizations data over the coming weeks.

I think perhaps the most informative case for the determining the effect of protests on overall transmission might be New York City. In recent days it has been at record lows in terms of confirmed cases/hospitalizations/test positivity rate, it has one of the highest number of tests per capita of any state/country, it has seen large protests, and it is has not reached the phase of reopening that involves activities known to be high risk (for now, it's just construction/manufacturing and curbside pickup — no schools, dining in restaurants, bars, etc.). I'm keeping an eye on this Metaculus question (which has already been mentioned elsewhere): https://pandemic.metaculus.com/questions/4590/how-many-confirmed-new-covid-19-cases-per-day-will-there-be-in-new-york-city-for-the-week-of-june-15th-to-june-21st/

As for scapegoating, I agree that it's likely and it's extremely unfortunate. I think few (if any) of us would disagree that this is a just cause. But it is nevertheless concerning that there are mass gatherings in the middle of a pandemic, whatever the reason for them might be. A big concern of mine is that the credibility of messaging by the public health community may have taken a huge hit. Many public health professionals were (rightly) telling everyone to stay home as much as possible and then seemingly disregarded their own advice to participate in/support the protests. If (in my view, when) a second wave comes, will public health messaging to again stay home be viewed through a far more partisan lens than has already been the case?

Replies from: Sean_o_h
comment by Sean_o_h · 2020-06-09T18:26:04.600Z · EA(p) · GW(p)
I'm not convinced that we would have already seen a significant uptick in reported/confirmed case numbers quite yet (weren't the largest protests this past Saturday?). The median incubation period is ~5 days, most people don't get tested at the time of symptom onset, and the PCR test turnaround time still seems to still be at least a day or two. Perhaps most importantly, most of the protestors seem to be relatively young and so many may be asymptomatic or may have mild cold/flu-like symptoms. I'm more interested in (and concerned about) any secondary transmission events that may involve older family members that protestors might live with/come into contact with. Many of these older folks would presumably have more serious symptoms and so would be more likely to show up in confirmed cases/hospitalizations data over the coming weeks.

Right. But with regard to R0 =0.9, I understand R0=0.9 was being used as the background R0 prior to the impact of the protest, rather than the R0 following the impact of the protests (if 'background' R0 is <1, then the impact of an R0-increasing event/set of events will have a lesser effect than if 'background' R0 is >1). It may be the case, as you suggest, that R0 has increased significantly since the start of the protests until now (whether due to the protests or in combination of other factors), in which case protests right now are happening against a higher R0 than these estimates assume - but we don't have the data. I agree that NYC will be interesting.

Replies from: Juan Cambeiro
comment by Juan Cambeiro · 2020-06-09T18:38:46.884Z · EA(p) · GW(p)

In an earlier Twitter thread, Trevor Bedford says "my rough guess would be that an infected individual would on average transmit to one further individual each day in the protest setting" — so I think he was using 0.9/0.95 as the R0 to estimate the impact of the protests. https://twitter.com/trvrb/status/1269395234762285056 But I don't think this is a safe assumption and I'm concerned that the R0 might be significantly higher than ~1 in a mass gathering.

In any case, the most reliable modeling I've seen (and that Trevor cites) estimates that the current overall background R0 in the U.S. is already back above 1. :/ https://covid19-projections.com/#current-us-projections

Replies from: Sean_o_h
comment by Sean_o_h · 2020-06-09T18:47:51.311Z · EA(p) · GW(p)

Thanks Juan, I hadn't seen that most recent R0 estimate you link to - concerning.

comment by Aaron Gertler (aarongertler) · 2020-08-17T11:44:02.204Z · EA(p) · GW(p)

I don't think this has been posted as a comment yet, so I'd like to link this study (shared with me by Hauke Hillebrandt [EA · GW]) which estimates the impact of protests on COVID-19 spread. The abstract (emphasis mine):

Sparked by the killing of George Floyd in police custody, the 2020 Black Lives Matter protests have brought a new wave of attention to the issue of inequality within criminal justice. However, many public health officials have warned that mass protests could lead to a reduction in social distancing behavior, spurring a resurgence of COVID-19. This study uses newly collected data on protests in 315 of the largest U.S. cities to estimate the impacts of mass protests on social distancing, COVID-19 case growth, and COVID-19-related deaths. Event-study analyses provide strong evidence that net stay-at-home behavior increased following protest onset, consistent with the hypothesis that non-protesters’ behavior was substantially affected by urban protests. This effect was not fully explained by the imposition of city curfews. Estimated effects were generally larger for persistent protests and those accompanied by media reports of violence. Furthermore, we find no evidence that urban protests reignited COVID-19 case or death growth after more than five weeks following the onset of protests. We conclude that predictions of population-level spikes in COVID-19 cases from Black Lives Matter protests were too narrowly conceived because of failure to account for non-participants’ behavioral responses to large gatherings.

(I haven't read the full paper and don't have further commentary to offer.)

Replies from: Larks
comment by Larks · 2020-09-05T21:49:04.356Z · EA(p) · GW(p)
I don't think this has been posted as a comment yet, so I'd like to link this study (shared with me by Hauke Hillebrandt [EA · GW]) which estimates the impact of protests on COVID-19 spread.

Thanks. I think this paper was actually already linked in a comment by AGB here [EA(p) · GW(p)]; I've also discussed it in the retrospective part of the post.

comment by willbradshaw · 2020-06-04T12:20:09.764Z · EA(p) · GW(p)

I think this kind of argument is broadly true (though the true magnitude of the effect is, as you say, very uncertain), but I think it's important to note that these kinds of arguments also apply to politicians, administrators, and the police.

If you could reasonably predict that incidents like this would lead to mass protests (which you could, because it's happened before), and that this could result in a severe increase in the number of pandemic deaths, then you have a duty (even more so than usual) to make sure this does not happen. As an administrator, you should put extra pressure on police to make sure these incidents don't take place. As a police officer, you should take extra care that it doesn't happen on your watch, and that your colleagues know what the kinds of consequences could be. As a politician, you should be making it clear that if something like this happens, heads will very quickly roll.

(Of course, all these people should have been doing this anyway, because incidents like this are a profound moral stain on American institutions. You could reasonably argue that if these groups were receptive to these kinds of arguments, they wouldn't be causing these incidents in the first place. But even given the status quo, everyone in law enforcement and administration should have been taking extra super special care right now.)

comment by judyfletcher · 2020-06-20T02:41:47.439Z · EA(p) · GW(p)

It's been two weeks since the post was written. Any update / data on whether or not the protests ended up increasing coronavirus deaths?

comment by PeterMcCluskey · 2020-06-04T15:41:53.201Z · EA(p) · GW(p)

Another risk is that increased distrust impairs the ability of authorities to do test and trace in low-income neighborhoods, which seem to now be key areas where the pandemic is hardest to control.

Replies from: willbradshaw
comment by willbradshaw · 2020-06-04T15:44:41.024Z · EA(p) · GW(p)

True, though the loss of trust seems to fall more on the authorities than the protesters to me.

comment by kbog · 2020-06-11T03:00:00.657Z · EA(p) · GW(p)

Here's my cost-benefit analysis. (I also posted it to my shortform, but I don't see a way to link directly to a shortform post.)


Replies from: Habryka
comment by Habryka · 2020-06-11T04:05:00.368Z · EA(p) · GW(p)

The timestamp with the small link icon is a link to the shortform comment itself. Here is the link for your shortform comment that I got this way:

https://forum.effectivealtruism.org/posts/siuZkxobSGtEqtwq2/kbog-s-shortform?commentId=fHxmf3G7Qn6H6pZpb [EA(p) · GW(p)]

comment by Sanjay · 2020-06-10T15:14:37.586Z · EA(p) · GW(p)

Thanks for this, very interesting.

Sorry if I missed it, but does this analysis factor in the fact that, barring a vaccine/cure happening soon, we would expect all those catching the disease now will still get the disease, only later?

Unflattening the curve is clearly a bad thing, but does this analysis acknowledge that this is an unflattening the curve effect as opposed to people-getting-the-virus-and-otherwise-they-wouldn't-have effect?

Replies from: Larks
comment by Larks · 2020-06-11T19:26:26.685Z · EA(p) · GW(p)

Good question!

I think of there being basically three extreme possibilities:

  • Really low R0
    • We successfully suppress the disease after the protests. As such, while protest-driven infections are a larger percentage of the total, the total number is much smaller, so it doesn't really matter very much. This is basically the R0 = 0.7 case I mentioned.
  • High R0, no vaccine in time
    • Basically everyone gets the disease. As such the protests and other re-opennings have limited direct impact on the number of cases, as you mention. The impact is largely limited to accelerating this, with some effect on hospital capacity and less time to learn about better treatments.
  • High R0, mass vaccination in medium term
    • The number of cases keeps growing, then suddenly falls when a vaccination is rolled out. In this case, accelerating the spread is basically the same as delaying the vaccination. Because of the nature of exponential growth, the majority of cases will be just before mass vaccination, so this leads to a dramatic increase in the total number of deaths. (This might be slightly offset by the fact that a higher incidence makes it easier to do clinical trials on vaccines, but I would expect this effect to be small).

In the modelling I assumed an R0 < 1, which is basically a less-extreme version of the first scenario.

comment by bfinn · 2020-06-12T20:29:59.413Z · EA(p) · GW(p)

There's also the complication that by deciding to protest, individuals incur a non-negligible personal cost (in time and risk), but only make a tiny difference to the size & hence effectiveness of the demonstration. Also any benefit arising mostly accrues to others. All that on top of the risk of you spreading the virus to others.

All told, it's far from clear it's worth people's while to demonstrate, even for major issues like this one. It depends on things like the size of the demonstration and your degree of altruism. I did a rough model of this a while back (excluding the virus spreading affect):

https://www.lesswrong.com/posts/juusvkkMvuC5D7Een/rationality-of-demonstrating-and-voting [LW · GW]

comment by etphone27@hotmail.com · 2020-06-04T00:57:16.135Z · EA(p) · GW(p)

Interesting post, although your post actually allayed my concerns rather then exacerbate them. From the post or comments it doesn't seem that you tried to quantify the number of lives saved by the protest. Obviously this is hard to do prospectively but it seems no harder than estimating the increased spread. For instance, if one outcome of these protests are that the http://useofforceproject.org/ policies are adopted then that might lead to 72% fewer deaths at the hands of police. Last year approximately 1000 people were killed in the US by police, meaning that after approximately 100 years the gains in your worst case scenario would be eroded. There could be many other outcomes, for instance lowering cycles of incarceration that cause much anguish aside from death. Of course this is all hypothetical but that is what we are dealing in here.

Also, a minor point is that the age of the protestors skews young, potentially meaning that the same death rate for the population cannot be applied.

comment by irving · 2020-06-03T21:40:04.354Z · EA(p) · GW(p)

It's worth distinguishing between the protests causing spread and arresting protesters causing spread. It's quite possible more spread will be caused by the latter, and calling this spread "caused by the protests" is game theoretically similar to "Why are you hitting yourself?" My guess is that you're not intending to lump those into the same bucket, but it's worth separating them out explicitly given the title.

Replies from: Larks, Pablo_Stafforini
comment by Larks · 2020-06-04T02:03:22.538Z · EA(p) · GW(p)

Thanks for your comment! I actually discussed the maths of this a bit in person, but omitted it from the article for simplicities sake, and because I don't think it affects the conclusion much; it is essentially another causal channel by which the protests could increase transmission.

I am sceptical that total transmission-once-arrested cases will be anywhere close to transmission-on-streets. For the period they are arrested they'll be in close quarters, so it's definitely true that that is bad, though the total number of people they interact with will presumably go down, which will be a positive. But most importantly I expect only a very small fraction of protesters to be arrested. Indeed at some protests not a single person has been arrested! Furthermore, I expect that anyone reading this article (or anyone being influenced by someone who has read this article) is significantly less likely than average to be arrested, so it is at least less relevant from the point of view of their personal decision making.

I sort of see your point about the game theory, but I am sceptical that "the police will have to treat me nicely because otherwise I will get coronavirus" will work in practice. Similarly, I don't recommend trying deterrence with the IRS, or the SEC, or many other US government agencies; they have quite credible pre-commitments to ignoring your threat.

Replies from: irving
comment by irving · 2020-06-04T15:36:12.413Z · EA(p) · GW(p)

Thanks, that’s all reasonable. Though to clarify, the game theory point isn’t about deterring police but about whether to let potential arrests and coronavirus consequences deter the protests themselves.

comment by Pablo (Pablo_Stafforini) · 2020-06-03T21:49:02.147Z · EA(p) · GW(p)
It's quite possible more spread will be caused by the latter

What do you mean by 'quite possible'? And what's your estimate of the minimum ratio of arrests to protesters needed for spread due to arrests to exceed spread due to protests?

Replies from: irving
comment by irving · 2020-06-04T15:38:50.668Z · EA(p) · GW(p)

“Quite possible” means I am making a qualitative point about game theory but haven’t done the estimates.

Though if one did want to do estimates, that ratio isn’t enough, as spread is superlinear as a function of the size of a group arrested and put in a single room.