Hacker News new | past | comments | ask | show | jobs | submit login
A Superspreading Event: The Sturgis Motorcycle Rally and Covid-19 [pdf] (iza.org)
58 points by localhost 20 days ago | hide | past | favorite | 96 comments

Figure 7 of the paper appears to show a large spike in cases in the "high inflow" counties, at 18-23 days and 24+ days post-Sturgis start (Aug 3).

However, a quick review of the maps in Figure 1 identifies those counties as Clark, NV (Las Vegas), Maricopa, AZ (Phoenix) and Jefferson/Arapahoe/Weld, CO (Denver Metro). None of those counties' own data reporting seems to match up with this supposed spike in cases that the paper suggests, starting around Aug 21.




Clark and Maricopa are clearly heading downwards. The Denver Metro graphs are a little more vague, you could maybe suggest there has been an uptick in cases but you'd be hard-pressed to definitively say there is a "spike" that contrasts with the long-term trend.

I'm having some difficulty squaring the paper's case data with the county governments' own counts.

If you look at the parameters for the "valuation" of a SARS-CoV-2 infection, they are hard to take seriously. They are based on a mapping of typical workplace injuries to viral infections. This is the paper referenced: http://ftp.iza.org/dp13632.pdf.

A purely asymptomatic infection is valued at $11,000

A symptomatic but non-hospitalized infection is valued at $33,000

If you are in the hospital for any reason with an infection, $512,000

If you are in the ICU for any reason with an infection, $1,100,000

If you are in the ICU on a ventilator, $2,900,000

> they are hard to take seriously.

Why? What convention would you use?

If you think these are estimates of related healthcare costs, you should probably re-read the paper.

I would not value the public health cost dollar estimate of an asymptomatic infection at 11k. I would not value the public cost dollar estimate of a non-hospitalized infection at 33k.

They do have some discussion in there about the option of valuing asymptomatic cases at $0. So it's not like they didn't think of this option. But they went with $11,000 as the default option in their model, anyway. This leaves me thinking we may be looking at a Chesterton's Fence situation: the reason isn't obvious to me, but that doesn't mean it doesn't exist or isn't a good one.

I'm guessing there is some subtext here that an armchair analyst isn't going to catch from reading just this paper. Perhaps, for example, the cost of asymptomatic cases is there because "asymptomatic" doesn't actually mean "no symptoms at all", it means "the patient didn't subjectively feel ill enough to seek care." In which case there probably is some real cost to factor into the average. Or perhaps the argument is that asymptomatic cases still carry some risk of secondary effects such as myocarditis that absolutely need to be taken into account and averaged into the group from a public health perspective.

It's hard to say for sure, since the paper clearly isn't written for a non-expert audience and therefore doesn't spend much time on defining jargon.

I've seen reporting recently that studies showed some large percentage (63%?) of asymptomatic patients had visible signs in chest x-rays. So asymptomatic doesn't seem to always mean "absolutely no measurable effects".

How does that compare to x-rays of people with the common cold or the standard flu?

I'm probably going to perma-delete my HN account after this, but this is my main complaint with popularized scientific research being parroted for political purposes. "The paper isn't written for a non-expert audience" yet one of the authors is going on Anderson Cooper tonight to talk about it to a non-expert audience.

Wouldn't that be the best way to communicate this to a non-expert audience? Bring the expert in to explain the report and fill in the context that the layman is missing compared to an expert?

The same researcher can speak to a popular and expert audience. And that researcher should absolutely say things in the paper that they don't say in the popular outlet, and vice versa. They're different audiences with different needs and different background knowledge.

Compare the experience of reading one of Einstein's scientific papers with the experience of listening to one of his interviews. It's almost like they're not even the same person.

That's the difference between a paper and someone explaining it to the layperson.


The authors thoroughly outline their approach (and its limitations) based on available literature.

What specific criticism do you have?

If you read the literature, they compare a symptomatic viral infection to having your hand or foot broken, or a pelvic fracture, or a mild to severe head trauma. I do not see any evidence of those things being equivalent.

For the person who suffers them, they are not equivalent. But that isn't what is being measured.

When measuring economic impact, they could very well be equivalent. A dollar lost when someone can't work due to a broken bone is economically equivalent to a dollar lost when someone bedridden with a virus.

The paper doesn't make those things equivalent.

Best case, your comment is an oversimplification of what the authors spent a dozen or so pages explaining. The limitations they describe perhaps encompass some of your concern. They also describe alternative calculations.

On the surface it doesn't seem like a bad comparison. Out of work for weeks, potentially protracted recovery including possible long term limitations. What's the problem?

Downvoted because this is not an argument, just a statement that you don't like their estimate. That doesn't contribute to the discussion. If you're asymptomatic and you make someone else sick and they die, the cost is a lot more than $11k.

How many subsequent infections are caused by an asymptomatic infection? How much effort is expended to prevent those infections, by the asymptomatic individual and others?

"We first use the Department of Transportation (2016) guidance on value per statistical life (VSL) and severity/injury estimates as a basis for our non-fatal valuations by category. After updating the figures for earnings and inflation the DOT guidance recommends using a VSL of about $11 million in 2019 dollars. We use the severity classifications in the DOT guidance as a basis for our non-fatal valuations. DOT (2016) recommends using six different severity categories in benefit-cost analyses including Level 1 (minor), which corresponds to using a 0.3 percent amount of the VSL, Level 2 (moderate), which uses about a 5 percent amount,Level 3 (serious), which uses about a10 percent amount,Level 4 (severe), which uses about a 27 percent amount fraction,Level 5 (critical), which uses about a 59 percent amount,and Level 6 (unsurvivable) which uses a 100 percent amount(the full VSL). We therefore value asymptomatic cases at about $11,000 (in 2019 dollars) each which corresponds to using a 0.1 percent amount of the VSL in DOT (2016)."

0.1% of a statistical life is about a month? It's also a third of the "minor" classification.

How would you value it? $0? Clearly, that's not right.[1] Less than $11,000? More?

[1] Consider a new infectious disease which has only minor symptoms immediately, but may have unknown health consequences in the future.

Despite the criticism, I think you’re asking a great question. These figures do appear surprising and I think a lot of people will feel the same skepticism. I’d hope that someone can help us understand the paper rather than chastise you for being impertinent. Thanks for reading past the headline and provoking some discussion on the details.

Simply repeating your disagreement is no substitute for a counter-proposal.

Why not? The US GDP dropped by over 2 trillion in just the second quarter. That's over $300,000 per infection right there. And that's if we don't count future quarters, Q1 2020, or any of the trillions in stimulus that has been pumped into the economy. It doesn't seem these numbers are out of line with the back-of-the-envelope math.

You’re off by a factor of 4. US GDP is a little over $20T/year, and dropped a little under 10% 2020q2, so a drop of about $500B.

Ah, I was looking at an annualized dollar amount. Still, it would be a reasonable back-of-the-envelope number considering that the economic impact is still present and is going to continue into the future at least a year.

Are you from America? Those numbers look pretty reasonable for US healthcare bills.

I didn't realize I would get a bill when not seeking medical intervention at all, especially for $11,000 or $33,000.

This isn’t an attempt to estimate a typical bill for an asymptomatic individual. It’s an attempt to set an expected value for the costs that will eventually be incurred across the population of asymptomatic patients. They are using a mean to describe how costs will scale with the size of that population (asymptomatic SARS-CoV-2 positive) and NOT as a way to describe the typical costs of individuals in that population. The cost distribution is almost certainly a power law. I’d guess the median is zero and the mean is many thousand dollars.

As others have said, there are unknown but already detected cases of asymptomatic impacts. One football team doc said he thought 1/8 had some impact on heart function long term. Those who see this as no big deal for some reason are stuck on arguing the exact percentage of impact.

Several studies have shown that asymptomatic infections may have long-term impact. The valuation may be taking that into account -- you might not get that bill today, but maybe you (or your insurance) will pay that over the course of your lifetime

> Several studies have shown that asymptomatic infections may have long-term impact.

To the tune of tens of thousands of dollars on average? That's not plausible.

Sure it is. If 5% of asymptomatic cases require care that costs $200k, you come to $10k a person. This is entirely plausible in the US, particularly if you are amortizing the cost over the lifetime of the patient.

Can you cite any paper that suggests serious complications for asymptomatic patients is anywhere even remotely close to 5%? All I've seen is a handful of rare anecdotes, nothing to suggest it's a remotely likely outcome.

You're looking at actuarial estimates. Overestimating unknown-unknowns is more pragmatic than underestimating them. No, the article doesn't show how that particular sausage is made; we're all out here guessing.

"Furthermore, event organizers ... warned attendees that some “South Dakota tribal lands may be restricted with checkpoints to protect residents from COVID-19 exposure”"

I bet that was fun.

"Using a synthetic control approach, ... [f]or the state of South Dakota as a whole we find that the Sturgis event increase COVID-19 cases by 3.6 to 3.9 cases per 1,000 population as of September 2nd 2020. This represents an increase of over 35 percent relative to the 9.7 cases per 1,000 population in South Dakota on July 31, 2020 (South Dakota Department of Health 2020)."

That's pretty substantial.

The amount of travel involved in this thing is staggering:

> The cellphone tracking data showed about 10% of Sturgis attendees hailed from within South Dakota, with about 19% from border states and 72% from across the rest of the country, with heavy attendance from Arizona, California, Colorado, Iowa, Minnesota, Nebraska, Washington and Wyoming.

I find this aspect far more upsetting than the stupidity of the event itself. It's one thing when a bunch of locals decide to congregate in an environment they really shouldn't be, but another thing entirely when it's 400,000 people biking across the country.

Ironically - or perhaps not - I saw an NPR article ~a week ago[1] indicating the infection rate in Sturgis and its county, while certainly higher than it was before the rally, wasn't as catastrophic as you'd expect. Apparently, many people who live in Sturgis left town or just tried to avoid contact with anyone that weekend. Can't find any up to date information on whether the infection rate has increased since then with more time/testing, though.

[1] https://www.npr.org/sections/coronavirus-live-updates/2020/0...

Arizona saw this early in the pandemic in cities along the California border like Lake Havasu.

They had zero mask restrictions for a while, and cases were low.

Then they opened restaurants and bars, which were flooded by thousands of Californians coming across the border every weekend and not taking any precautions and Arizona's ICUs quickly reached capacity.


"While many locals are grateful for the tourism dollars amid a global downturn, some are worried that the crowds could cause a rise in coronavirus cases locally and overwhelm the city’s only hospital, which has just 16 ICU beds for its 55,000 residents."

The article and the quote from the article you linked do not support your statement: "Arizona's ICUs quickly reached capacity".

The article is dated May 11.


2822 excess mortality for Denmark in its flu season of 2017/2018. Extrapolating Denmarks population to Swedens population we would expect 5000 excess mortality for a comparable bad flu season. Current Sweden COVID related deaths are 5800. Less than 16 deaths a day in Sweden since July 2. Less than 8 deaths a day since July 19.

Swedens COVID related deaths will not go above 6000. Seems to be a 0.06% mortality rate, heavily weighted towards the aged. 6000 deaths is still a lot, yes.

So, article was May 11 when the influx started.

This is June 25th.

"Fewer than 200 ICU beds available in Arizona as state reports more than 60,000 coronavirus cases."


This is July 3rd.

"State health officials say the capacity of hospital intensive care units is at an all-time high of 91%.

The number of people hospitalized Thursday due to a suspected or confirmed case of COVID-19 was 3,013, according to the Arizona Department of Health Services. It’s the first time reaching 3,000."


That's a pretty scary timeline. Things got real serious real quick in less than a month.

Interesting, I agree the 91% of ICU capacity does seem alarming, but what aspect of that is hospitals optimizing for maximal usage of beds for financial reasons? I would want to see how much an ICU was used in previous years and how often it reached 90% capacity.

Also given it was 91% on July 3rd, I'd be interested in knowing what the situation was on August 3rd and September 3rd.

That's a great question.

I found this plot from Arizona Department of Public Health.'Hospital Bed Usage & Availability' sub-tab 'Number of Intensive Care Unit Available and In-use at Arizona Hospitals.'


Everything is reported in percentages, so I don't know how the absolute number of beds shifted as a result of temporary increases in ICU capacity.

Great graphs on that site, 57% of ICU in beds in use by COVID patients on 7/13 to 13% on 9/7. Seems like the peak has passed, but like you said it could be due to an increase in beds provided.

That's an article about a single tiny town. Not the entire state.

Sure, it's meant as an example of how large influxes of people into an area can potentially lead to superspreading-type events, and how common these are. That tiny town is an example of thousands of such places and events throughout the country.

Check out this graph from Arizona department of health 'Hospital COVID-like & Influenza-like illness Surveillance.'


Maybe it's spurious correlation, but I'd bet money the May 25th memorial day weekend had something to do with that big spike you see ramping up May 31. Memorial day also happened to feature thousands of people from another state flooding into Lake Havasu and other, similar cities without mask restrictions, and visiting bars, restaurants, etc.

Bear in mind, there were no mask restrictions in AZ before that peak.

Nobody has an RCT, so we're all just whiteboarding it to some extent.

> Maybe it's spurious correlation, but I'd bet money the May 25th memorial day weekend had something to do with that big spike you see ramping up May 31. Memorial day also happened to feature thousands of people from another state flooding into Lake Havasu and other, similar cities without mask restrictions, and visiting bars, restaurants, etc.

Maybe this is a spurious correlation too, but I'd bet money that the May 26th George Floyd protests had something to do with that big spike you see ramping up May 31 https://en.wikipedia.org/wiki/George_Floyd_protests.

I believe this is a replay of spring break in Florida...

"400,000 people biking across the country" ... and periodically stopping at gas stations, restaurants, hotels...

Truly, this is upsetting. Especially when you consider how violently they shun facemasks.

I travel by motorcycle a lot, so I'm familiar with the routine. My level of interaction at home (Berlin) and on the road are not that different. I'm not sure if it would be much worse than staying home.

As you say, gas stations (pay at the pump, or go in with a mask), hotels (mask on, often no interaction) and restaurants (yep, that one is bad). The rest would be spent on the bike.

On the other hand, unlike at home, I don't meet anyone else, I don't go grocery shopping, and I don't use public transport.

Of course I could also indulge in a bit of whataboutism about what the locals are up to, but I don't think it's a compelling argument.


If you'd like to have a better idea of what the rally itself was like and its attendees, All Gas No Brakes did a bit on it.

I doubt as many people were infected as they state as only one death has been reported so far. If as many were as infected as they claim many more deaths would have occurred by now.


Hold up, did you look at a graph of cases? It has an incubation period. Absolutely every single one of those things is associated with a spike.

Good point. I see a spike in cases in mid-to-late June. Now ... what happened in early June that might lead to a spike in cases. Involving large crowds of very random people congregating together, perhaps shouting. Hmm. Well ...

It's my understanding that the spikes are associated with earlier events than the two-ish week single infection incubation period because for the case spikes, you have to have a few rounds of infection aka exponential growth before the infection rate impact makes a blip.

As in: If you got 10k new cases in mid/late June, it's not "because" 10k people got infected at one event in early June, but because 10k ppl got infected in early June from 5k people who got infected in late July from ~3k people who got infected in early July from ~1k people who get infected in late May from one event.

Note: Specific numbers pulled out of my ass as an easy way to convey the idea.

Protests were found to not be correlated with covid. Spring break however was.

Things that increase the risk of infection:

- spreader lacks mask

- infectee lacks mask

- indoors

- spreader speaks

- length of time spreader and infectee spend in contact

The speaking part is a wash -- plenty of speaking at both protests and spring break. The rest of them all are chiefly problems with spring break, chiefly because people frequently socialize indoors as they did at Sturgis.

https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm, about half-way down.

Excess all-cause mortality in 2020 vs. previous years. (Note: data from recent weeks is incomplete.)

"Experts": 400K people for a few days is a "superspreader event"

Also "experts": there is "no evidence" hundreds of thousands of people yelling slogans in the street for 100+ days are spreading COVID.

The two can't be true at the same time.

I am not sure, in many photos I see, the protests are outside and have high mask compliance.

Sure they can - read this comment: https://news.ycombinator.com/item?id=24411270

No evidence was presented in that comment, other than evidence of political bias.

In the second paragraph, yes. The first paragraph pretty clearly explained the difference:

"The inside/outside dynamic is not the only part of the location that matters for the dangers these events present. The big protests mainly occurred in big cities and were mainly attended by locals. The Sturgis rally happens in the middle of nowhere, people come from all over, and then head back home to places that are generally more rural. Sturgis is therefore much more likely to spread the virus to people and places who wouldn't have had much exposure otherwise."

> mainly attended by locals

Significant percentage of the arrestees are _not_ locals. Look it up. Booking records are public.

...which is pretty much completely irrelevant when it comes to the percentage of locals at a protest, because the number of people arrested at a protest is only a tiny percentage of the number of attendees and is not a random sample.

It's not "irrelevant" when it comes to spreading COVID though.

In Texas, thousands of people attend drag races without masks.


Attendees do not seem to care, the venue does not enforce their policies, authorities do not take action...

I would not be surprised if there is a massive outbreak COVID-19 in that town.

sacks2k 20 days ago [flagged]

really? And many thousands of protesters all across the country (many not wearing masks and certainly not social distancing) aren't a 'superspreading event'?

I don't think I've even seen one article mentioning it...because racism.

Almost every bit of protesting footage I've seen has the majority of people wearing masks, which is a huge deal. In the footage of the motorcycle rally in sturgis, I have seen almost no one wearing masks. On top of this, people protesting were almost exclusively outdoors whereas there are tons of stores, bars, etc in sturgis that people went in.

Your comment is just ridiculous political pandering, trying to focus the topic other events that aren't similar in the slightest.

>On top of this, people protesting were almost exclusively outdoors whereas there are tons of stores, bars, etc in sturgis that people went in.

The inside/outside dynamic is not the only part of the location that matters for the dangers these events present. The big protests mainly occurred in big cities and were mainly attended by locals. The Sturgis rally happens in the middle of nowhere, people come from all over, and then head back home to places that are generally more rural. Sturgis is therefore much more likely to spread the virus to people and places who wouldn't have had much exposure otherwise.

Also we can't have this conversation while ignoring the false equivalency of these two events. There is a fundamental difference in the urgency of attending a protest versus attending the Sturgis rally. If they both presented the exact same danger to the population, Sturgis would still be a more egregious gathering due to being an entertainment event. We are in the middle of a pandemic. We can cut down on these large entertainment gatherings whether it is Sturgis or Coachella.

> There is a fundamental difference in the urgency of attending a protest versus attending the Sturgis rally.

The constitution deliberately makes no such distinction.

How is that relevant? No one is talking about the Constitution. Neither I nor the parent comments I responded to ever mentioned that Strugis rally should have been shutdown by the government. We can criticize something or point out that it shouldn't have happened without us needing to have the Supreme Court weigh in.

From a moral perspective, getting people killed in order to have some fun is less defensible than getting people killed in support of civil rights. I wouldn't think that is a controversial statement.

> How is that relevant?

The relevance is obvious.

How are the first amendment protections on the right to protest not an explicit distinction?

"Congress shall make no law respecting an establishment of religion, or prohibiting the free exercise thereof; or abridging the freedom of speech, or of the press; or the right of the people peaceably to assemble, and to petition the Government for a redress of grievances."

These are enumerations, not distinctions. One person's right to protest is not of any more importance than another's right to associate freely at a motorcycle rally.

> There is a fundamental difference in the urgency of attending a protest versus attending the Sturgis rally.

That's irrelevant. The virus doesn't care about the urgency. If both presented the exact same danger to the population, they should be handled exactly the same. Otherwise we are not arguing science but politics.

The virus doesn't care about urgency, but people do. Some things in life can't be put on hold for the virus while some things can. Some dangerous activities are still worth doing because they are important. Some relatively safe activities are not worth doing because they are so unimportant.

The Sturgis rally is on the long list of things we can skip for a year because its importance does not justify the risks involved.

Are you arguing that the protests are more valuable because they are a political event, or because they are a political event that you agree with? I'm interested in whether you would make the same argument in favour of, say, a Trump campaign rally.

Protests are a political call to action. A political rally is more a celebration of a specific political personality. Those two also do not have the same level of urgency. I am currently against holding any large political rally regardless of the candidate. I am open to protests from the right if they are willing to conduct them in a manner similar to how the recent protests from the left have been conducted.

I saw lots of protest images in June, and July, and lately, featuring very few people with masks.

Then you saw a set of images curated to support a specific narrative not representative of the larger group.

I live in Seattle and visited CHAZ/CHOP. While most protestors had masks, a large number of them had the mask below their nose and frequently not covering their mouth either. Nobody curated what I witnessed firsthand.

I think the biggest difference is the protests skewed young and motorcycle rallies skew boomer.

I have zero data to support this, but I've thought about it a lot and it seems to me that the majority of respiratory droplet secretion would be through the mouth during vocalization as a result of the vocal cords vibrating.

+1 for the boomer skew.

Masks are NOT a fool proof solution. Why do people think mask wearing removes all risk? Masks are just one component. Packed crowds in 90 degree heat with yelling and plenty of physical contact are just as worthy of scrutiny.

Now for my generally useless tidbit, In my city it was very hot and humid during the early summer protests. Mask compliance was probably about 50/50, with most of the folks not complying were wearing them incorrectly or taking them on and off repeatedly. Demonstrators used the time as an opportunity to socialize and share meals, with lots of incidental physical contact.

Who said masks removes all risk?

The data and official guidance I have seen supports the idea that masks (and being outdoors) significantly reduces risk.

I interpreted the comment I replied to suggest that masks remove a lot of the risks by itself. Is that not a fair reading of that comment?

The data I've seen suggests that being outdoor reduces risks, I have not need any data or studies that compare being outdoor with or without a mask. I'd love to read your studies if you have them.

It's surprising, but you can't argue with data - there were cases associated with Sturgis but not with the BLM protests. It kind of makes sense - that one thing Sturgis had and BLM didn't is bars. People catch COVID when they spend time in an enclosed enviroment next to an infected individual. Why bars and nightclubs are still open is a mystery from a public health standpoint.

It's true that you can't argue with data, but what if you're not allowed to collect the data. How can you criticize what you can't measure?

As an example, NYC contact tracers were forbidden[1] to ask if people attended a protest.


NYC contact tracers were forbidden to ask if people attended a protest

Refusing to collect that data point is completely understandable from a public health standpoint. There is substantiated concern that people may end up on cop shitlists for attending a protest, we know that there is police overreach, and the only way to keep the pigs out of the trough is not providing it at all. NYPD is notoriously corrupt. If any single person ended up being harassed by the cops as a result for telling the truth no one will ever work with a contact tracer again.

Sad that the cops can't restrain themselves in a time of public emergency but that's the reality we live in.

Very relevant point. We can't trust the science if the science isn't allowed to see certain things.

Not relevant at all. The government cannot be trusted to gather lists of protest attendees since they would be immediately used for retaliation.

Statistics can be looked at in aggregate. Infections did not go up in areas with large protests.

The BLM tribe has safer Covid behaviour in general than the other side, though, outside their protest attendance. That's a pretty big confounding variable.

that's why the authors of these studies have been using anonymized cellphone data to study concentrations of people and subsequent changes (if any) in case loads.

And the difference in mask wearing.

Yep. I went to only one BLM protest, but everyone was very conscientious about mask wearing.

Actually it was pretty easy to find a paper on that exact subject by the same authors.


> "we find no evidence that urban protests reignited COVID-19 case growth during the more than three weeks following protest onset. We conclude that predictions of broad negative public health consequences of Black Lives Matter protests were far too narrowly conceived."

Should concerns over Sturgis also be too narrowly conceived, then?

This is the subject of the article.

It is a separate event with much more in door crowds and much less mask wearing. Reality is not a negotiation where two different agendas agree on what they will pretend is the truth.

I think there's only a few white papers on the protests, here's one not otherwise mentioned in the thread:


Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact