The Geographic Spread of COVID-19 Correlates with Structure of Social Networks as Measured by Facebook

There is a new NBER working paper (by economists) on Covid-19:

We use anonymized and aggregated data from Facebook to show that areas with stronger social ties to two early COVID-19 “hotspots” (Westchester County, NY, in the U.S. and Lodi province in Italy) generally have more confirmed COVID-19 cases as of March 30, 2020. These relationships hold after controlling for geographic distance to the hotspots as well as for the income and population density of the regions. These results suggest that data from online social networks may prove useful to epidemiologists and others hoping to forecast the spread of communicable diseases such as COVID-19.

That is by Theresa Kuchler, Dominic Russell, and Johannes Stroebel.


However, among the largest American clusters of Covid19 are Sioux City SD, a nuclear aircraft carrier, and the Cook County Jail. Somehow, one can reasonably doubt that Facebook would have been particularly helpful in at least two of those cases, and considering the workplace structure of a large US meatpacking plant, Facebook would likely not have been too helpful to Smithfield Foods either.

Anything based even on week old data is pretty irrelevant as the pandemic spreads. Think of it as a fire in a welding supplies store. It may be relevant to track the initial progress of the fire from a faulty electrical device, if only to learn how to prevent such conditions in the future, but using that information after the tanks of acetylene are exploding is basically a waste of time.

On the other hand, more people vacation and take business trips to Northern Italy and the New York area than to Sioux City SD, a nuclear aircraft carrier, and the Cook County Jail.

Which shows that a virus does not care about how many people vacation or take business trips once it begins spreading from its initial starting points. It does look like South Dakota will be providing another counterpoint to the lockdown strategies. It is Smithfield shutting down its plant, while South Dakotans remain to free “to exercise their right to work, to worship and to play. Or to even stay at home.” At least for all those not employed by Smithfield in Sioux City, since those 3700 employees no longer have a right to work there, though they remain free to exercise their right to stay at home.

We are easily a couple of months beyond the point where it matters whether someone went skiing in Switzerland and got infected.

Austria, my friend. 'Twas the filthy Öschis.

But it's a lot cheaper to shut down a meatpacking plant in South Dakota than it is to shut down New York City.

A lot of the underestimation of the severity of the problem was based on assuming that, sure, it would pop up here or there, in some meatpacking plant or jail, but we could deal with that.

But instead it took over New...York...City, the economic capital of the world.

That's really bad.

It will be cheaper to shut down Sioux City too. Especially now that fracking is starting to flatline in the Dakotas.

Oh wait, we don't talk about that when talking about the economy and unemployment claims in places like North Dakota, do we?

Enlighten me...what do New Yorkers eat when there's no food to eat? The meat packing shutdown list grows daily.

Food magically appears on their shelves. Duh.

Al grew up in NYC, one of 12 kids. Many years after Al passed, I asked his son if Al ever talked about the depression. "Al checked out a noise he heard in the middle of the night. Al went to the kitchen and saw his mother at the table sobbing. There was no more food or money."

No food...unlikely. I said unlikely a few months back on C-19, but Trump may yet prove me wrong.

"New...York...City, the economic capital of the world." Nah.

Perhaps the Financial Instruments capital of the world.

since those 3700 employees no longer have a right to work there

Get a grip. It's not going to take that long to disinfect the plant. The problem is the supply chain back ups on protective equipment, and the time necessary to redesign production processes in order to contain infection.

Having those workers return - not counting the less than 10% that are sick, as of a couple of days ago - might be just a bit more challenging.

You fancy they'd prefer to not earn a living? OK.

I am sure they would prefer to earn a living - without catching a disease. And that many of them will try to earn a living somewhere where the chance of being infected is much less than 1 and 10 (and one assumes rising every day - they are likely to be much better positioned to know who they work with immediately gets sick)..

Pandemics cause people to change their behavior - most definitely including doing their best not to get sick next.

Economists discover that geographic spread corresponds to those spreading a disease.

Boy, the epidemiologists are going to have egg on their faces again, aren't they?

No, you missed it. What it says is: "Friend let friends catch Covid-19 from them". This is why, during the Black Death, teenagers ran away from their families (Google this, anecdotal and thus perfectly sound data in the literature that families broke up over the Black Death, to escape disease).

Wow, this is revolutionary, epidemiologist will have to learn about network analytics.

The only ones that appear to be surprised are persons who don't know that epidemiological models use network analytics.

Based on the post, and the links, my guess is
That it is
Economists who do not know that epidemiologists
Use network analytics.

I have already succeeded in getting MRev to publish this observation.

And referencing the examples above, coworkers catch it from coworkers (meat packing plant example), crew members catch it from crew members (carrier example - twice, including the French one), and prisoners catch it from prisoners (Cook County Jail).

It looks like we are going to find that actual mobility in the US is a bit more limited than some of us might have thought. That there are a lot of rural areas where people dont travel much and outside people dont travel to those places. So we are likely going to end up with low rates of infections in most of the areas. However, if one of those areas does get hit, it looks like they can get hit hard either because of work conditions or because of the belief that they were safe since they weren't urban.


It only took a few weeks for there to be registered cases in all 50 states. North Dakota has more cases than some provinces in South Korea. To the extent that growth across geographic barriers starts to slow, that is probably because governments are restricting travel and private individuals are postponing trips.

There's a funeral home in Newfoundland, Canada with over 200 infections tied to one person who came to a funeral from out of the province. This thing is spectacularly unforgiving of even small mistakes.

Still, being remote and unconnected buys you time, which is valuable. New York was bound to get hit hard, because of its international connections, because of its density, and because of its terrible, terrible local and state government. Small towns in Vermont and the Dakotas will have to get unlucky. Eventually many of them will, but hopefully not before there are better treatments or before the excess capacity built up to handle the urban surge can be repurposed to handle the relatively slow rural percolation.

I am amazed at how people keep forgetting how one person can put a funeral home, a packing plant, or a distribution center out of business.

What all of the herders/truthers seem to be willfully ignoring is what happens when the economy opens and a school teacher, a warehouse worker, or a restaurant worker turns up positive and infects everyone around them.

It transmits through close personal contacts - family homes, nursing homes, hospitals, jails, ships, offices, churches, and parties. You will likely not get it from touching surfaces that others have touched.

Preventing big parties and crowds looks like it will slow the spread. But forcing people into their homes and bringing anyone but the deathly ill into hospitals could turn out to have made things worse.

As demonstrated at least by the meatpacking industry, where two plants - a Smithfield one in South Dakota and a Tysons one in Iowa - have been closed indefnitely.

The Smithfield plant is one of the largest single clusters in the U.S., and has gone from 80 infected workers to around 240 in less than a week. There are plenty of workplaces where people work in fairly close proximity over a shift. The food industry is full of them, but who considers food an essential industry?

What we need are data on the risks of infection by occupation, which we could get a lot of from data-mining vast medical databases like Kaiser-Permanente's. That would likely violate HIPAA, but data maestro Raj Chetty has shown that he can get around privacy laws on your 1040 tax returns, so we need some economists (such as Chetty himself) to get to work on figuring out how to get their hands on patient data.

If we could tell which jobs are risky and which are not, we could re-open the latter soon while making the former wait for better protective gear and better safety codes.

Epidemioligists use network analytics in their models all the time, so this is nothing new.
Using Facebook is new, but it was preceded by
Cambridge Analytica.

Rather than Facebook links, Nextdoor App would have been a better choice for neighborhood transmission analysis and maybe such things as church membership rosters or recurring events attendance lists.

If they are using Facebook anonymized data, I would be interested in persons who link to right wing media that claimed covid was a democratic hoax, or that social distancing was unnecessary, and examine whether those persons had a higher rate of infection than those who linked to articles supporting social distancing etc.

If they are using Facebook anonymized data, I would be interested in persons who link to right wing media that claimed covid was a democratic hoax, or that social distancing was unnecessary,

That's not of interest to anyone but the most puerile sort of partisan Democrat.


That would be a great research question.
And, you could go further:

Ask what kind of person would call this research
The kind
Pursued by the
Most puerile sort of partisan Democrat.

What does that say about you?

It says I'm an ordinary American and you're a puerile partisan Democrat who fancies he's an ordinary American with a curious mind.

One expects an Art-.and-Crafts sort of person to know all about glue and sticking.


I hope that someday you will come to understand the goodness of people and learn to respect the views of others and be willing to support and conduct research.

As someone else would say,
I will pray for you.

PS. You don't get to decide who is an "ordinary American".

Cambridge Aanlytica was preceded by the more egregious Facebook 2012 campaigns on behalf of the previous administration.

This isn't some groundbreaking thing....

Figure 2 is just sad, and in line with the other work I have seen from Stroebl and co-authors.

Comments for this post are closed