Coronavirus: Prevent Seeding and Spreading

Should We Restrict Travel? Close Music Festivals? Schools? What Businesses Should We Open? What Social Events Should Be Allowed?

Tomas Pueyo
31 min readMay 13, 2020

This is Part 5 of Learning How to Dance (Part 1, Part 2, Part 3. Part 4 unpublished), a series where we go in depth to understand what countries need to do to Dance, to reopen their economies without new outbreaks. Translations available at the end of the article. New translations welcome! To receive the next articles, sign up here.

Countries are reopening. They want to imitate the success of countries like South Korea or Taiwan. We now know they can do that with great testing and contact tracing, isolations and quarantines, by mandating home-made masks, with hygiene, physical distancing and public education.

These measures are all extremely cheap compared to closing the economy, but are they enough? Do we also need to keep schools shut? Businesses? Music festivals? Tourism? Are there other measures that are expensive for the economy but necessary to keep the virus under control?

The bad news first:

  • We will likely need to heavily slow down national and international tourism for months
  • Big events like business fairs or music concerts will need to remain closed for now

The good news:

  • We should still be able to travel for one-way or very long trips
  • There are ways we can accelerate the reopening of tourism
  • We can probably reopen schools
  • A clear order is emerging for which businesses should reopen. The most important to keep open are likely banks, grocery, and general stores, and the least important are likely cafés, dessert parlors, and gyms

Ok, let’s do this.

Imagine that your country has applied the Hammer reasonably well and now it’s getting ready to Dance. Some areas are still in quarantine, but the rest of the economy is slowly opening. Authorities are careful: mask-wearing is strongly enforced, all contacts are traced, all sick people are isolated in hotel rooms, and quarantines are enforced by tracking mobile phone movements.

Your country now has two goals:

  1. Protect safe areas from infections coming from outbreak areas. I’ll call that “prevent seeding” of new cases.
  2. Prevent these seeds from developing into outbreaks. We can call that “prevent spreading”.

Countries must prevent seeding and spreading.

To prevent seeding, you need travel restrictions.
To prevent spreading, you need social gathering restrictions.

Since most countries are opening their countries and starting to allow social gatherings, it’s more urgent to talk about spreading. We’ll cover that first, and then we’ll explore seeding through travel.

The Coronavirus’ Favorite Events


As I’ve already covered, you’ve probably heard of the superspreader from South Korea:

Source: Reuters

A single person, Patient 31, caused over 5,000 cases. That was the vast majority of cases in South Korea at the time. As of the middle of May, this figure still represents over 40% of the cases in the country — not counting all the spreading that might have come from it since. If this person hadn’t caused this outbreak, maybe South Korea would never have had one, like Taiwan.

You might have heard that this was caused by Patient 31 going to a megachurch twice. What you might not know is what such a service might look like.

Service at a Shincheonji church. Source.

At this church, thousands of people sit close to each other for long periods of time, forbidden to wear eyeglasses or masks, singing and praying loudly. The church doesn’t allow them to miss the service, even if they’re sick. That’s why patient 31 went two Sundays in a row to a service while sick.

In Part 2 of this article series, I explained how the virus spreads, via droplets and droplet clouds, expelled by coughing but also singing, or even talking. We also discussed how the virus prefers confined spaces shared for long periods by people who interact closely with each other.

This megachurch environment met all the requirements that increase the rate of transmission — talking, singing, and being in confined spaces for long periods of time — especially since nobody was allowed to wear masks or glasses.

As we have also seen in Part 1, something similar happened in Singapore’s dormitories.

As a reminder, this is a picture from the main dormitory, taken after the lockdown in that country.

Same thing: These are places with lots of people in close quarters having heavy social interactions that likely include lots of talking.

Migrant workers in their room at the WestLite Toh Guan dormitory in Singapore. Photo: EPA. Source: South China Morning Post. As you can see, the spread might not have been their fault, but rather an unavoidable consequence of their living conditions.

The brutal spread of coronavirus at social gatherings includes many other such examples. One is a choir in Washington State, where people sing loudly, close together, for a long period of time. 45 out of 60 people fell sick.

Over 4,000 people were infected across Europe through what might be a bartender in Ischgl, where apres-ski events are common.

Trofana Alm, an après-ski at Ischgl

The German area around Gangelt suffered 1,500 infections after a carnival, where people commingled, talked, sang, and kissed each other on the cheek.

This is an image of the event:

Source. Photo: Heinz Eschweiler

The other biggest outbreak in Germany was also due to a festivity. The German district of Tirschenreuth, with 600 cases, has the most cases per capita in the country after festivities to celebrate beer.

The list keeps going:

  • At Bondi beach in Sydney, at least 30 people became infected after attending a crowded dance party the night before gatherings of 500 people or more were banned.
  • In France, a four-day celebration at the Porte Ouverte Christian church gathered 2,500 worshippers, infected many, and seeded outbreaks across the country, including 250 people working at the Strasbourg University hospital or 263 cases in Corsica.
  • A soccer game in February between Italy’s Atalanta team, from Bergamo, and Valencia, from Spain, seeded what later became Bergamo’s outbreak, one of the biggest in Italy.
  • Even in Seoul, South Korea’s capital, which has some of the best coronavirus management in the world, 2,100 bars and clubs closed in the middle of May after a single infected person visited five clubs over a long weekend, infecting at least 100 so far. Over 1,900 contacts are currently being traced.

This is not limited to massive celebrations.

Weddings, funerals and birthdays generated 10% of early spreading events in the US. For example, this man from Chicago infected 16 other people at a funeral and 3 of them subsequently died. This makes sense, since people at these types of events often spend hours together, touching one another, hugging, talking. In this 50-person party, over half got infected, and these individuals then spread the virus across the US and the world.

These super-spreader events also happen in the business world.

At least 115 meat processing plants have had outbreaks in the US, affecting at least 5k workers.

Packing plant workers ( USDA ). Source.

A Biogen conference in Boston in early March caused around 100 infections and spread the virus to six states. At a fish factory in Ghana, one worker infected over 500.

And it’s not limited to blue-collar work. An call center worker infected around 100 coworkers (~45%) in South Korea.


Say goodbye to the open office.

All around the world — we see the same pattern time after time. Close proximity, social interaction, singing, talking, dancing, hugging, kissing, all for long periods of time in a confined environment… A pattern consistent with the model that most infections happen through droplets or droplet clouds while coughing, talking or singing. We have not heard of outbreaks coming from movie theaters, for example, where people are mostly silent and stay in one seat the entire time.

A scientific paper (pre-print) led by Dr. Carl Juneau, PhD — I’m a co-author — reviewed hundreds of other papers. For the flu, the evidence was mixed, but it looks like the type and timing of events mattered. What we found was that the flu is more likely to spread in events that are long, crowded, and indoors. Another study suggested that, for flu, limiting mass gatherings was only helpful just before or at the beginning of an outbreak (but would have little effect 20 days after the infection peak). Here are some relevant quotes from more recent publications:

“The best-available evidence suggests multiple-day events with crowded communal accommodations are most associated with increased risk. Mass gatherings are not homogenous and risk should be assessed on a case-by-case basis.” Nunan and Brassey (March 2020), Oxford University, not peer-reviewed.

Results demonstrated that event cancellations can reduce COVID−19 infectiousness by 35%. Sugishita et al. (March 2020), not peer-reviewed.

Ferguson et. al. from the Imperial College believe the impact of limiting public events is a reduction in the transmission rate of around 12.5%.

With all this evidence, it’s clearer and clearer that the coronavirus loves social gatherings and crowded spaces. For those of us who love feeling part of a large group and meeting new people, we might be forced to say bye to them for the next few months. This is especially true if we are meeting lots of new people.

The Social Network

Another paper looked at how infectious diseases actually spread in social groups. The bigger the group, the faster the infection spread to the entire group.

Let’s look at the green line that says n=20, which represents a group with 20 people. This line shows that, as the transmission rate of a virus increases, so does the share of individuals in the group that become infected. And it does so faster than the normal population (the bottom dotted line). However, a group of 100 people will get most of them infected, even with a low infection rate.

Let’s compare a music concert with 10,000 people with a classroom of 20. In the concert, there’s a higher probability that at least one person is infected than in the classroom. On top of that, each person will interact with many more people, whereas in a classroom it’s capped at 20. And on top of that, the people in the music concert don’t know each other. They come from very different social groups, so the virus will then spread across all these groups, and these people will take the virus back to their groups, infecting them. Conversely, in the classroom, since it’s always the same group meeting, if somebody gets infected the infection largely remains contained in that group.

Big events are not just worse than small events. They are much, much worse.

Over the next few months, better research might shed light on exactly how much different social gathering events are impacting the spread of the coronavirus. In the meantime, what we can guess is:

  • Social gatherings should be avoided if a lot of people are close to each other, sing, talk a lot, or commingle.
  • Indoor, confined areas are much worse than outdoors activities.
  • Time matters. A short time is probably ok. Hours probably not.
  • Mixing different social groups is especially bad.
  • Conversely, small outdoors events where people don’t talk or interact are probably safe, especially if combined with measures such as masks, physical distancing and hygiene.

How can we use these insights to decide what types of events to allow?

Social Gathering Restrictions: Prevent Spreading

We can establish some rules of thumb. For example:

  • Cinemas or classical music concerts are probably much safer than rock concerts. In the first ones, a limited number of people are sitting close to each other but barely interact and don’t talk. They can also use masks if needed, and maybe seat 6 feet apart. In a massive rock concert, it’s hard to not sing, move about, or interact with lots of people around you.
  • Schools or small business meetings with the same people over and over are much safer than big fairs, travel for work, or big sports events. If somebody gets infected in the school network, many members might get infected too, but at least that doesn’t spread to many other places outside of that network.
  • If you have a wedding coming up, either cancel it, postpone it, or have a small ceremony in open air where people wear masks, don’t hug each other, and avoid talking as much as possible, especially from up close. Since doing all that would be very hard, it’s much safer to postpone it. In any case, if you do hold a wedding event, tell grandma or your diabetic cousin that you love them very much but that they really should not come physically. It’s a poor substitute, but under the circumstances a videoconference is their only safe option.

All of the above are early conclusions based on current — but scant — evidence. These conclusions will change. But with them, you can picture how risky an event is. What we’re trying to do is get to a place where we can estimate how many infections will happen at that event.

Note: All these numbers are guessed based on everything I’ve read. I’m sure they’re wrong. However, these are the decisions we’re making today: educated guesses. I hope this will spur new research quantifying this properly.

This table tries to assess how many infections could be caused by different events. We can see that watching a movie in the cinema might cause only one infection, while holding a big fair might cause 20,000.

This is only one side of the equation: the cost of allowing social gatherings. The other side is the benefit.

The Cost-Benefit of Social Gatherings

Once you have the cost per event — your infections — you can calculate the benefit per event, and finally compare everything apples to apples.

Once you have a version of this that reflects reality, you can reach many conclusions:

  1. Are events worth allowing, based on what we know about costs and benefits? For example, in this model, the opera has a value of $33k per infection. That is the highest value per person for any event (again, surely wrong. This is illustrative.) But what’s the cost per infection for society? If we establish based on our healthcare system that the cost is only $10,000 per infection, then we should allow the opera, theaters, cinemas and big conferences and congresses, but we shouldn’t allow events like big fairs or music concerts since their value is below $10,000 per infection.
  2. How can we account for value that goes beyond strict monetary value, such as psychological, ethical, or legal value?
  3. How does that change with changes in prevalence? What is the prevalence at which I can start opening events? How does that change per region? Can I open up events in areas where there are very few cases? At what point in the prevalence can I open up what types of events?
  4. How can I influence the risk of every type of event? Can I enforce mask-wearing? Can I force people to only interact with a few other people?
  5. How do these values per infection compare to those from tourists coming to your country?

Allowing these events might be more worthwhile if intelligent measures are taken.

For example, if scanning a QR code is mandatory to enter an event, and people need to have bluetooth enabled at all times to register who is close to whom, if we later discover one person was infected, we can trace all the other people at the event that were close by, contact them, test them, and isolate or quarantine them. If there are temperature checks at the entrance, this can reduce the number of potential infections entering (but probably less than half). If you can succeed at making people wear masks, your event might become acceptable.

Perfect is the Enemy of Good: Why an Imprecise Framework Is Needed

This might sound awfully hard and imprecise, so it’s worth pausing for a moment, take a step back, and realize why we’re doing this.

Most decision-makers are guesstimating all of this. Even if they don’t do the math formally, they are doing this calculation in their heads. The only thing we’re doing is extracting it from their heads and allowing people to debate it through a common language.

When a group of scientists and politicians gathers to decide what size of crowd they will allow, what conversation do you think is happening? It’s a cost-benefit, but they might not realize it.

“If we ban events above 1,000 people, we’re banning lots of sports, fairs, and music events. That’s hard.”

“Yeah, but these are the events that cause the most infections.”

“But what about all the jobs that we’re going to lose?”

“Think about all the deaths you’re going to prevent!”

The value of formalizing into a cost-benefit is not to be precise. Precision is impossible with the data we have. The value is to put all of our biases and pieces of data and knowledge and insight and goals and problems and solutions in one single place, in front of everybody, with one common language, so we can all discuss it and reach the best guess on what’s most valuable to allow for society.

Which Businesses Should We Open or Close?

Researchers are already applying this approach to determine which businesses should open first:

Source. Notes: danger is calculated taking into account the number of visits, the number of unique visitors, and person-hours of visits above two density thresholds. So banks are quite safe because few people visit, those who do tend to be the same all the time, and they don’t spend a lot of time. Conversely, lots of different people go to gyms and cafes, and they spend a long time there — in the case of cafés, talking a lot face to face. The importance comes from a combination of how much people care about different businesses based on surveys, and how much wealth they generate based on employment and receipts. Note that this model isn’t perfect because it doesn’t account for face-to-face conversations or singing. Further models should account for that, and ideally for other things such as empirical evidence.

Based on this analysis (which looked at real-world data such as mobility, consumer preferences, and government statistics) banks and grocery stores are the most important to maintain open, while gyms or cafés should remain closed and be the last to reopen. It is not perfect, but it gives a broad priority for the order of reopening:

  • If they were closed, banks, finance companies, grocery and general stores should be the first to reopen
  • After that, reopen department stores, colleges and universities, clothing and shoe stores, and auto-dealers and mechanics
  • Next would come furniture and home good stores, electronic stories, barbers and salons, hardware stores, places of worship, casinos, office supply stores, and movie theaters
  • Next come amusement parks, book stores, museums, pet and supply stores, and liquor and tobacco stores.
  • The last to open should be sporting goods stores, gyms, cafes, and dessert parlors

Interestingly, the analysis confirms the types of businesses that countries have already prioritized intuitively, even with imperfect data.

Models like this are not perfect yet. For example, I disagree with their conclusion on eateries. It’s true that fast food restaurants should open before sit-down restaurants, since people stay for much less time and talk less, but this paper suggests we could reopen some of them at full capacity early on. That’s because they don’t account for the huge danger that is added in eateries of people sitting face to face for long periods of time. They also don’t account for creative ways businesses could adapt to reopen safely. Gyms, for example, could limit attendance, reorganize floor space, require people to maintain a 6-feet distance, and/or have someone on staff keeping all spaces clean full time, like in this gym:

Businesses could also impose stronger restrictions during exclusive business hours for older clients. This is just an example, and in the end, the broader question might not be “Which businesses should open?” but rather “How can businesses reopen safely?”

Families and Transit

In China, outside of Hubei, around 80% of infections involved families and close contacts in the household. This is surely an effect of home quarantines, since mass gatherings were limited. But it illustrates how much the virus spreads in families.

Among these contacts, the highest risk by far is transmission to spouses: 28% of them end up infected, compared to 17% for other adults in the household and just 4% of kids below 18 years old.

Nevertheless, we can probably draw conclusions that are applicable elsewhere:

  • Stopping infections at home can have a dramatic effect on the epidemic
  • You are not very likely to catch the virus from random people on the street.
  • You are, however, very likely to catch it from your spouse, your kids, your parents or the friends you visit.

Transit also has an impact, although it’s unclear if it’s due to traveling with family members or caught in transit from other people.

So treat your family members with caution. Try to avoid interacting physically with members that are at risk. Call them instead. If somebody has symptoms, try to isolate that person. It’s hard to do that with a family member, but the alternative can be more sickness, a visit to the hospital, or death. So if there are symptoms, avoid kisses, hugs, or even talking face to face without a mask.

Thank you Dr. Muge Cevik for your round-ups of papers.

What about Schools?

The role of children in the coronavirus epidemic is unclear. How likely are they to get and pass on the coronavirus? How does it affect them?

On one hand, it looks like they are infected less frequently, and when they are they don’t die of it or infect other people much. As a result, it looks like school closures would only have a small effect on limiting the spread of the virus. On the other side, we’re not sure, because most countries closed them, so it’s hard to compare, but there are some emerging news of children with special symptoms, and some outbreaks are reported at daycares.

One country that only closed schools in a few regions and only for a few weeks is Australia, which has now controlled the epidemic.

According to their prime minister, the expert advice they received was that keeping schools open wouldn’t greatly impact the spread of the virus, it might even be better since kids wouldn’t be spreading it across many different social groups (mostly their classroom / schools), and it might be beneficial for parents to be able to work.

What is the impact on education? We know that students who lost on education during World War II had severe reductions in their lifelong earnings. Nowadays, every year of education adds 10% in annual earnings. School closures also have an inordinate effect on inequality, since kids from richer parents have more support, while those coming from poorer families, fed in school, might not get to eat at home.

Given all that, it’s likely a good idea to reopen schools cautiously. The Economist suggests a reasonable order:

  1. Start with young children, opening childcare, pre-k, crèches, and primary schools. These kids are learning the most, they’re the least impacted by the coronavirus, their parents get the most relief from being able to work, and their age is prime for reducing inequality through learning and nutrition (note: conflict of interest. I have three toddler kids.)
  2. Continue with older kids, especially those with important exams: They are most able to follow physical distancing guidelines, and their future is more at stake.
  3. Measures must be taken to reduce their infectious spread, such as temperature checks, tests for educators (and ideally kids), reductions in class sizes, and ideally attendance could be split over shifts or days.

Colleges and universities are different from schools. They gather many people from different social networks. Students commingle in high-density environments where they touch and talk with each other. This is true for nearly all aspects of their lives, from classrooms to study groups, dorms, or parties. Students are also older than kids, so more susceptible to the virus, and they’re more able to study through online learning.

As a result, the risk is likely higher than for schools, and the benefit might be lower. Despite that, universities are in a bind. If they don’t open by fall, they lose half their yearly revenue. If they open online only, they are making the case that on-campus learning is not as useful as they used to say, which will be hard to walk back once they can freely accept students on campus.

This will likely be an advantage for online schools, while some campus-based ones with existing financial difficulties might have a hard time remaining open.

This is why they’re so eager to open in the fall. They will do whatever it takes to open the campus while reducing spread.

They might be able to do it, since students are more able to follow social distancing rules, while universities are in a position to change their environments to limit the spread of the virus and to enforce the use of smartphone apps by everybody on campus, making contact tracing, isolation and quarantines much easier (note: conflict of interest. I work at Course Hero, a company that helps college and university students study online.)

Summary of Social Gathering Restrictions

This is what we’ve learned:

  • The coronavirus spreads mostly in indoor, confined environments where people spend a lot of time together talking, singing, breathing, or touching each other.
  • Big events that mix lots of social groups are riskier. They should be limited. This is especially true of things like business fairs and music festivals, whereas cinemas and operas might be ok, if physical distancing is respected.
  • We should stagger the reopening of businesses. The highest priority is always groceries, banks, financial services, and general stores, while the lowest is gyms, cafes, dessert parlors, bars and clubs.
  • High density workplaces are a recipe for contagious disaster. Physical distancing measures must be taken, such as physical barriers, temperature checks, staggered shifts, remote work, or open air if available.
  • Be very careful with the family. You’re most likely to catch the virus from your partner, other adults in your household, or friends. The biggest risk is at home or during transit. Avoid seeing grand-parents and isolate yourself as much as you can if you have symptoms.
  • Reopening schools should be a priority, always being careful, and starting with younger kids.

Countries must prevent seeding and spreading.

We’ve reviewed how to stop the coronavirus from spreading in a community by limiting social gatherings. Now we also need to make sure we’re not seeding the community with cases from abroad. Let’s look at travel restrictions.

This section has been heavily influenced by the work done by Barthold Albrecht. Thank you!

Travel Restrictions: Prevent Seeding

The Impact of Travel Restrictions in East Asia

As we saw in Part 1 of this series, Singapore was doing many things right, dancing nearly perfectly, but then workers from abroad seeded the country with coronavirus and kindled an outbreak. The lax travel restrictions seeded the disease, the lack of mask requirements spread the virus, and the lack of limits on social gathering size created super-spreader events.

The country controlled seeds from China really well early on. As you can see, a few blue seeds in January were stopped after the ban on Chinese visitors. But it was then too slow stopping visitors from countries like Italy, France, Spain and Germany in the first half of March, and then the UK, the US, and other countries later. The result is that these few blue cases then created the outbreak (in pink) just after that.

Conversely, as we saw in Part 1, Taiwan was extremely on top of travel bans, updating them daily. The country banned travelers from Wuhan the day of the lockdowns, and all Chinese nationals two weeks later, on February 6th. On March 14th all European resident visitors had to quarantine for two weeks upon arrival. It was expanded to all visitors three days later, and by March 19th all non-resident foreigners were simply banned. As of April 18th, non-essential travel was forbidden and all incoming travelers must self-quarantine for 14 days.

South Korea had a reverse ban. “Thanks” to their early epidemic, 171 countries banned travel to and from the country, which eliminated a lot of potential seeding.

Another country that has officially managed the crisis really well is Vietnam. It declared the situation an epidemic as early as February 1st, as soon as they discovered community spread. They stopped all flights from China. They then quickly banned flights from most countries. In fact, they were so fast that they closed the country to visitors from Spain, France, and the UK before those countries announced their lockdowns. As of April 18th, the country has 270 cases, in a population of 95 million.

This graph shows when different countries in East Asia took different travel restriction measures. Each country has a lane, and the colored lines show restrictions targeting different geographic areas: Hubei, China, South Korea, Northern Italy, Europe and the world. It’s difficult to really graph this, because there is an infinite number of measures that can be taken, from temperature checks to health declarations, regional travel restrictions, quarantines for certain visitors, exceptions for others… But we tried to simplify as much as possible taking what was closest to either a mandatory quarantine for all visitors from a certain area, or a full-on ban — despite frequent exemptions, such as diplomats or others.

In this graph, we can see when different countries from East Asia took different travel measures. Each line represents a restriction from a different area, from Hubei to the world. We can see that Taiwan and Vietnam, which both have fewer than 1,000 cases, tended to restrict travel from most other countries faster.

Singapore was reasonably quick, but not quick enough to stop the seeds, mostly from Bangladesh and Indonesia. Luck also has a role: If Singapore hadn’t been so cosmopolitan, it might not have this outbreak. Thankfully, the amazing management of the country has already reduced the outbreak.

South Korea is a special case: Although it wasn’t always very fast restricting travel, it did restrict travel from China quickly. By the end of February, dozens of countries had already banned travel from South Korea, which also made it much harder for travelers to go to the country. This “reverse ban” surely helped the country.

Although Thailand was slower, it had some measures, such as temperature checks and health declarations. It might also have been lucky to prevent more seeding.

Japan was in general the slowest of them all, which likely explains the slow growth in cases through February and March, with the final outbreak at the end of March. Thankfully, the country has also reacted well and appears to be controlling the outbreak now.

With all of this, we can see that the East Asian countries that restricted travel fastest were also in general the ones to have less seeding — and hence less spreading.

The problem, obviously, is that travel bans are very expensive. For some countries, heavily dependent on tourism, it’s vital.

Travel Bans vs. Quarantines

Tourism accounts for ~10% of global GDP. For a country like Spain, it’s 15%. The country receives 83 million visitors a year. The lockdown in March and April has likely cost the country 2% of GDP just in 2 months. It’s desperate to open up now: 6% of all GDP comes from tourism between June and September. Will Spain be able to open up in time? How should the country think about travel restrictions?

There are many different types of travel restrictions, such as temperature checks, health declarations or testing, but two have a major impact on seeding: quarantines and outright bans. Quarantines require visitors to be secluded for two weeks. Bans simply prevent them from entering the country.

If the quarantine is as exhaustive as in Taiwan, it is reasonable to assume that very few infections happen from travelers.

Unfortunately, this is unaffordable and impractical for most tourists. Who wants to spend two weeks of their holiday locked in a room, only to spend two more weeks in quarantine when they’re back home?

If tourism countries like Spain are entering the dance, they need to limit the seeding from abroad. For example, let’s take travelers from the UK. In 2019, 18 million of them went to Spain. Let’s assume that 1.2% of brits are infected, and that sick people are 50% less likely to travel. That means that, over a year, Spain would import 108,000 coronavirus cases from the UK, an average of around 300 per day. And that’s just one country. You can imagine what a disaster it would be to open up the country to tourists.

So how should a country like Spain think about this?

The key is to think in terms of cost-benefit. Your cost here is new infection seeds in your country. That’s what you want to minimize. So you try to understand the number of infections that travelers from a foreign country would import. The benefit is the money they bring to the economy.

There are probably two types of travelers: those willing to go through a quarantine, and those who aren’t. The ones who are willing to go through a quarantine are more likely to live in the country and to participate heavily in the economy. Thanks to the quarantine, they are also much less likely to add cases to your country. So they cost less in terms of infections and are worth more in terms of economic activity.

Therefore, as a rule of thumb, countries that can afford the foreign seeds should first switch from travel bans to travel quarantines for all visitors, no matter the country they’re coming from. Spain has finally realized this.

Countries that do a good job at quarantining long-term visitors make sure they have a residence. Otherwise, they must stay at a government-sanctioned hotel. Some countries even pay for it.

They also ensure that the transit from the airport — or border — to the residence has as few opportunities for local contagion as possible. In Taiwan or China, you either use your private transport or government-sanctioned transport. Any other transport — including walking on the street — is forbidden.

Once this operation works well, countries should monitor the situation to ensure few infections are imported, and those are always contained. They might convert travel bans into quarantines country by country, depending on the cases they’re importing and their ability to contain them.

Let’s take this imaginary situation:

Imagine that a country like Spain had to choose with which countries it opens its borders. Country H is usually a great source of tourists (10 million!), but it has not controlled the epidemic and right now around 5% of their population might be infected. Assuming these people tend to feel sicker and travel 30% less than others, that might mean that nearly 350,000 infections could enter the country. No way.

Conversely, Country D is dancing successfully, and has very few infected travelers. It makes more sense to open up to that country first. But which country should we open then: D, O, or W? You need to account for the value brought to the country per infected traveler.

Here, Spain is applying a quarantine to all long-term travelers, which we’ll assume reduces the number of seeds by 75%. We can see that opening up the borders to those from country H brings around $1.7 million per infected person to the economy, but opening Spain to visitors from country D brings $700 million! It also becomes obvious that, if Spain has to choose where it gets long-term visitors from, it should prioritize first country D, then O (around $11 million per infected person), and then W and finally H.

Spain would decide from how many of these countries it can receive seeds based on the number of cases it can tolerate. Here, opening to countries D, O and W would bring around 1,250 cases per year to the country, or around four per day.

If that works, the next question for any country becomes: Can we even lift quarantines for travelers from certain countries? This is the holy grail of tourism. Without that, it’s in trouble.

The calculation is really the same: What’s the value per tourist you’re bringing, and what’s the cost per infection? Open up to countries with the best value per infection, up to the maximum number you decide your healthcare system can process, and as long as the benefits outweigh the costs.

This can illustrate, for example, that it makes sense to allow for tourists from Country D and O, even without a quarantine, before allowing any long-term traveler from Country H, simply because Country D and O are doing a good job at dancing: Their tourists bring $25 million and $2 million to the local economy, compared to only $1.7 for long-term visitors from country H.

With that, we can prioritize what countries to get visitors from, and what kinds of visitors to get.

If Spain can afford to bring 10 new cases per day to the country, it can get long-term visitors from countries D, O and W, and tourists from country D.

There are a few important details. For example, don’t just consider the country of origin, but also the travel path. You need to feel confident about the share of infections you can prevent through quarantines. You need to guess the prevalence per country and know the value per tourist or resident.

More importantly, understand the numbers that matter most and how to impact them. For this model, the key numbers are the prevalence of infected people in origin countries and the share of infections that you can catch or isolate. If you enable temperature checks and PCR tests in airports to catch who is infected, for example, you can suddenly open up the country dramatically more. This is the result if you catch 75% of the potentially infected:

Suddenly, long-term visitors and tourists from most countries are welcome! This might be one of the reasons why South Korea has been able to keep their borders open for a very long time. This is their process:

Austria already offers PCR tests for incoming and outgoing travelers.

These important details are what the next few weeks — and months — of work should resolve. These numbers are all made up. But if the orders of magnitude are broadly valid, this type of model can give us many different different insights. For example:

  • The two key factors here are coronavirus prevalence (share of the population that’s infected) from the source country and the success at identifying and isolating cases. After that, it’s the value per visitor or tourist.
  • If there’s a lot of cases worldwide, even quarantining all travelers might still overwhelm the system, and we might want to pick from which countries we allow visitors. If most of the world does a good job at controlling the coronavirus, the prevalence will be so low that countries will be able to open up to long-term visitors from many other countries.
  • Countries that pursue herd immunity might have so many cases that even a quarantine of their long-term travelers is not good enough to allow them in foreign countries.
  • Long-term visitors, who are likely to bring more value per person to the country than tourists and might be willing to suffer a quarantine, might nearly always have a higher priority than tourists. If this is true, we are likely to first open up to countries through traveler quarantines, and once the system can cope with that, we can start allowing countries to send travelers without quarantines.
  • If a country is amazing at identifying traveler cases and quarantining them, and can boast a score of 100% quarantining success, it can open up to receiving long-term visitors from across the world.
  • Tourists might in general be allowed in after residents. But citizens from countries that have a very low prevalence might be free to go wherever they want. For the tourism industry, this might be hard: Unless prevalence is really low, countries get really good at catching coronavirus cases early on, or the tourism industry can figure out ways to distance all customers physically, it could be hit for the entire time until there’s a vaccine or effective treatment.

This is all based on this basic and theoretical model, so any of the conclusions above might be wrong. The point is not as much to reach these conclusions, but to highlight how such a way of thinking can help make policy decisions.

This is broadly how Hong Kong thinks about the problem

From their official plan:

“The Government has conducted detailed risk assessments prior to implementation of port health measures in view of disease outbreak in other countries or regions. Other than considering the number, distribution and rate of increase of infected persons, the Government would also take into account the surveillance and control measures implemented by the authorities of that particular country/region, as well as the frequency of travels between Hong Kong and that particular country/region. The Government would suitably review and rationalise the relevant measures having regard to the latest development of the outbreak.” Legislative Council Panel on Health Services Prevention and Control of Coronavirus Disease 2019 in Hong Kong

A Tale of Two Worlds

This suggests that countries might split in two classes: dancing countries and infected countries.

With hard work, dancing countries will have eliminated most internal coronavirus cases. They will welcome travel between them, with few limitations. However, they will shun countries that go for herd immunity, for fear of causing new outbreaks. Infected countries, in turn, might welcome visitors from all countries, since they have nothing to lose. Unfortunately, they might mostly receive visitors from other infected countries, since those from dancing countries won’t want to catch the virus and their countries may not let them come back without a quarantine.

This is another reason for countries following herd immunity to beware. They might have a hard time enabling their citizens to travel abroad, and struggle to keep their tourism industries afloat due to extensive travel restrictions to and from their origins

So far, I’ve only talked about countries, but the exact same logic is valid across regions.

Within the United States, some dancing states have very few cases, such as Alaska, Oregon, Montana, Idaho, Hawaii or Maine. Infected states with bad outbreaks include Illinois, Iowa, New York, New Jersey, or Maryland.

We could imagine dancing states opening up for travel between each other, but preventing visitors from infected states or forcing a quarantine, like Hawaii and Alaska already do. States like California, Nevada or Washington might work hard to get to a level of prevalence low enough to get accepted to the club.

Conversely, some states like Iowa, Nebraska or Illinois might have such intense outbreaks that they decide to go for herd immunity. These could be joined by Southern states like Georgia, Tennessee or Mississippi. States in the middle might decide whether they want to belong to the dancing or the infected club.

Something similar is being considered in many countries.

In Australia, local and regional travel is allowed as regions enter Step 1 (some already have). Step 2 will allow some interstate travel, and Step 3 will include international travel.

In Spain, where different regions will have different rights based on their levels of prevalence, travel to and from dancing provinces will be allowed, while infected provinces will have to work harder to stop their epidemics. The central government does not allow for provinces trying to reach herd immunity.

This section draws on the work from Jorge Peñalva.


Countries that want to dance successfully might be able to do so with just testing, contact tracing, isolations, quarantines, masks, hygiene, physical distancing, and public education. On top of that, they might need to limit social gatherings until coronavirus prevalence is low and they figure out ways to limit contagions in these events. Throughout the dance, they will have to limit infections coming from abroad through quarantines, filtering measures at the borders, or outright bans.

The countries who do all of that well will be able to open their economies as well as travel between each other.

For the countries that don’t and end up pursuing herd immunity, their suffering will not be limited to the loss of travel from dancing areas. It’s worth remembering the cost in deaths, chronic conditions, ongoing outbreaks, and economic impact.

If the Transmission Rate R is 2.7, around 65% of the population might get infected, or around 214 million Americans. If the Infection Fatality Rate (IFR) is between 0.5% and 1.5% — which seems likely based on the most reliable seroprevalence studies — the death toll across the US could be between 1 and 3 million people.

On top of that, there might be chronic conditions that emerge from the effect of the virus on lungs, kidneys, blood, brains… Reaching herd immunity would take months, during which the economy would be depressed since people will not want to go out, catch an infection, and spread it to their loved ones. Around 45% of Americans have conditions that would make the coronavirus very dangerous, such as diabetes, obesity, or simply being older.

All of that will likely be worthless anyways, because they won’t be able to prevent further outbreaks: herd immunity through sickness doesn’t mean the population is now free of it. Outbreaks frequently re-emerge, especially if the virus behaves like other flu viruses and our immunity disappears after 1–2 years.

So the herd immunity strategy would be bad for the health and the economy. Instead, avoid seeding and spreading.

If you want to receive the next installments of the article, sign up for the newsletter.

If you want to translate this article, do it on a Medium post and leave me a private note here with your link.

Italian (alternative)

This is Part 5 of our Article, Coronavirus: Learning How to Dance. In Part 1, we discuss best practices from Taiwan, Singapore, China and South Korea. In Part 2, we discuss masks, hygiene, physical distancing and public education. In Part 3, we cover testing and contact tracing. In Part 4, we will talk about Isolations and Quarantines. In Part 6, we will put all of it together. We will give specific recommendations on each, including a warning: Most countries are not approaching the Dance well. If they continue their current path, they will end up like Singapore.

This has been a massive team effort with the help of dozens of people who have provided research, sources, arguments, feedback on wording, challenged my arguments and assumptions, and disagreed with me. Special thanks to Barthold Albrecht, Jorge Peñalva, Genevieve Gee, Matt Bell, Carl Juneau, Mike Mitzel, Christina Mueller, Elena Baillie, Pierre Djian, Yasemin Denari, Claire Marshall, and many more. This would have been impossible without all of you.



Tomas Pueyo

2 MSc in Engineering. Stanford MBA. Ex-Consultant. Creator of applications with >20M users. Currently leading a billion-dollar business @ Course Hero