I spent an inordinate amount of time this year subtweeting articles, commenting on policy and discussing corona with folks on Twitter. Because I read studies and listened to weeks worth of podcast content on the topic in three languages (e.g. Das Coronavirus Update and This Week in Virology) I’d like to believe I have at least a somewhat informed take on the issue. I’ll try to summarise some of that here.
I may improve and expand this post in the future, based on objections I encounter, so if you want to write a scathing review, perhaps link to a snapshot as well.
Government vs. society
Since my audience includes many libertarian leaning folks, I’ll start with a brief treatise on the role of government. If you like longer essays, go read Nassim Taleb’s Principia Politica, which I think provides a good framework. On the one hand it leaves room for countries, states, cities, etc. to run things according to their own preferences. On the other hand it points to a few rare but extremely important phenomena which always require some heavy handed top down action. Pandemics are one such phenomenon.
There is an artificial distinction I often see people make between government and society as a whole. This leads to incoherent statements like: “Don’t take action, just let it spread”
The implication here is that “the government” shouldn’t take action, but this ignores the existence of the rest of society. A better question to ask is: what should society do? Once you ask that, it becomes clear there’s no such thing as “don’t take action”, because action is always taken.
Society as a whole can decide to distribute the burden more equally, or let everyone fend for themselves. It can decide to deal with the problem and try to keep the total damage limited, or decide to roll the dice.
Government is just a means to coordinate such a decision, as are markets.
It is society as whole that decides whether everyone stays home for a short time, or that only people who fear for their lives (or that of their loved ones) stay home for a very long time. There’s no meaningful definition of “voluntary” that covers that second option; it’s forced upon that group by society as a whole. This is where ethics come in, and terms like geronticide (but plenty of younger people suffer too).
In a healthy democracy this coordination is not a one way street. The people of Hong Kong forced their government to close the border, the Dutch government initially didn’t want to close schools, but parents were pulling kids out so fast they had no choice.
Whether a society acts based on government guidelines or a more heavy handed approach varies wildly. In Sweden people do the former, even if those guidelines are dead wrong, in America some people refuse to wear masks because they want to see liberal tears.
Even in non-democracies things aren’t as black and white. Totally ignoring the will of the people has always been a bad idea, ask 18th century French aristocrats. There was no mass revolt in China after their heavy handed measures, and it’s not like they don’t have a history of those. Propaganda alone cannot explain that outcome.
What if you had a group of people dedicated to monitoring the planet for pandemic threats, preparing plans for hospitals, border control and all these other things? What if everyone chipped in a tiny amount of money to pay their salaries? What if the amount was so small you didn’t even know you were paying for it, because someone else made sure that this system was in place? That’s pretty much my take on government: they take care of stuff you don’t want to spend you life obsessing over. Don’t worry, it leaves plenty of room for personal responsibly.
Without coordination most people act too late in a pandemic, perhaps only when a good friend or loved one gets really sick. By that time they’re already infected themselves. This may vary with experience: a single news article about a suspicious virus in Wuhan will get everyone in Hong Kong on their toes and spontaneously abandon bars and restaurants. But Dutch people went skiing and celebrated carnival by the millions. You’d think the first wave changed that, but mobility figures suggest they’re even less worried about it than the first one.
A well prepared government can act, and cause people to act, early. As time goes by, society can debate whether that was necessary and adjust.With a bit of luck the outbreak is contained without most people even noticing it. The corona virus is too contagious and widespread for that, but plenty of countries have not seen their hospitals overrun by it. More on timing later.
Personal responsibility vs pathological selfishness
A final point on personal responsibility: it’s necessary but not sufficient in a pandemic. It might work if mortality was 90% and death occurred too quickly to endanger friends or medical staff, because then your personal interests are well aligned with those of society as a whole. But that’s not the case here. For the vast majority of people the (perceived) cost of risking the disease for themselves is much lower than the cost they impose on society. A single person can cause millions of deaths and trillions in economic damage downstream, without paying a price for any of that. Most of the time they’ll spread it by accident, but even if it was intentional, there is no conceivable way for them to compensate the damaged parties. Without a way to retrieve damages, personal responsibility loses its meaning. Pandemics are an inherently collective problem.
Some people use the language of personal responsibility and localism to obscure the fact that they’re pathologically selfish. This is easy to discover: ask them if they’d accept a light measure, such as a mask mandate, if it is was proven to save billions of lives, and they’ll call you a fascist. Or notice how they suddenly complain about government measures right when it hits them, but show no interest in the pandemic at other times. Or how they protest masks using fancy tricks, such as pretending to have a disease or sincere religious objection, rather than actually risking a fine. Or they brag how they took their mask off in a bar, i.e. on private property against the wishes its owner, and intimidated staff. Almost all of them live in very comfortable democracies and merely fantasise about sea-steading and citadels. They oppose all government action, from global down to their own neighbourhood, not because of their beliefs, but because they truly do not care about others.
Bitcoin draws these people like moths to a candle, perhaps because financial independence is a prerequisite to remaining an asshole, but rest assured there’s plenty of decent people around that candle too.
Just the flu?
Cult followers can drag you and your followers into never ending arguments. With their zombie army numbers, they can drain 100% of your time.
I have blocked hundreds of people – or, more accurately, dozens of people and hundreds of troll farm accounts, for claiming covid is just the flu. Most parrot this claim without pointing to any evidence, which tells me they’re in a cult. I do not have the patience it takes to rescue individual people from cults. It can be done, see this interview with a former white supremacists or this book about leaving the Southern Baptist church. But at this point it’s not the best use of my energy.
It is however important to remain open to being wrong. For that I believe it’s sufficient to refrain from blocking a small number of people. Preferably those who can actually articulate their opinion and point to evidence. People who have a reputation for not always towing the party line. Always leave a few holes in your filter bubble.
Anyway, back to this flu issue. There’s a basic fact these people need to explain: why do hospitals get overrun despite the most draconian anti-pandemic measures since the middle ages? And/or: how were previous incidents of this happening hidden from the public?
They don’t do this. Instead, if they provide any argument for their claims at all, it tends to be one of these three, in order of sophistication:
- looking at total deaths per capita, ignoring the role of (voluntary) measures and luck
- pretending everyone has already been infected, so it’s over and we can just relax now (e.g. Sunetra Gupta from Oxford)
- a “sophisticated” – but incorrect – analysis of the infection fatality rate
Nassim Taleb and Jordan Peterson, each in their own way, point to the danger of intellectuals ignoring common sense by sophisticated reasoning. Sometimes as paid shills, but often fooling themselves in the process. Taleb calls them IYI’s: Intellectual Yet Idiot. Peterson mentioned a very articulate but dangerous character, might have been Mephistopheles, but I can’t find it back, because he talks too much 🙂
In case you’re still interested in those numbers and all the caveats that come with them, the best paper I’m aware of is Assessing the Age Specificity of Infection Fatality Rates for COVID-19 and its most important chart is this:
However, death is not all there is to life. The initial signal from Wuhan was overwhelmed hospitals, that could even happen with a disease that’s non-fatal when properly treated. Death also doesn’t cover longer term harm. Imagine if HIV was discovered just recently, and we knew nothing about it. If a billion people were infected and we counted deaths a few weeks later, we might mistakingly think it’s not very deadly.
Some of these sophisticated but wrong analyses are made by naive scientists, usually not virologists or epidemiologists. The above paper points out many of their flaws. But regardless of the author’s motives, they’re spread for propaganda reasons, typically by the same kinds of people who downplay climate change and tobacco. People who defend short term business interests, for a tiny set of industries and not rarely at expense of other businesses, either have no ethnical standards or they’re useful idiots.
You’re not going to see the study I mentioned on Fox News. Not because they would actively censor it, but because the PR agencies involved in downplaying the pandemic will reach them first. It’s convenient news for them, so they have no incentive to dig deeper.
To act early or late?
I’ve already established that there’s no such thing as “not acting”, since nobody controls the behaviour of everyone in society. But you’ll find that some parts of society act earlier than others. Some people published a paper on how to handle this pandemic in January. Others were actively talking about it on Twitter and stocking up on masks. At the same time today there are people who express surprise to their doctor about how bad the illness can get. So not all of society acts at the same time.
A well prepared government, and ideally a population keenly aware of the risks, can help a society to act early. It is easier and cheaper to act early, even if you account for false alarms (see the above paper).
If China had acted earlier there’s a good chance the virus would not have left their country. From what we know the first introduction from China to another country was to Thailand on January 13th. For the USA it was on January 20th, with Europe a few days later. It may have happened before, but as far as I know, none of the current circulating virus descends from these earlier events. We know from counting RNA mutations that the virus first jumped to humans in October or November of 2019. Li Wenliang sounded the alarm on December 30th, and if China had a free press it probably would have been public knowledge earlier. Luck always plays a role, but it’s conceivable this story would have ended with news of China suddenly closing its border and locking down a few small areas in December 2019.
But the cat’s out of the bag. Although China as a country responded very late, many areas inside of China arguably responded extremely early. Just consider the amount of territory, not all desert, that did not experience a full lockdown:
Those areas restricted travel, perhaps of their own choosing or with orders from Beijing. China closed its land borders to the west, to countries you may have never heard of, but that doesn’t make them uninhabited and unconnected. Not much is published, but I assume everyone installed their mysterious app with a QR code that displays green, orange or red based on criteria only Xi knows. They did that despite case numbers being (near) zero; that is, early.
The lockdown brought cases down dramatically, whether you trust the precise numbers or not. Lockdowns were effective, though certainly not efficient, pretty much everywhere in the world.
But what happened after the lockdown is more interesting. Where in Europe and the US numbers bounced up again after lockdowns, though sometimes with a few nice calm months, this did not happen in China. The occasional small outbreaks that occurred were acted on early and strongly, with millions of people tested, and (at least) hundreds isolated, after just a handful of cases. Some of that made the news.
Some people, usually those uninterested in honest dialogue, object that you can’t trust statistics from China. This points to a sad truth: it’s very difficult for other countries to learn from China, because even if they are completely honest, their government has ruined trust so badly, it will take decades before anyone can trust them again. But in this case we can say that their approach is successful, even we don’t fully know what the approach actually is. Although they can potentially hide small outbreaks, they can’t hide massive outbreaks. You can’t hide lockdowns that are visible from space, unless they also found a way to fake air pollution. In the west we’ve seen time and time again how small outbreaks lead to giant outbreaks, which lead to lockdowns. So either China effectively stops small outbreaks, or they don’t have them at all. Either is good news, but at the same time not very actionable.
Fortunately there’s a whole bunch of other countries that did really well, especially before summer. They all have in common that they acted early, by closing borders, quarantining inbound travellers and (relatively) broad testing. If something got through the cracks, they took strong measures early, but usually much less drastic than the lockdowns we saw.
When to put out a fire
Fire fighters don’t wait until the whole neighbourhood is on fire before acting. They also don’t stop once the fire is under control, especially when there’s lots of flammable material very close by. A fire is generally completely put out, and then some.
When a small flame comes out of the window, they don’t start with a plant sprayer, wait an hour to see if it was enough, then move to buckets, wait another hour and only then pull out the big gear. You generally throw much more water on a fire than you think is strictly necessary.
The reason for this overkill behaviour is exponential growth. If you don’t do enough, by the time you find out the problem got way worse. If on the other hand you did too much, you’ll achieve the result.
One mistake I see in my own country and elsewhere is the idea of waiting to intervene until things are already bad in hospitals. This is the equivalent of waiting until the whole neighbourhood is on fire.
Another mistake is to take measures that seem “proportional” given their economic impact. This may require several attempts to get it right, while the infection keeps spreading. This is the equivalent of trying the plant sprayer first. Maybe a “partial lockdown” like we have today wasn’t necessary in mid August, but the measures then should have leaned more towards overkill.
So act early, and do more than you think you need. It’s preferable to deal with people who complain you “panicked” than with the consequences of doing too little, too late.
Cure worse than disease?
The dynamic here is similar to the previous section. Any analysis after the fact will result in the same accusation of “panic” I mentioned in the last paragraph. If you understand that, the following will seem repetitive. If not, enjoy!
There’s a Dutch newspaper article calling the Ebola response from a few years ago “panic”. That’s an especially strange choice of words given the quote the headline ends with: “This could have ended badly”.
TL&DR: woman returns from the Congo with suspicious symptoms, goes to a hospital. She was not immediately isolated and spent time in a waiting room, but then rushed into isolation where people in moon suits did their thing … false alarm. Pfew.
Presumably she encountered a lot more people in the hours and days before that. So was it “panic” for the news media to cover this (which they didn’t always do in the past)? I don’t think so. If she did have Ebola, then anyone she came into contact with would have been on high alert. That could make the difference between an outbreak and someone self isolating just in time. We’ll never know. I doubt it caused more economic damage to society from distraction than a medium quality cat video.
There are scathing reports blaming the government for spending too much money on a vaccine for the Mexican flu – back when we named viruses after the place they were first seen. This flu strain turned about to be so mild a vaccine wasn’t necessary, but that wasn’t clear when the order was placed. Concern over repeating the same political embarrassment if Corona wasn’t severe enough, seems to have played a rule in the initial slow response of the Dutch government. The main factor was overconfidence.
The lesson here is that you should not judge the price paid by the actual outcome. Not even by the observed average outcome of recent pandemics. Most of the time it’s not too bad, but occasionally it’s extremely bad. One day we’ll have a pandemic with the potential to wipe out 90% of the population. By the time we realise that, if we rely on an evidence based approach, we’ll be too late. So, for the initial response at least, when the virus is unknown, there’s almost no conceivable maximum price to pay to stop it.
But should we shut down all of society for every new virus that shows up somewhere in the world? Well, let’s start by shutting down international air travel and cancel large gatherings, for a few days or weeks until more is clear. And we don’t have to do it for all new viruses. Taleb in a recent podcast suggested a heuristic: “Don’t worry until 1000 people have died.”
That doesn’t happen a lot, so it’s OK to spend a few billion dollars when it does. Nothing insurance companies can’t handle (if they limit coverage to false alarms). This particular threshold might actually be too high, since SARS 1 killed less than that in total. But some heuristic can be decided upon, such that it’s a small price to pay for having otherwise uninterrupted global travel. Spending $1 billion a year to prevent an occasional $10+ trillion in damages is a good deal. But people will consider it a waste of money and hassle 99 of 100 times. It takes discipline to ignore those people (blocking helps).
OK, but we already know a bit about Corona. So can we now say the lockdowns were a waste of money? No, because we have to judge the decision by the information available at the time, which in late February and early March was practically nothing. The upper bound estimates I saw back then were 5-10% mortality. Worse, these are often biased downward due to lag in reporting of deaths. It could also be biased downwards if only a small percentage of infections is detected. That depends on what percentage of people have mild cases; not something you know a priori (e.g. mild Ebola is rare). So it’s either biased upward of downward; as a decision maker, you have to assume the worst.
OK, but about the new lockdowns? We’ve measured (short term) mortality better, so we can do better cost benefit analysis? Still not really. Long term morbidity is unknown, permanent heart, brain and lung damage even for mild cases is not off the table.
Even ignoring this unknown long term risk, something you should never ignore, you have no basis to compare actual economic damage to. It’s baffling how many people look at the reduction in GDP, count the number of people who actually died, and the divide those. Why would you possibly be interested in the cost per dead person?
The least insane thing to do is to compare economic damage to what would have happened with unmitigated spread. So if the entire country got infected, how many would die, how many hospitalised, how many not hospitalised due to lack of capacity (redo your IFR calculation to take that into account) and how many non-covid patients would have missed a life-threatning diagnosis and treatment. Add the cost of social unrest to that, which of course is completely unpredictable; will sick Dutch people cross the border into Germany and overrun hospitals there? Will the military try to stop them?
Regarding missed diagnosis, it’s necessary to remind people: more covid patients mean less capacity for regular care, which causes more diagnosis and treatments to be missed. Although the anticipation of more patients could lead a hospital to overestimate that and prematurely scale down regular care, it’s still the (anticipation of) the virus that’s causing the health damage here. Lockdown measures save lives here. The same goes for patients avoiding healthcare settings out of fear of getting infected: the lower the prevalence, the less concerned people are.
But you’re not really done there. As I brought up several times, there is no such thing as “doing nothing”. That’s an illusion created by seeing government as the only entity that does things. In reality the economy will take a massive hit, because lots of people shelter in place on their own accord and stop spending. This was very clear early on from e.g. drops in restaurant bookings before lockdowns were announced. It happened again the second wave; transactions in Dutch shops dropped in areas with increased infections, despite no government measures at the time. It was also clear during the first lockdown: most shops weren’t closed, yet many saw their revenue cut in half. It was less clear for bars this summer, because they weren’t allowed to operate at full capacity. But from what I gathered, it was difficult to fill up indoor seating, and so revenue took a beating on rainy days.
The only way to counter the above spontaneous dynamics would be a massive propaganda effort to downplay the virus. If you can get people to underestimate the risk, they might keep spending and they won’t avoid hospitals for regular care. This appears to be the strategy in much of the US. Personally I think such propaganda is evil, and as with communism reality tends to catch up to people.
When a government takes early measures, or like in Hong Kong, when people spontaneously change their behaviour at the slightest uptick in cases, those measures can be far cheaper than a full lockdown. Perhaps all indoor seating in bars and restaurants are closed for 2 weeks every 2 months. If that keeps the virus under control it will look crazy expensive by the actual case counts. No hospital beds will be filled. But it’s actually cheap compared to a full lockdown if you had waited longer. And even cheaper compared to letting everyone fend for themselves.
The biggest challenge is dealing with pundits who don’t get this, which at this point includes half the Dutch government.
The 1953 North Sea flood (watersnoodramp) killed 2551 people, of which 1.836 in The Netherlands. It also caused tremendous economic and social damage, but let’s ignore that, as people who focus on IFR do. Billions, in 1950’s money, were spent on very impressive infrastructure to prevent the sea from ever doing that again.
Although there is an economic case to be made for this infrastructure investment, as opposed to e.g. evacuation, the main motivation for it was to ensure this human tragedy could never happen again. In addition, rather than trying to calculate the precise damage future floods may cause, preventing floods is a simpler and more robust strategy.
So rather than trying to calculate the most economic approach for each pandemic as learn about the virus, it’s more robust to have single policy for all new viruses worse than common cold: wipe it out.
As an aside – in addition to fire fighting and war analogies – Dutch people like to use water analogies when describing the pandemic. In addition to using the word “wave”, “containment” is translated as “indammen” (embanking), “monsterstromen omleggen” (rerouting [PCR] sample streams), and testing & contact tracing are described as dykes (which flooded).
I’m going to be brief here, though perhaps I’ll expand later. The topic of Sweden is almost exclusively brought up by people with no knowledge of the pandemic in general and Sweden in particular. It’s one of those cult memes.
Sweden is currently best described as a voluntary dictatorship under a mad scientist. In any case it’s uninteresting, because they’re doing worse than their neighbours in terms of things we can measure (deaths, GDP). Because it’s culturally unacceptable to criticise government, there’s currently no functioning free press to tell us what’s really going on.
What I know from Dutch journalists who went there is that many old people, who could have survived, were given morphine instead of oxygen. Nursing home staff were forbidden from wearing masks even if they wanted to. So they’re also not teaching us lessons in morality.
Sweden eventually got their case numbers down – to the degree we can trust their numbers; they’re not testing too eagerly. But see my point about China and lockdowns visible from space. Ultimately their streets and bars were about as empty as in places with a formal lockdown. This is where voluntary compliance with social distancing comes in, which doesn’t work in all societies.
This will never end! Just give up.
Quitter talk. So far I have no reason to believe a vaccine is particularly difficult to develop. The only challenge seems to be to do it 20x faster than usual without compromising on safety. Most likely all we need to do is buy time until a vaccine is widely available. The challenge is to make that time as bearable as possible.
That said, if at all possible, it’s best to develop a strategy to deal with this virus on a permanent basis, under the assumption that there’ll never be a vaccine. If not extermination, then at least repeated local elimination. Like in The Matrix, we can get increasingly efficient at that.
The western strategy so far is best described as letting the virus circulate, doing a half-assed job at trying to contain it, then when hospitals are in danger, we lockdown for a few months, gradually relax and repeat the whole thing. I think this is a terrible strategy, and it requires society to not completely collapse before the vaccine arrives. Many countries have been through worse, survived and thrived, but I prefer a better plan.
“give up” is as meaningless as “don’t take action”. Who are you asking? Are you asking the 65+ and immune compromised people to kill themselves? Are you asking government to stop defending the interests of a double digit percentage of its constituents? Are you asking hospitals to refuse those patients so non-covid care doesn’t collapse? Or are you just asking people to stop confronting you with this problem so you can close your eyes to it?
Here’s some countries that endcoronavirus.org mentions as “winning”:
I’m guessing you haven’t even heard of many of these places. You probably heard about Sweden, but not Laos or Bhutan. In general the media pays more attention to countries that take extreme measures (because they were very late), and then kind of ignore them as things get better. Occasionally they’ll show a pool party in Wuhan or a full stadium in New Zealand, because those are really extreme examples. But there are dozens of countries that do way better than the US and most European countries, with far less restrictions. Pay attention to them. Learn from winners, not losers.
Many countries, states, cities and villages have been trying different approaches. Here’s just one way to look at the intermediate results, but remember this thing is not over yet:
If there’s any particular measure you don’t like, you’re bound to find a country in the top left corner that managed to avoid it. Don’t like lockdowns? Check out Japan and South Korea. Want more privacy, check out New Zealand. Want to walk your dog without police permission, The Netherlands suppressed its first wave without bothering dog walkers.
But there are going to be trade-offs. Test, trace and isolate are key ingredients. Maybe if you test and isolate more, you can trace less, which makes sense if you value privacy above physical freedom. On the other hand if you can trace people’s whereabouts and contacts with high precision, you can massively limit the number of people in quarantine.
Border restrictions seem almost impossible to avoid, though Iceland has tried a strategy of more aggressive population testing in order to catch the occasional outbreak from a tourist.
Rapid testing is becoming easier and cheaper, and Slovakia is experimenting with that as an alternative to a full lockdown. In addition to some other measures, everyone who tests positive, or who doesn’t want to be tested, has to isolate for two weeks. That’s strictly better than having everyone isolate, even if the tests aren’t terribly reliable.
We live in great times, in terms of being able to communicate across the globe and learn from people trying all sorts of different things. Drive-through testing seemed like science fiction when South Korea first demonstrated it, yet today seems perfectly normal (the discussion has shifted to making it easier for cyclist and pedestrians, and having vans that can visit less mobile people in their homes).
Given that, there’s just no excuse to shrug your shoulders and say the only solution for your particular area is to let everyone fend for themselves, because you find specific measure X to be completely unacceptable. More likely you didn’t think things through enough.
The game is rigged
If you found none of the above persuasive, I have to disappoint you: your opinion doesn’t matter. The outcome was decided by China on January 23rd when Wuhan went into lockdown, or at least in the weeks following that when they decided to aim for elimination rather than merely mitigation.
This is where Taleb’s concept of the Dictatorship of the small Minority comes into play. This dynamic exists at a global level between countries, but also within much smaller communities. It boils down to this: China will never allow travellers from covid-suspected countries to enter without strict quarantine, if at all. The same goes for many other countries that (nearly) eliminated the virus. Currently people from New South Wales find themselves banned from entering other Australian states, until they step up their elimination game.
There is asymmetry here, no country fanatically wants to spread covid, but some countries fanatically want to keep it out. The non-fanatic countries will eventually concede to the fanatic countries and do the extra work required. This might be under pressure from their own business community, their tourist industry, or their own people who would finally like to travel without hassle.
Severity of the disease doesn’t matter for these dynamics. If China decided tomorrow to eliminate the NL63 common cold coronavirus, because we did something to offend their panda, then the world will have to deal with it. At minimum travellers to China would be screened for it, and a vaccine developed to reduce this travel hassle.
At smaller scales there are people, e.g. the elderly, who fanatically do not want to catch this virus. They will stop hanging out with people who don’t behave similarly. They won’t spend their retirement money on bars, shops and restaurant that don’t take strict precautions. I occasionally listen to Dutch radio commercials while driving, and all of them spend precious advertising seconds to emphasise how safe they are. Even though the majority of their customers are relatively young. Many businesses depend on that last 10% of customers for the difference between profit and bankruptcy.
There is of course a fanatical group who e.g. protests against masks, and some businesses could cater specifically to them. Until staff sues the owner for hospital bills.
One notable exception are poor countries, or regions, desperate for tourists. Many islands kept the virus out, only to let in tourists, watch it spread and needing lockdowns themselves. Much of the second wave in Europe stems from Spain, presumably bar owners could not afford to turn away the uncivil young tourists, spending their parents’ money (and wiping out their own pension). Many expected Thailand to suffer the same fate, but their economy doesn’t depend on tourism as much as people think, and they kept borders tight. Now they’re carefully restarting tourism, but they’re picky on who they let in, how, and from which countries.