Suppression is Working, R is Declining

The reproduction factor is declining. We need to push it below 1 for the virus to start to fade away and then we can move to safety protocols and mass testing.

These estimates are from the Centre for the Mathematical Modelling of Infectious Diseases at the London School of Hygiene & Tropical Medicine. More details here.


so when does Trump can Fauci and claim victory ?

Trump can claim victory when he uses Magic to lower COVID-19 deaths from the current 67 per million to 5 which is a little more than what Australia may end up with. Australia is a nation which received a similar amount of warning and had roughly comparable -- although fewer -- resources.

Density of Australia's biggest 5 cities:
Sydney: 423 per square kilometer
Melbourne: 508 per square kilometer
Brisbane: 155 per square kilometer
Perth: 321 per square kilometer
Adelaide: 413 per square kilometer

Density of US's biggest 5 cities:
New York City: 10,715 per square kilometer
Los Angeles: 3,124 per square kilometer
Chicago: 4,572 per square kilometer
Houston: 1,354 per square kilometer
Phoenix: 1,238.30 per square kilometer

The disparity remains if you look at more cities. Your logic is bad logic.

Your logic is worse.

Seoul: 17,000 per square kilometer.

Your implicit premise that virus deaths in each country are determined by the density of cities in that country is, so to speak, complete bullshit.

Density of Strasbourg - 9,300/sq mi
Density of Grand Est Alsace - (250/sq mi)

The worst affected part of Germany (on the border to the Netherlands) has around 405 people per sq km or around 1200 sq mi

Coronavirus seems completely unconcerned about population densities when spreading, at least in the context of Europe. And Australia's cities (apart from Brisbane) seem to be on the same order of density as the worst effected regions (as contrasted to cities) in France and Germany.

It's not a bad idea on the face of it, but obviously in practice it depends a bit on whether for a particular region it works out that population density = natural social distancing, or whether population density = Australians like to drive (as we all know from the Guzzoline Wars of 2001).

Realistically, how often do people go to stores, children to school, people to work - population density is unlikely to reflect the density found in something like larger retail stores on weekends, schools on weekdays, or various places of employment (banks or warehouses likely have the same respective general staff density regardless of local population density).

I'm not sure what exactly you're arguing here, but people living in low population density places probably do have smaller school sizes, smaller stores within them at any one time, smaller social networks etc. And the people who are in those at any one time are more likely to be the same people on a repeated basis.

School size? Sure. Classroom size? Unlikely.

As for France or America, the Hypermarches or Walmarts are very present in rural areas, and if anything, attract more people than the more urban versions, where the number and variety of stores in the area is much greater.

Germany is a special case. The virus is where the international China business is:
NRW (lots of supplier purchased by Chinese and Chem industry)
BW (automotive Industry)
Bavaria (automotive industry)
The further away you go from high Density locations and travel hubs (eastern Germany) the better it gets.

In the US there's a large cluster near Detroit which would seem to mirror that logic.

Being in the middle of summer might have helped as well, as though that remains subject to debate.

Warmer temperatures help by causing virus particles to break down faster. Just how much isn't clear but it is obviously just one factor helping keep transmission rates here low.

Also, the relative humidity is believed to help slow the spread. Aersol spread is easier in dry weather, just like driving down an empty street is easy than driving in rush hour.

It's a dry heat.

Dry heat in some parts. But the centre of Australian infection is Sydney, and in particular its affluent seaside suburbs, which are sub-tropical and have been humid from February (when a long drought broke) to early April.

causing virus particles to break down faster.

So that's what viruses are, particles? Aren't they actually some kind of really little, like nano-organisms that reproduce? They're already really, really small, do they, in fact, break up into even smaller particles, like teensy-weensy grains of sand or something? Then what happens to them?

Viruses don't meet the generally accepted criteria to be a living organism. Hence the use of the term particles. If a virus particle breaks into smaller particles then it don't work no more, the exception being when a virus leaves behind its outer coat when entering a cell.

More usefully it will be subject to fact, when the facts emerge. Whenever that might be.

Being relatively isolated on frequent world travel routes to Europe, and the relatively successful response in East Asia, does seem to have helped NZL and AUS.

Though ScoMo would probably be pleased to take all the credit (and perhaps Crikey would be happy to give it to him...).

Scomo was slow, but the difference between here and the US was expert opinion informed the political process. After literally thousands of doctors and other medical professionals called on the government to take more action they had no choice otherwise they would be destroyed in the next election if the outbreak got out of hand.

Looking at the US response, expertise was either ignored or given equal weighting with crackpots and politician's gut feelings.

The Sydney cruise ship was a giant mistake.

For decades now the party currently in power has been promising to "Stop the Boats" and they couldn't stop the one boat that mattered.

The US response breaks down into New York City and everywhere else. Non-specific criticisms notwithstanding, Trump has basically followed expert advice as it evolved. Mayor DeBlasio, on the other hand, actively encouraged large gatherings and dragged his feet on closures until far too late.

Trump did nothing of the kind.

CDC director Dr Fauci was telling people to go to the malls, movie theatres and the gym on NBC on February 29 . He is now saying President Trump did not follow the advice of the expert's during February to lock down the country. I think Fauci is confused about the timeline of his own evolution of how to deal with corona virus as more data came in. You can see the NBC tape of Fauci on Feb 29 at Breitbart.

Crikey, Will, get a room.

Kids really seem to engage routinely in the sort of social interaction that favors spreading a noel pandemic virus.

"The Japanese island of Hokkaido announced a second state of emergency on Sunday, weeks after allowing schools to reopen.

“We are facing a crisis of a second wave,” Hokkaido Gov. Naomichi Suzuki told reporters, according to the Japan Times.

The prefecture first declared a state of emergency on Feb. 28 after a number of coronavirus cases were linked to tourists attending the Sapporo Snow Festival. On March 19, as the spread seemed to slow and worries about the damage to local economy mounted, the order to stay home was lifted and schools were allowed to gradually reopen."

Germany: 3,022 deaths

Japan: 102 deaths

You know, some places better than others.

What do deaths have to do with Hokkaido reinstituting a lockdown after seeing a second surge in a short period following reopening schools and businesses, leading to the prefecture acting to stem the number of new infections?

As of April 9th, Hokkaido, with a population of 5 million, has had 226 known coronavirus cases and 10 deaths as a result. The media are not reporting what the sudden increase in cases has been but instead report the increase in cases in Tokyo. I did find that Hokkaido had 156 cases on March 18. So there have been 70 cases added there over the past 25 days.

Always difficult to know how much to quote - "While Hokkaido hasn’t been hit hard by the virus, a double-digit jump in infections last week worried local officials, who on Sunday urged residents not to make nonessential trips outside the home. Schools have once again closed, and aren’t slated to reopen until May 6. The emergency declaration also asks those who live on other parts of the island to avoid traveling to Sapporo, the prefecture’s capital city."

One can assume that the prefecture officials are now completely familiar with the failure of the EU and the US to contain coronavirus when there are only a small number of cases. In part due to a failure to institute lockdowns early enough in the pandemic's spread.

The real point, not emphasized enough in the first post, is the apparent role of schools in virus spread, something not discussed among a group of almost exclusively male commenters here.

I think schools do play a potentially substantial role. Think there is some research out of China consistant with that premise. Obviously the data could change. Unfortunately, I think we’ll witness some natural experiments in Europe that make the role of schools in transmission clear. I really don’t understand why Germany is going to lead with schools reopening. China has done the opposite. I thought governments were supposed to be risk averse. I guess the economic risks are weighing more heavily on their minds.

i am no expert, but my amateur understanding is that, in order for these charts to be reliable, you'd have to know pretty much everything there is to know about the virus.

Yes, Alex is saying that "since R is below one with suppression, we can move ahead to other steps - he proposes safety protocols and mass testing. But this is a complete non sequitur. Knowing R under one regime gives virtually no useful information about what it would be under another regime. If R for safety-protocols-mass-testing is under 1 then we should be doing that already now. If it is greater than 1, then the fact that suppression is working is not a very good reason to adopt s-p-m-t,

Spain will provide an excellent experiment to see what a small change in social interaction leads to in terms of R, and I am certain that this data will be given the proper attention it deserves here, as this blog should be extremely interested in presenting unbiased factual information, such as that which Spain will be providing in a couple of weeks, using the formulation of the highlighted link that new cases lag by around 10 days in the reporting data.

Why no (lagged) inflection point after lockdown?
Surely one of the main intentions was to change the trajectory?

That was my first thought. The dropoff in R should be nearly immediate as lockdowns go into effect.

This is a very big deal if not explained. Perhaps there's some explanation I can't think of. Maybe it's so difficult to estimate R that it takes a month for the estimate to go from its previous value to its true value. But if that's correct, it means all the "middle" values are wrong. How do we know the end values aren't still wrong "middle" values?

Or, maybe R stayed high because it was medical staff getting infected, and their particular Rs were high. Or it took time for the "super-infectors" to gradually decide to stay home.

In any case, the gradual decline cries out for explanation.

I wouldn't trust this analysis. Look at the country names. They didn't clean their data. For example, Czechia and Czech Republic are the same country (

The graphs look pretty, but it's GIGO.

+1, nice find!

Good spot. This was just pointed out to us on Twitter and we have opened an issue to fix on the next update ( Looks like it was caused by a data transition. If you find any more issues feel free to open an issue or submit a pull request to the open-source repo this is all hosted in.

The Vietnam war, the War on Poverty, the War on Drugs, the War on Terror...and now the War on COVID-19.

The virus is patient. It will wait for lockdowns to end. It is becoming endemic, and not just in human populations. Google dogs and cats and coronavirus.

We can flatten the curve, win a few battles, but we will lose the war.
Maybe the cavalry will come in 2021 or 2022, the vaccine.

But by then the unemployment rate in the private-sector could be 50%. A recovery could take a decade (like after 2008), and for many, will never come, in some regards. A guy five years from retirement, for example, might never find another jobs, and will just move into Social Security when he is 65. A youth will have his career start delayed five years, and lose five years income.

Meanwhile, maybe no more meat. Meatpacking plants closing.

Lockdowns are an economic catastrophe.

This crisis really is like porn for some people.

The one saving grace with this coronavirus is the slow rate of mutation, which should make a vaccine viable. But won't mutations increase greatly if it hops frequently between species (cats and humans)? Isn't that what common flu strains do when they spread among pigs and birds and humans?

Mutation rates wouldn't change. Selection for different variants would increase (so selection for mutated variants), but those variants would not have any particular reason to then be successful in infecting humans and evading human immune response. They might more probably be less effective.

Looking at the first three graphs, there is absolutely no way that France or Spain (!) achieved R=1 at the end of March. Particularly as Spain has overtaken Italy with the most European new cases, even though Spain's population is a third smaller. This explanation seems to involve a bit of handwaving, to be honest - "In other words, today’s case data are only informative of new infections about two weeks ago." Lacking uniform standards between countries, it it seems a bit strange to make a uniform prediction, however the data that one sees for Spain at makes it appear that R=1 in Spain is not a fully tenable position. Spain will be providing a natural experiment in how quickly the virus can respond to even a small increase in social interaction. Something that France, Germany, and Italy have zero interest in, until it is much more certain that R is reliably approaching 1, as compared to making models.

And after reading this, nothing like using 10 day old data when posting something today - 'Estimates are shown up to the 2020-04-02.' Germany has been on a real roll coaster ride for new cases, though at this point, the trend line looks better than between April 6-8. Though the start of page says "Using data available up to the: 2020-04-12," the predictions use data that is 10 days older.

Just to clarify this uses up to date data but all plots are shown by date of infection and are hence lagged by the report delay and the incubation period. We have tried to flag this as clearly as possible as it understandably causes some confusion. If you have a suggestion for how we could further highlight it please feel free to open an issue.

That explanation seems to open a new can of worms, to be honest. To me, then this looks like a model built on a model. There is no question that reporting data is inconsistent, as data is subject to a variety of delays before becoming part of a single day's reported cases.

At least the source isn’t the Imperial College. At this point, isn’t it clear that too much faith is being placed in models?

Apparently not, models seem to be the latest distraction available in American based blogs like this. In other European countries, models are generally employed to compare actual data to various projections, to see whether public health measures are being effective. In contrast to Austria, for example, the current UK measures are completely inadequate based on data, regardless of whatever any models might show to be the case.

When the Imperial College came out with their model showing 500k deaths in the U.K. and then, a few days later, reduced it to 20k, supposedly reflecting social distancing, I figured they were using that as distraction to cover up a bad initial forecast. Here in the USA, the models have been gradually ratcheting down their estimates. We’re down to 61k deaths which is getting in seasonal flu territory.

I’ve been following this site for over a month. and they are estimating IFRs in the 0.2-.0.4% range. They also approach the estimates from analysis of a wide range of cases. You can also read their analysis and decide whether you agree with their methodology.

"When the Imperial College came out with their model showing 500k deaths in the U.K. and then, a few days later, reduced it to 20k, supposedly reflecting social distancing, "

I might be wrong, but I thought both the 500K and various lower figures were always in the model, just that the 500K figure (no extra precautions taken) was reported widely, even though precautions were already well under way by that time.

What's with all these models anyway? They're just skinny dames with serious, haughty expressions slinking around in impractical clothes that nobody wears. Sure, maybe they study epidemiology and statistics while they're having pedicures but still.

What I recall is that he revised the numbers down by 96% just two days after the mitigation actions were taken. I just found that pointed out by Alex Berenson, who is now reviled as a turncoat.

I think it is clear that their estimates were grossly wrong, yet very few are pointed out the obvious.

Here's the original study:

If you look at page 13 you can see the estimates for deaths in GB:
Do Nothing - 40-550K
Do a lot - 5.6-40K

Well, that is interesting. I don't remember hearing the low end of the Do Nothing estimate. Do you?


I just scanned the paper and I don't see a 40K estimate for the Do Nothing. I do see a number 0f 410K for DN, R=2.

Sorry that was a typo on my part:

Should have been 410-550K

"Well, that is interesting. I don't remember hearing the low end of the Do Nothing estimate. Do you?"

I remember people criticizing that the author had lowered the numbers and then looked for a source to the paper at that point. I didn't check at the time of the original news stories.

Thanks. This is the type of useful and friendly discussions I like.

Seasonal flu? That is 60K per year, where C19 is 60K until August. And that is only modeling the hospital deaths. The flu numbers will include hospital deaths and deaths in the wider population, which are estimates. If you add in those numbers and scale by the number of months to carry the calculation over, you aren't in seasonal flu territory at all. Plus C19 60K estimate model presumes active social distancing the duration of the model. It makes no predictions about what happens when that requirement is lifted.

Tabarrok and others have noted significant regional differences within the U.S., NYC vs. LA for instance. What we don't know today is what will happen in, for example, Florida: Florida has lagged behind in new cases/deaths, but they are still on the rise. Will there be a spike? Or will there be a flattening out? I fear Florida is on the cusp of a spike. While Florida has an abundance of hospital space, the available space likely won't match the geographic incidence of the virus, requiring patients to be transported long distances (the driving distance from Key West to Pensacola is 832 miles). If the worst is still ahead for places such as Florida, how will that affect relaxation of social distancing in other places?

The data, at least in Florida, can be misleading. For example, Florida's death count only includes Florida residents. Of course, Florida has many non-resident visitors (including snow birds that stay in Florida for months). Is this meant to understate the spread of the virus in Florida?

Wikipedia's Florida coronavirus page shows the percentage increase in the number of cases has fallen from 18% on March 31 to 7% on April 12, so the virus has been winding down.

Florida deaths increase from the previous day:

April 1.....101
April 2.....144....+43%
April 3.....170....+18%...+31%
April 4.....195....+15%
April 5.....221....+13%....+14%
April 6.....254....+15%
April 7.....296....+17%...+16%
April 8.....323....+7%
April 10...419...+13%...+10%
April 11...446.....+6%
April 12...461.....+3%...+5%

"For example, Florida's death count only includes Florida residents. "

As far as I know deaths are reported as to where they were recorded. Do you have a source that says differently?

This is a good point. US is collection of states/regions and we will have different policies in different states/regions depending on state R.

Doesn't a declining number of daily new infections (which must have happened 2-4 weeks before a declining number of daily deaths) necessarily mean that R was less than one at that time?

... to objectively assert that a process-intervention ("suppression") is "working" -- requires causative proof that the specific intervention is the key variable in changing the process outcome.

That proof is totally absent here.

There are many variables in this process. A control-group is probably available in the mash of world data -- but difficult to isolate and/or ignored.

Odd that R is following a similar decline in many countries that have taken very different strategies towards lockdowns.

Apart from the UK and Sweden, basically all European countries have taken the same broad approach to lockdowns.

Has the UK been drastically different than everyone else.

Not yet as strict at least, as the UK numbers show the result of not learning quickly from the experience of Spain, Italy, or France.

Imperial put together a nice little timeline in one of their papers -

Doesn't cover a number of countries though.

You'd need to evaluate it relative to confirmed and estimated cases / deaths to understand if measures were relatively earlier / later in the true infection curve etc.

Death rate increase has gone linear with new deaths roughly level since last week. That's probably a better indicator of flattened real infection rate than confirmed case count increase, which probably reflects increase in testing rates. 100% what you'd expect given the lockdown timing and the high compliance rates (very comparable to other European nations).

If I'm drawing the correct conclusion from the graph, the UK is pretty standard but delayed by 10-11 days.

For the lockdown measure, sure seems like that. Closure of public events and schools lagged heavily (but probably didn't have a big effect), while case isolation and social distancing recommendations early-normal. (Though whether case isolation actually works depends of course on testing prevalence, which I have heard worked very well early on in late Feb-March, but probably ran into low testing resources / policy in mid-late March).

The title of this post suggests that suppression is the principal—perhaps the sole—cause of the decline in R. But no evidence is provided for that suggestion.

I'm wondering why R is not below 1? It feels as if we eliminated 80-90% of human interaction. Wouldn't that turn an R=3 to R=0.3?

I've been curious about that as well. We are well past any incubation lag from when shutdowns started happening.

We'll know in a week.

Seasonal flu and other respiratory viruses have been eliminated two months early.

The question is why this respiratory virus has not.

Answer: pets. We need to cull the pets.

" It feels as if we eliminated 80-90% of human interaction. "

You are living in a bubble. When's the last time you drove your car to the nearest shopping center? Or checked out rush hour traffic? Have you seen the pictures from the NYC subway? It's still crowded during peak times.

Reports are the mass transit use is down 90%. My human-human interactions have dropped dramatically. A weeks worth of interactions today was easily done in a morning before and what interactions I do have are as distanced as possible.

+1, your information is more up to date than mine. As of this week Subway ridership in NYC is down 90%, it was much higher just a few weeks ago.

I believe they have reduced the number of subway runs in NYC; that is what my son who lives there told me.

Alternatively R0 might be closer to 6 than 3.

It seems to have taken 13-14 days for the new cases to peak from the lockdown. There’s the incubation period ( 5 days mean), the fact that some people share a home and can infect each other, some oriole in stores and still some essential workers about , some of whom are often exposed to the virus ( Health care workers ). Wuhan took 12 days to peak, Italy 17.
After the peak there’s often an equilibrium plateau for a few days.

Speaking of modeling and epidemiology, this is fun. We have "missing" heart attacks:

"A model looking at activations for ST-segment elevation myocardial infarction (STEMI) showed a drop of 38% (95% CI 26-49) from roughly the year prior to the outbreak to the first month of it."

This too:

"I've spoken to more than a few ppl who would have had knee/shoulder Sx 4 weeks ago... now they are better."

Your second link could point to proof that American medicine is over utilized and one of the benefits of single payor queuing is that people don’t use medical services at several margins. And the results seem to be no worse than if they received medical treatment.

Or staying home, no traffic, lower the number of heart attacks that would have already happened.

Also I suspect the human body has a limited capacity to put some medical issues 'on hold'. For example, you hear about people who die after Christmas or some life event like their child getting married.

One does wonder why they would list Czechia and the Czech Republic as two separate countries.

Hospitals in Fairfield co hit hard tho emergency rooms free of car accidents ....this is a great piece that’s harder to explain

Please post the code and data source(s) used here. This would be very helpful.

Looks like a Rorschach Test for detecting cognitive bias.


I meant for all of us.

There's no community testing at all in the UK and I believe not much at all in the US. So for both countries, there's no empirical evidence to based those graphs on. For Germany (and hopefully some other countries mentioned), the evidence is much stronger - but their measures are far stronger. In the UK, millions still work in unsafe conditions, many with no possibility of proper social distancing. No doubt that social distancing can work (albeit in combination with testing, tracing, quarantine, treatment), just questioning the UK and US figures.

Good old confirmation bias.

R always declines after you reach a certain level of immunity. You want R to be high in the beginning, so as to achieve level of herd immunity, and then R declines rapidly. This is how all virus epidemics, including annual flu, works.

+1, my concern and I could not track down the article. And we have no data yet on immunity period, we assume seasonal?

Sure, if we all go back to the Stone Age or at least the 4th century and live like nomads in tents or quasi-like igloos, of course social distancing will work. Is this what we want for one of the more advanced economies in the 21st century? Sorry, I know it is taboo to talk like this.
But you have Bill Gates saying that *everything* should be shut down for 18 months - and nobody said even a whimper.

Bill Gates should learn to be quiet. Just shows how dumb people can be outside of their areas of expertise.

M.E. Trump may declare victory now. I bow before you. You have dog walked me. Now go away, all three of you. I neither like, love, nor hate any of you. Be gone. NOW

ALL of the HOTSPOTS on the planet appear to be located in one of four areas:

(1) The ancient core of China

(2) The Western Roman Empire

(3) Places in America with immense mileage in tunnels (NYC has an exceedingly high density of tunnels and on top of that still relies on a combined sanitary/storm sewer system ~ particularly in the older parts ~ Manhattan, Queens, Brooklyn.

Let me suggest that these immense tunnel systems built by both old and modern major civilizations allow the virus to be moved rapidly from Patient Zero to everybody in the region quickly and efficiently!

BTW, Iran has nearly 300,000 miles of qanat...... Italy is a honeycomb of ancient watercourses, as is Spain, plus Spain has been heavily mined for thousands of years. The Chinese core area where China was founded is similarly covered with tunnel systems.

Models based on assumptions of how rapidly a virus may pass from one human for another likely take no accounting of the massive tunnel based distribution systems that allow viruses to pass rapidly for dozens to hundreds of miles....

Comments for this post are closed