Houston

Houston. Yes, I suppose they had to have a public memorial for George Floyd. That was 19 days ago.

And, I've linked to Political Calculations, who estimated that an upturn in cases occurs 13-16 days after an event.

Now this:

Texas Medical Center censors hospital ICU availability data. That gives us the right to assume the worst, until proven otherwise: no more beds? way over capacity? They don't trust the public with alarming news? Entire IT staff out sick with COVID? 1/2https://t.co/ejj1P1b6oG

— Jody Lanard MD (@EIDGeek) June 28, 2020

Texas Medical Center in Houston is one of the largest hospitals in the world. It is believed they hit their ICU capacity the other day. Then the data stopped.

Anyone recall in late February when I mentioned in class that it was really weird that Iran had "gone dark"?

Sweden

Remember when Sweden went a different path? They decided to not seriously lock down, and to let people handle the pandemic in whatever way they thought best?

One country looks different from the rest... #covid19
(via @ECDC_EU which, ironically, is based in Stockholm: https://t.co/JX8iHgfBBv) pic.twitter.com/BmzPTyTPhu

— Kai Kupferschmidt (@kakape) June 26, 2020

Rates, Rates, Rates … If It’s Not In Rates, Beware

The world’s most famous pop-statistician gets it:

The relationship between the current rise in COVID cases and the partisanship of a state/city/county is much weaker and more complicated than the commentariat seems to assume. https://t.co/m5wOcjZFTS

— Nate Silver (@NateSilver538) June 26, 2020

Do not get me wrong: it’s bad in Arizona, Texas, Florida, and other states right now.

BUT, most of the things you may see saying that it’s super-wrong in those places are based on either total cases, cases not scaled by population, or both.

But what you need to look at is new cases. And you need to scale those by the population of the region they’re measured over.

When you do that, you find COVID is getting worse in red states and blue states.

Testing Numbers vs. Case Numbers

Trump asserted that more testing leads to more cases. Not necessarily. Let's think this through.

If cases were distributed randomly, and tests were done randomly, then there should be no relation between the two, right?

If cases cause some infected people to feel ill (and note that it need not be all of them), and feeling ill motivates getting a test, then yes you'd expect to see a positive relationship.

On the other hand, if cases is what encourages people to get tested, and the capacity is there ... perhaps you end up testing a lot of people who weren't likely to be infected in the first place. In this case you'd expect to see a negative relationship.

Here's the thing: Trump's story (shared by many) is about looking at just two variables, but the stories I outlined above have extra variables.

Not sure about that? Look closely. The phrases "... feeling ill motivates getting a test ..." and "... cases is what encourages people to get tested ..." imply extra variables(s). That implies that both Trump and the smart people at the COVID Tracking Project who produced this graph are oversimplifying and are probably subject to omitted variables bias:

The Trump argument is probably best exemplified by Louisiana and Georgia: differences in rates of tests and rates of positives. But for the most part, the rest of the chart shoots down the Trump logic. Now it doesn't show his logic directly, since it does not have a number or rate per 100K for cases along the horizontal axis. In fact, Trump's story will check out if we were to regraph this to show cases rather than rates, and just the states to the right of the dashed line.

But even so it would miss the most important part of the story told here. Look at the scale on the left. Arizona, for example, is testing roughly 200 out of every 100,000 people every day. That's less than 1%. What's going on with the other 99%? Trump is right: if we tested all those other people the number of cases would go up. But that's infantile: what we really need to know is by how much.

As always, drawing attention to Utah, and ... we look pretty crappy right now. We're just not testing that many people, and we get a lot of positives amongst those we do.

How Utah Is Different from Some Other States

I have been meaning to post about this for 3 weeks. Ironman at Political Calculations came up with some interesting skyline charts for state level data on COVID-19.

Originally the data was from June 2nd. They've now been updated to June 9th. I hope he continues to update.

Anyway, you read these by going top down: top is the past, the bottom is the present.

The width of each panel is scaled to represent 2% of each states population. So, wider bars mean a worse outbreak. Do note that the horror story that has been New York, is, proportionally only 2-3 times as bad as most other states.

And it's colored: black is deaths, dark brownish orange is people in the hospital, green is people who've been in the hospital and were discharged, and the paler orange is number of positive tests.

The top row is fairly easy to figure out: lots of deaths but they're not growing much anymore, many people testing positive. Not really surprising for those northeastern states.

But I'd like you to go down to the 7th row, where Utah is on the left. The chart shows very few deaths, and not many hospitalizations. But discharges are almost as large as tests.

What this tells me is that just about no one is getting tested until they need to go to the hospital. But once there, they're getting checked out fairly quickly.

And there's 3 levels of illness we should be concerned about:

1. You're infected but not sick enough to go to the hospital.
2. You're infected, and sick enough to go to the hospital.
3. You're infected, you're sick enough to go to the hospital, and you're so sick they won't let you leave.

I am OK with Utah hospitals pushing people out the door: they are worried about getting overloaded, and also about exposures of healthier people. By all accounts, many people can recover from COVID-19 at home. I trust hospitals to keep the people who should be kept, and in Utah we haven't had too many of those.

What worries me is that Utah is mostly people in # 2. (I'll skip Arizona which has exploded since this data was updated). But when I look at other states' charts, I see a lot of places where it's mostly # 1's: Kansas, California, Ohio, Nevada, Washington, Florida, Colorado, Missouri, Georgia, Virginia, and so on. Those are the ones I'm not really watching for news about.

Where are the states that match up with Utah (again, not counting Arizona in this one)? Iowa, Michigan, South Dakota, Mississippi, Minnesota, Tennessee, North Dakota, New Hampshire, Wisconsin, North Carolina, Arkansas, South Carolina, Texas, Maine, Oklahoma, Idaho, Wyoming, West Virginia ... those are the ones I'm watching for news about. And none of them has really blown up yet, so maybe this all means nothing. We'll see.

SARS-CoV-2: It Only Wants Targets

Would you believe this was Adam Elkus posted this to Medium on March 13th, our last day of face-to-face classes? Before New York City?? Now that's prophetic.

... The virus has ... a power that makes it uniquely dangerous to Western society: it is utterly stupid. Scientists and philosophers debate whether viruses are even properly counted among the living. But whether it is alive or dead does not matter.

It exists, and the only thing it wants is targets.

It does not think, it does not feel, and it lies totally outside the elaborate social nuances humans have carved out through patterns of communication, representation, and discourse. And this, above all else, makes it a lethal adversary for the West. It has exposed how much of Western society – but American society in particular – is permeated with influential people who have deluded themselves into thinking that their ability to manipulate words, images, and sounds gives them the ability to control reality itself.

They implicitly or explicitly assume that by attaching labels and names to things, they can control them. They implicitly or explicitly behave as if control over narrative is control over the things narrative is attached to. The virus therefore was a problem of psychology before it was a problem of microbiology, because people did not have the “right” attitudes and words for something that in and of itself was incapable of having attitudes or making words. And from the President on down, politicians behaved (and are still behaving) as if it was something that could be spun or narrativized away.

...

But I would like to return to the obsession with using words to control reality. There were endless attempts early on to compare it to a less-threatening entity, the flu or even the common cold. In doing so, institutional actors tried to take something new and uncertain and fit it into a tame pre-existing mental model that they preferred. Acknowledging the virus as a creature of fate – of fortuna – would be to admit that it could collapse the elaborate machinery for making narrative and reveal the narrative-makers as utterly impotent.

Managing public health and disease was one of the core tasks that helped build the legitimacy of industrial era government in the 19th and 20th centuries. When civil servants are too burdened by bureaucratic red tape and the need to perform political face-work to properly pursue this endeavor, it is a sign that Western society has traded the substance of political competence for its appearance. And more generally, a society that cares more about declining trust in institutions than what institutions have substantively done to deserve trust – and which devotes far more effort towards managing the behavioral psychology of risk than actually reducing risk – is engaged in narrative-making as a singular pursuit above all else.

Which is where our virus comes in. It is a very simple creature, unburdened by all of this discursive weight. To the extent it can be said to have desires and needs, they are very humble. It only wants targets. We lack a working vaccine and estimates vary about how fast we can get one, but it was born with a natural immunity to our capacity to distract ourselves with our silly little language-games. As this seems to be the most powerful weapon our society had up to this point, we will have to go to Plan B: actually doing something to alter the situation instead of hoping that things will change if we come up with nicer-sounding words to describe it.

I have quoted most of this article. There's an extensive run into some quasi-Marxist mumbo-jumbo in the middle that I cut out.

Via bookofjoe, which almost never posts about stuff like this.

Arizona Is the New COVID-19 Hotspot

Yeah ... I know ... you already knew this. But I don't think you know as much about it as Ironman writing at Political Calculations. He knows the data better than politicians and the media, and he does the stats more insightfully than most dashboards.

This chart is interactive. Go and play with it here. Do note that the vertical scale does not need to be logged because it is a rate. Even so, I'd prefer it if it was, since the rate is cases divided by population, and the latter isn't really changing over the sample period.

Oh ... and where's Utah? It's the 4th highest pink dot on the far right. Uh-oh.

Death rates lag case rates, so Arizona merely has a high death rate, but it doesn't stand out yet. However, he estimates a 20-28 day lag between policy changes and deaths, and the uptick in deaths in Arizona starts 23 days after the state re-opened on May 15.

He also estimates a 13-16 lag between an event and it showing up in cases. That makes the above look like a George Floyd protest spike. Not sure why it's worse in Arizona (other than the virus was already heavily seeded into the state through the spike on the Navajo reservation in early May).

One thing that's new about Arizona is the proportion of young people being diagnosed. It's up in all groups, but has more than doubled for people between 20 and 44.

Computing and Data Science Master’s Programs

A list of programs from Economics Job Market Rumors.

I Am Scott Alexander*

Scott Alexander† has deleted his blog because the most influential progressive publication in the legacy media threatened to use its reach to dox him. Anyone with an open mind should be offended.

The blog was Slate Star Codex. It was so high brow it was tough to read sometimes: literally exhausting to process.

Slate Star Codex had been around for perhaps 15 years. I have linked to it many times, both on this blog and others. I still have some Slate Star Codex webpages open in tabs on one of my PCs (and probably more in my caches).

Scott Alexander titled his blog Slate Star Codex to retain anonymity. He is a psychiatrist in private practice, who lives by separating his professional life from his private one.

Slate Star Codex was to be the subject of an article in The New York Times. The employee putting the article together told Scott Alexander that he would reveal his true identity in the article (this has happened to other bloggers, like those behind Zero Hedge).

That employee claimed this is a policy of that newspaper. This is false. As noted by Geoff Lewis, one telling example is that the Times has published over 1,500 about Banksy while preserving his anonymity. Other examples are easy to find.

Scott Alexander’s response was a post entitled “NYT Is Threatening My Safety By Revealing My Real Name, So I Am Deleting The Blog”, and then he did. The post contains details about the situation, fears, and advice to others.

Do note that, of course, The New York Times can still dox Scott Alexander. They hold all the cards, because in a situation like this all that counts is the willingness to play them.

I think I’m going to puke.

* I am shamelessly copying Tyler Cowen.

† Apparently this is part, but not all of his real name.

************************************************************************

You wanna’ know why I read stuff like Slate Star Codex? Here’s an example of something I’ve been ruminating on since I first read it there a month ago, but have yet to post about.

What is it? All our experts need to explain Switzerland. Switzerland has a national policy for COVID-19, that … didn’t work equally well everywhere. Gee, do you think that might be important?

it turns out that German areas of Switzerland have far lower death tolls than the French and Italian regions (no word about the Romansh region). This matches the outcomes in Europe so far, where Germany has far fewer deaths than France or Italy. It is astonishingly important to establish how shared language and culture can influence the progress of a pandemic, and furthermore seem more important than the well-intentioned policies of bureaucrats and politicians.

You simply did not find insights like that if you, say, only read The New York Times.

BTW and an Apology

Way back in January, BD raised the possibility in class that SARS-CoV-2 was the product of Chinese experimentation. I knew about this claim, but at the time I was dismissive.

Also around this time, the financially-oriented blog Zero Hedge made assertions that such work might have been done by specific scientists at the Wuhan Institute of Virology.

Twitter accused Zero Hedge authors of doxing, and banned them for life.

As we've gotten more information about the beginnings of the outbreak, I've taken a Bayesian position that my posterior probability is being driven towards it being a human-made problem (if maybe not quite a human-created virus). Apologies to BD.

Also ... Twitter quietly rescinded their lifetime ban on Zero Hedge a couple of weeks back.

Some History of Lockdowns and Social Distancing

In class in January 2020 I emphasized that it was a hugely unusual move for China to lock down healthy citizens. I noted that quarantine was what you do to the ill, not to the well.

In class in March I noted that extent epidemiological models (SIQR models) have not even been developed which incorporate isolating the well.

By this time, Italy — following China's example — was considering isolating the well. They did so in stages, beginning on February 21st, and extending to the whole country by March 9th.

In the U.S. system of federalized government, lockdowns are the decisions of governors and mayors. San Francisco's mayor was the first to impose a lockdown on March 17th. The Governor of California followed suit 2 days later.

***************************************

Three months on, it's not clear how useful lockdowns have been. I don't think anyone doubts that they help fight disease, but reasonable people can and do disagree about whether that is worth the cost (here is an anti-lockdown estimate).

The American Institute for Economic Research (AEIR), an austrian outfit, has the historical details. It's ugly.

Slogans like “flatten the curve” massively oversimplify social processes and human choices, and presume to know far too much about cause and effect.

...

Remarkably, none of the models even considered the impact of the virus on long-term care facilities, and hence contributed to gross neglect of the population that should have received the bulk of the focus.

...

... Epidemiology similarly took a turn for the worse around 2006, when agent-based modeling strategies displaced the accumulated wisdom...

And what happened in 2006?

Let’s start with the phrase social distancing, which has mutated into forced human separation. The first I had heard it was in the 2011 movie Contagion. The first time it appeared in the New York Times was February 12, 2006:

If the avian flu goes pandemic while Tamiflu and vaccines are still in short supply, experts say, the only protection most Americans will have is “social distancing,” which is the new politically correct way of saying “quarantine.”

But distancing also encompasses less drastic measures, like wearing face masks, staying out of elevators — and the [elbow] bump. Such stratagems, those experts say, will rewrite the ways we interact, at least during the weeks when the waves of influenza are washing over us.

Maybe you don’t remember that the avian flu of 2006 didn’t amount to much.

And what about lockdowns? It gets worse.

The administration of Bush II was concerned about pandemics, so it tasked an intensive care doctor and an oncologist with no experience in pandemics to formulate a plan.WTF?

It gets worse. They relied on a seminal piece of work — based on a high school science project — by a young woman who now refuses to talk about it. She'd written an article with her dad (a serious mathematician, with no experience in medicine or economics) which showed that because school kids have bigger networks, shutting down schools is the most effective tool to reduce infectious spread.

It gets worse.

The Bush II administration ended up choosing lockdowns as a policy over the objections of epidimiologists. More on that in a moment.

The import of this is that the Obama administration inherited a "playbook" for pandemics that included lockdowns. It was a major news story for a few days in mid-May that the Trump administration had ignored it. But the thing is ... apparently they didn't. They may have thrown out the paper version but the mental model of how to proceed was ingrained into the swamp.

The New York Times has the whole political story. Why Bush II? Homeland security concerns. But here's what the epidimiologists wrote:

There are no historical observations or scientific studies that support the confinement by quarantine of groups of possibly infected people for extended periods in order to slow the spread of influenza. … It is difficult to identify circumstances in the past half-century when large-scale quarantine has been effectively used in the control of any disease. The negative consequences of large-scale quarantine are so extreme (forced confinement of sick people with the well; complete restriction of movement of large populations; difficulty in getting critical supplies, medicines, and food to people inside the quarantine zone) that this mitigation measure should be eliminated from serious consideration…

They were outvoted.

New Southwest Utah Numbers

The numbers are now posted, including both yesterday (phew) and today.

In raw terms, they are still increasing. This is not good for hospitals. Doubling in both counties is 17 days (around where it's been for a couple of weeks).

Garfield County is up to 9, that's 3 today, and 1 yesterday. Still very low, but a disturbing uptick. There's also one more in Beaver County.

And there was another death in the region over the weekend, so we are up to 10.

Coming to a Red State Near You

COVID-19 originally hit blue states. Now it is mostly in red states:

That was new cases. Here's total cases (this one is hard to interpret because the scale is not logged):

Because the data isn't logged, you can't interpret the slopes on the left as growth rates, making it impossible to compare with the first chart. Basically, once the red line passes the blue line in the top charts, then the reddish border at the bottom right will start angling up more steeply.

These are from the Washington Post piece (probably paywalled) entitled "Coronavirus Has Come to Trump Country".

It’s Not Race, It’s Public Transportation

Quite a bit of casual discussion over the last 2 months have focused on disparities in death rates between African-Americans and non-African-Americans.

It is important to emphasize what makes this discussion casual. Roughly, this comes from binning deaths by race, then dividing by the population of that race, and comparing the resulting rates. This is easy to do. But it is also fundamentally univariate, and thus subject to the most important form of bias in statistical work — omitted variables bias.

Pretty much no one should do analysis this way, and yet the media and politicians emphasize this sort of junk prominently.

There’s new research on this, where rates of death are regressed on a host of available data at the county level. I’ll quote from the conclusions:

The main findings as follows:

(i) For all four minorities, there is a strong positive correlation across counties between the minority’s population share and COVID-19 deaths. (ii) For Hispanic/Latino and Asian minorities those correlations are fragile, and largely disappear when we correct for education, occupation, and commuting patterns. (iii) By contrast, for African Americans and First Nations populations, the correlations are very robust. Regardless of what other factors are controlled for, a racial disparity in mortality rates persists. (iv) Surprisingly, for these two groups the racial disparity does not seem to be due to differences in income, poverty rates, education, occupational mix, or even access to healthcare insurance, which has been hypothesized by many observers to be a key source of the disparity. (v) A significant portion of the disparity can, however, be sourced to the use of public transit, which also explains a large fraction of the difference in mortality between Los Angeles and New York City.

Of course, while use of public transportation is generally associated with income within cities, it is not associated with income across cities. So this is not an absolute poverty result (which they did correct for), but rather a relative poverty result. In short, New York City was hit so heavily, particularly amongst minorities, because while it is on average a rich city, it is one in which the poor and minorities ride public transportation the most.

Do note that this is consistent with evidence reported in my February 26th post that evidence from Wuhan showed that shutting down public transportation we effective in containing the outbreak there. Once again, this tends to confirm poor choices made by political officials in New York City as leading to high death rates there. It is almost certain they would have had access to the same information that I did.

This also dovetails with a post I made last year about why infrastructure projects cost more in the U.S. than in other countries. That detailed a high degree of chauvinism amongst U.S. city politicians: if an idea does not come to them from within the U.S., they don’t listen.

COVID-19 # 68

Mexico:

I’m very worried about the pandemic in Mexico.
This is the test positivity rate of the country. Pretty much linear growth, now at 56.7%.
This means the country confirms a case for every 1.76 test they do.

The source is our #COVID19 profile for #Mexico https://t.co/LbsZ5RS6uj pic.twitter.com/pn0sk6pPmd

— Max Roser (@MaxCRoser) June 21, 2020

New York City, at its worst, was over 55% positives for the 2 weeks around April 1st. This result is for the whole country of Mexico.

There has been some hemming and hawing that Mexico isn't doing that many tests. And this is true: they had about 4,700 positives yesterday, so they must've tested about 8K people. But that's roughly how many test NYC was doing around April 1st too.

******************************************

Chile:

Chile has 5-6 thousand new patients daily. System can't handle it. We will soon be in the Italian scenario. https://t.co/xED9k2ZUFE

— Crawford Kilian (@Crof) June 21, 2020

******************************************

Arizona:

Since last Sunday, 20% of the reported tests in Arizona have come back positive.

Right next door in New Mexico, only 2.5% of tests are coming back positive. pic.twitter.com/CnAuYgvpUD

— The COVID Tracking Project (@COVID19Tracking) June 20, 2020

Arizona is doing about 15K tests per day, so about twice as many as Mexico, or NYC (when it was just starting to ramp up testing).

******************************************

Brazil:

Set a record today for most positives in a single day, surpassing a record set by the U.S. in April.

FWIW: the U.S. is still getting daily numbers of positives that are running at about 80% of the record rate. That's the big slack off people believe in.

******************************************

Southwestern Utah:

Our numbers were not updated today. Sometimes that can be a very bad sign.

Dr. Osterholm's Thoughts

I'm linking to this because it supports what I wrote the other day.

Dr. Michael Osterholm, the director for the Center of Infectious Disease Research and Policy at the University of Minnesota, warned on Sunday that the country was likely to experience one long stretch of cases, hospitalizations and deaths. 

“I don’t think this is going to slow down. I’m not sure the influenza analogy applies anymore,” he said on NBC’s “Meet the Press,” referring to a report he and colleagues wrote in April using influenza pandemics as a model for understanding the virus. “I think that wherever there’s wood to burn, this fire is going to burn it.” 

“I don’t think we’re going to see one, two and three waves — I think we’re just going to see one very very difficult forest fire of cases,” Dr. Osterholm said.

I think the idea of waves sounds pretty. But I have trouble conceiving of a mechanism by which it would happen.

Moreover, let's take New York City as an example. Currently they have just under 400K positives, and a population of 8,000K. That's about 5% of the population. So, if herd immunity kicks in around, say, 80%, NYC will have 16 more waves this size.

As I wrote ... waves sounds pretty. Does 16 more of them sound pretty?

New Evidence from Italy of Earlier Spread

Italy's first community spread case occurred in mid-to-late February. Prior to that, there had been some cases linked to travel from China.

But now, tests of sewage samples collected in Milan and Turin on December 18 show evidence of the virus. Other cities showed evidence from January.

(Don't ask me who keeps sewage samples around for 6 months, but I guess that's what public works scientists do).

These dates make it harder to maintain the official story from China that the first case was in late November.

However, they are consistent with the evidence that there was something going on in the health system in Wuhan before mid-October (I covered that here), and the conjecture that international spread was related to auto shows held in Wuhan in early November (I covered that here). Wuhan is the center of auto manufacturing in China, and Turin is the center of auto manufacturing in Italy.

Here is a post detailing the results.

China's New Outbreak

China has released genomes of virus samples from their new outbreak in Beijing. These show that they are a strain from Europe, not from Wuhan and/or Hubei.

They're pretty sure that strains are close enough to each other that getting one confers immunity to all of them.

But, this points to the problem that we can't think about this in terms of countries. We're forced to do that because of the way the data is collected, but that isn't the way a pandemic works.

There are now roughly 3,000 strains of COVID-19 out there. The one currently breaking out in Beijing originally developed in Europe (it's closest relatives were most commonly found in England). The thing is, it broke off from those strains perhaps as much as 3 months ago. So it's been simmering somewhere (probably Beijing, but really along any path connecting airports).

The place to go to see this Nextstrain.org (which I linked to about 4 months ago). It's like a family tree.

The oldest strains are on the left. Blue is strains collected in Asia. Each branch is an estimate, based on the number of similar and different mutations between strains, of where each one began. For example, perhaps we start with AAA, which mutates to ABA and ACA, and then the first of those mutates to both ABB and ABC, while the other mutates to ACB and ACC. So it's basically a family tree.

The current outbreak in Beijing is that blue line near the top that extends the furthest to the right (most of the ones near it are in Bangladesh).The olive green color is Europe. So you can see that the European strains break off from those in Asia in late January, mutate most wildly there, and then went back to Asia.

The sense in which this has been circulating for a while, presumably in Beijing, is that those blue lines stretch back into early March. There are over 100 cases there now, and you know how it works: 1, then 2, then 4, then 8, then 16, then 32, then 64. So this has doubled about 6 times before they caught on to it.

This is part of the reason that I am very discouraged that we will ever get thing under control. Strains come and go, but there are just too many out there now, in too many places.

UPDATE

I don't feel that I emphasized this enough. Just because a strain in one place is close to a strain ni another place, it does not mean it "came from there". So China should not blame the UK.

Instead, it means that the strain observed in one location at one point in time, was similar to strains observed in another place at another point in time. There may have been a ton of travel in between, and even development of intermediate strains that were not observed.

Here's a look at the detail, similar to the one above, from a different source:

Combined with the above chart, this indicates that it's probably been a while since relatives of the strain currently in Beijing were in China. And relations were definitely popping up in Germany, England, and Taiwan along the way. But it isn't the same strain in Beijing as in any of those places.

BTW: the GISAID initiative, as of today, has received about 49K submissions of viruses from around the world, but only about 3K of them are different from one another.

Not Going Away In the Summer

Colds go away in the summer. The common cold is caused by a family of viruses.

Influenza goes away in the summer. Influenza is caused by a family of viruses.

SARS-CoV-2 is a virus. Therefore COVID-19 will go away in the summer.

I hope you can see the non sequitur there.

Nonetheless, this is the way a lot of people have been and continue to think.

I hope these two chloropleths put that to rest. Here's where case rates are increasing the most:

And here's the average temperature (this is for May, for June the colors would change but the gradient would stay the same):

FWIW: I lived in New Orleans for 9 years, and Tuscaloosa for two. It's June 19th: in those cities you pretty much did not open up the windows from mid-May to mid-September. It's almost like being cooped up inside in Utah in the winter. Last I checked, the virus likes that too.

Another Death In the Region

The Southwest Utah Public Health Department announced a 9th death in this afternoon's release. They note that further details are pending.

A Natural Experiment with Anthropogenic Climate Change

You all know what climate change means. Anthropogenic is the modifier that means caused by humans.

Almost all of what you hear in the news about anthropogenic climate change is based on correlation analysis. Correlation is not causation, but of course causation always has correlation if you dig deeply enough. To make things worse, we're working with trending data when we compare pollution to temperature, so we have to worry about spurious correlation.

Sooooo ... smart people look to natural experiments, and we've just had a doozy:

Look at the source: this data is not from climate change deniers.

Here's the thing: if we don't see a dip in global temperature trends that's correlated with this, then we can probably reject the anthropogenic part of climate change from future discussion.

COVID-19 # 67

Of course, we still have cases going up just about everywhere. The world is now over 8 million positives. But you've got to keep your eye on days-to-double-cases. It took us about 5 weeks to go from 4 to 8 million, and that's slower than all the previous doublings. That's good.

Here's the thing though. To get from 1 to 8 million, you need to double about 23 times. To get from 8 million to 8 billion ... that's everyone on the planet ... we need to double only 10 more times. I don't think we'll do all those in 5 weeks a piece, but if we did, it would take about a year. When I look at numbers like that, I think we're likely to get to global herd immunity before we get a vaccine. Of course, the hard part of that is we might get 100 million deaths along the way.

Most readers of this blog have had me for class, and know I'm really optimistic about nearly everything.

I'm not optimistic about COVID-19. I'm increasingly coming around to the view that organizations like the CDC and WHO are ... actually sugar coating the situation.

While I'm an optimist, I also tend to be fairly contrarian. It's an ugly feeling to be have many people around me suspicious that things are not as bad as we're being told, and to be thinking to myself that the math seems worse than the worst case scenarios in the legacy media.

**************************************

In case you missed it last week, the President of Burundi died. There's is strong suspicion it was due to COVID-19. If so, he would be the first leader of a country to succumb.

They had a cluster at Orlando International Airport, so they tested all 800 employees of the airport. A third of them tested positive. And, if you missed it, Florida fired the person in charge of publishing their data because she repeatedly stated publicly that they were lowballing their numbers. She is now publishing a renegade site.

Brazil has pretty much stopped making an attempt at publishing honest numbers.

The U.S. is re-opening their consulate in Wuhan. This is a good sign.

Meanwhile, China has an outbreak in Beijing. And the BBC has noticed that the scale of lockdown does not seem like a good match for the small number of cases.

Rumors

I was going to write something speculative about this the other day.

The uptick in cases in Iron County led to the lowest days-to-double-cases (7), between June 4th and June 6th. This is 14-16 days after local high school graduation ceremonies. It seems plausible that these events are related. Particularly since days-to-double-cases was at 32 three days after graduation before declining ten of the next eleven days.

*************************************

However, I've now heard the following rumor from more than one source, and from unrelated sources as well.

In generalities: several high school students who participated in a camping party related to the end of the school year later tested positive, and there was some further transmission within their households.

This would also corroborate why the proportion of people wearing masks at Walmart got very about 2-3 weeks ago, and has since slacked off.

I Wrote Too Soon (Regional Deaths Continue to Double Every 3-4 Weeks)

Today, the Southwest Utah Public Health Department listed another death from COVID-19. But this was apparently a reevaluation of an old case; the person had died on May 13th.

Anyway, combined with the deaths last week, this means that our region reached 8 deaths by around June 12th.

All this revision now means that we just doubled our death toll in 23 days, after doubling it in 25 days the time before, and in 28 days the time before that.

Not good.

Washington County Death (Now It's Deaths)

There was another COVID-19 death in Washington County. Southwest Utah Public Health Department is no longer providing dates, but it was reported on the 12th, so I'd guess it was on the 11th.

I posted on May 19th that I was worried that deaths had doubled in the last 25 days. All deaths are a tragedy, but it is a good thing that it took 23 days for the death toll to go from four to five rather four to eight.

*********************************

I wrote too soon.

Today's update shows 2 more deaths, so we're up to 7 now. Both occurred on June 12th.

Also, there's now a case in Beaver County. This may as well be lumped in with Iron County, since Beaver County has zero ICU beds.

Fastest Turning Point Dating, Ever

In the U.S., we measure the length of recessions by declaring certain quarters/months to be turning points: peaks or troughs. A recession is the period between a peak and a trough.

In most of the rest of the world, recessions are said to have begun after the 2nd consecutive quarter of declining real GDP.

Turning points in the U.S. are established by the NBER’s business cycle dating committee.

In classes, I note the personal observation that peaks and troughs are misnamed. Peak evokes a v-shaped change in direction for the economy at its top. My life experience over the last 40 years have been that those peaks are much more rounded, and difficult to spot in real time. By the same token, trough evokes the image of something u-shaped. And yet, in real time, troughs can be fairly easy to feel and notice (unless one gets excessively focused on the unemployment rate — a badly lagging indicator). Having said that, they can also be hard to spot since troughs are sometimes w-shaped rather than v-shaped. In short, it pays to wait.

This is reflected in what is often a rather long delay in declaring turning points in the U.S. Penn macroeconometrician Francis Diebold, writing at No Hesitations shows this in tabular form. The last 5 peaks were declared 6 to 12 months after the fact.

But this time around, the peak was very pronounced. On Monday the NBER declared a peak in the U.S. in February.

I’d narrow that down even further to somewhere between the initial stock market sell off in the third week of February, and the widespread closure of universities in the second week of March.

Of course, this point is mostly academic: a date that is most probable will be important to put in the historical record (I can remember ferreting out the dates of the turning points from library books — pre-internet — while working on my dissertation in 1989). In the real world, we’ve all been fairly certain we’ve been in recession for 3 months now. Other countries will have to wait until late in summer to be “official”.

Evidence for Earlier Chinese Outbreak

From the Director of Boston Children's Hospital's Computational Epidemiology Lab:

Clearly, there was some level of social disruption taking place well before what was previously identified as the start of the novel coronavirus pandemic.

This is about new evidence about what was going on in Wuhan months before the Huanan Seafood Market shutdown.

Images from private satellites show traffic at multiple Wuhan hospitals was up as much as 90% by early October.

This correlates highly with the evidence released about a month ago (from cellphones) that the high security area of the Wuhan Institute of Virology was completely shut down from October 7th through 24th. Over the past month, no evidence has come up to disconfirm this research.

Corroborating this is evidence from Baidu searches showing an uptick in searches for symptoms. This began as early as August.

Tooting my own horn: I posted about the shutdown of the virology institute about three weeks ago, and about unusual internet searches in China way back in February.

Not Sure If This Is a Good Thing or Not

A permit has been issued for the proposed Collin Raye concert at Iron Springs Resort.

I was out that way a few days ago, and it looks like they've been doing some dirtwork and graveling for a big parking lot style concert. Did not see a stage, but those are easy to rent.

I am torn about this. The country is torn between the seemingly polar viewpoints of pandemic safety and the macroeconomy-is-not-to-be-trifled-with. In my position I need to be cognizant of both, and frankly they both suck.

Jeff Swigert has a great quote on his door (I'm paraphrasing here): there are no bad choices, only bad choosing. We're in a situation where all our choices look bad. Just keep in mind that there's still an optimal one.

Everything You Ever Wanted to Know About SARS-CoV-2 But Were Afraid to Ask

Marginal Revolution recommended this very readable summary from Nature entitled “Profile of a Killer: The Complex Biology Powering the Coronavirus Pandemic”, about how SARS-CoV-2 works.

I learned quite a few things. Why SARS-CoV-2 is worse than SARS itself (it’s better at getting a foothold in the nose), why SARS was easier to stop (you aren’t contagious until you show symptoms), why that furin site that people keep talking about is important, why a vaccine ultimately won’t be necessary unless we get one and start vaccinating everyone (it doesn’t mutate quickly, so you should get some immunity once infected), where pangolins and bats fit in (and that this thing has been out in the wild for a long time before it found us).

Interestingly, it also mentions that a pandemic in 1890 that was believed to be caused by the flu, may instead have been the first occurrence of one of those other coronaviruses that now just causes colds.

A Sense of Scale

It's easy to dismiss the small numbers of positive cases in Iron County (and other places).

Here's how I scale it. The U.S has a population of about 340,000,000 people. Iron County has a population of about 34,000. That means we are 1/10,000 of the U.S.

So we had 5 positive today. Scaling up, that would mean 50,000 positives for the country.

But the country did not have that many today ... just 23K.

To me, that suggests that what we are feeling on a "better" day in Iron County is worse than average for the rest of the country.

We just haven't had our 1/10,000th of national deaths yet. That would be about 10 for our little slice of heaven.

The Bad Antibody News

I wrote last week that my entire family had tested negative for antibodies to SARS-CoV-2.

I'd also indicated in class when we were still face-to-face that my wife suspected she had had it in January. Her symptoms were a bad viral infection that had progressed into walking pneumonia. The suspicion I voiced at the time was that, although it would be fortuitous to be done with the disease, the absence of widespread illnesses in February and early March did not seem right for the virus to be in Cedar from January. Suspicion confirmed.

Anyway, we got our test as a result of the study done by Red River Health and Wellness. They released their report this week (although I can't find a downloadable version). But, the news coverage is common in Salt Lake (try here).

One of the things they found was that the virus entered the state through Logan (!!!) in mid-February. This was a month before the first case of community transmission (a bartender in Park City) was detected. Keep in mind that community transmission means either 1) that it can't be traced to someone who traveled from an area with an outbreak, or 2) to someone local known to be infected. And Cache County didn't even have its first positive test results until March 18th, with Iron County following 5 days after that (Washington County is a bit sketchier because they had that cruise ship passenger that returned to there, but they got their second case 3 days after Cache and 2 days before Iron).

The picture this paints is that when Utahns were most worried about the virus in late March ... it was barely beginning to spread.

And now that we are getting more cases, we're less worried.

WTF?

******************************

Iron County had a good day today with just 5 cases. Washington had an average day with 25 new ones.

Figuring Out Nursing Home Deaths

The U.S. is not the only country with a lot of COVID-19 deaths in nursing homes (they are called care homes in many other countries). And yet some countries do not have this problem at all. Why is that?

It turns out that those that modeled their response on SARS did well, those that modeled their response on the flu did not.

The thing is, influenza isn’t that infectious. It’s R-nought isn’t much more than one. For SARS-CoV-2, it’s more like three (without social distancing).

This actually should seem like common sense. Now, isolate you’re thinking to the full-blown flu. Not oh-I-had-a-24-hour-flu thinking, but rather the full-blown, real, I-felt-like-I-wanted-to-die-for-9-days flu. What would have happened in your family when you were a kid if that infection was particularly contagious? Can you imagine all of you getting it within a few days of the first person, and every single person being out of commission for an overlapping period of almost a week? Truth is … that happened to almost no one because the flu isn’t that infectious. SARS-CoV-2 is.

So, in the U.S., the U.K. and some other countries, the it’s-just-like-the-flu policy was to make patients who tested positive for SARS-CoV-2 go back to the nursing/care home, where they infected … everyone … and lots of people died. In contrast, places like Hong Kong quarantined anyone who tested positive, and monitored for 14 days anyone who’d come into contact with them.

An Insightful New Dashboard

MJ forwarded a new dashboard for me. It estimates current cases (tested or not) and projects hospital and ICU loads.

For Iron County, it projects there are currently 112 active cases. That's consistent with what I wrote yesterday: that we can only get super-explosive growth if there is a pool of untested cases that suddenly start showing up for tests.

They also forecast that Iron County will hit its peak ICU load on June 26, and that we will not go over the capacity of the hospital, but that we will get down to zero free ICU beds.

This month, don't get in a car accident, or get appendicitis, or whatever ...

Iron County Update

The news continued to get worse today, with 15 new positive cases in Iron County.

I wrote yesterday that case growth in Iron County was "worse" than "explosive". Technically, "explosive growth" is not something that is clearly defined. It's just a fun thing to say. And when people see something arcing up on a graph, that's what they call it.

More correctly, we talk about geometric growth or compounding (like interest). Here's some examples with different growth rates:

 

What you get with geometric growth is arcs that get steeper. Eventually they call go close to straight up (like the purple one here), but sometimes it takes a while (like the blue one).

But, if I superimpose Iron County's numbers on the same graph, I get this:

What's concerning about that turquoise curve is that was more or less matching the green curve from  a little after 50 days into the outbreak, until almost 70 days into the outbreak. But recall from above that the green curve would eventually look "explosive". What's happened over the last 10 days in Iron County is that our outbreak has started growing faster than geometric. This means there's something new and special and ... not good ... happening: perhaps travelers introduced new cases, or there was a superspreader event.

Don't forget that I speculated that Iron County will be in trouble when we get up to about 75 cases per week. Looking above, we're at 57 in the past week. An unlucky few of those people are going to be hitting the ICU soon.

One of the reasons we take logs of growing series is because they linearize geometric or compounded growth. If I do this to the hypothetical numbers from the top panel, I get lines of different slopes.

Here's what's important and what's bugging me about our local outbreak. When we take logs, we get lines even when the data might be described as "explosive". But if I add our local case numbers, we see this:

That uptick at the end does not seem large, but that's all our current problem in a nutshell. In short, we were on one path from the early 50's to the late 60's (that was a little flatter than the red line), and over the last ten days we've shifted to a new path that's quite a bit steeper than the red line.

This can also be shown by our days to double, which is now down to seven:

Washington County continues to slowly improve. They had 10 new cases today, but their doubling time has been on the rise for about 2 weeks now.

Iron and Washington Counties Update

Iron County: numbers are starting to get scary.

Washington County: still getting worse.

BTW: I went to Walmart in Cedar City this afternoon. I've been counting masks on customers on recent trips, and it got as low as 6 in the whole store a few weeks ago. Today the store was very busy, and I lost track of the number of people wearing masks in the high double-digits.

This first chart shows Iron County. The data on the vertical axis is logged: if cases are "exploding" the data should be an upward sloping straight line on the right hand side. It isn't. It's worse than exploding.

That isn't supposed to happen on a graph like this: a non-technical explanation for taking logs is that it gets rid of that upward curvature that people call explosive. It couldn't correct for that because the growth is faster than explosive; I know that doesn't sound like it makes sense, but those are the best words I have for the current situation.

I'm speculating, but I think it means that there are more sick people out there than we realize, and they are finally getting tested. An epidemic, by itself, shouldn't show growth like this. There must be a human factor involved.

And in Iron County, days to double continues to plummet. It's easy to say "oh, we're just getting back to where we were two months ago", but the thing is back then we were going from 6 to 13 cases in a week. Now we have 7 times that many tests coming back positive each day.

I stand by what I wrote a few days ago: the hospital in Iron County is going to be in trouble when we get up to about 75 cases in a week. We're up to 46 in a week now. We had ten the week before that.

For Washington County, the top chart shows that it never got to flat.

The roughly constant slope on the right hand side indicates steady growth rates. However, since those are rates, and they compound, that means the number of positive cases is continuing to rise and new cases are getting larger every day.

The bottom chart shows that days to double is still inching upwards. This is a good sign.

But, of course, Washington County is still worse off than on May 1 when things started re-opening.

COVID-19 # 66

The Pew Center produced an interesting graphic showing why COVID-19 has a partisan divide (with Democrats tending to think it is worse, and Republicans not so bad). It turns out that most of the hardest hit Congressional districts have Democratic Representatives:

This is an interesting graphic that's becoming more common. Each Congressional District is represented by a square of the same size. These are grouped into shapes roughly approximating the shape and position of states. Because the population of Congressional Districts are close to equal, this data doesn't have to be scaled.

Since most deaths are in the northeast, and the Democrats are strong in the northeast, 41 out of the 44 hardest hit districts are Democratic.

Here's another interesting tidbit: Democratic districts are getting their death rates under control:

Republicans ... not so much.

*************************************

Israel reopened schools a couple of weeks ago. And now they have a superspreader event in a high school, with 103 cases.

Offical figures from Iran indicate they are in a second wave that is nearly as strong as their first wave (which officially peaked around April 1). Dissidents claim that deaths are about 6 times higher than the official numbers.

In other upper level economics classes I'd guess that you talk about "natural experiments" (these are rarer in macro). Well, the U.S. is going through a huge natural experiment right now with all the protests: we're going to find out the hard way how transmissible SARS-CoV-2 is in the open air and sunshine. My guess is ... not very.

Japan looked really bad a month ago. How'd they fix that?

Interesting thread. Japan drove down its cases without large-scale testing/tracing. "Instead, the country focused on finding clusters of infections and attacking the underlying causes, which often proved to be overcrowded gathering spots such as gyms and nightclubs." https://t.co/QGzGFCKnD6

— Tanya Lewis (@tanyalewis314) May 28, 2020

Do you remember a month ago when the models were forecasting 100K U.S. deaths by August. Yeah ... we hit that mark 3 months early ... and people think things are getting better. They are, barely. But mostly what is going on is that people are ignoring the truth. Sorry to have to write that.

China has concluded that the wet market was not the source of the virus.

And in Washington, cases are rising amongst those under 40.

One More Time: It's Not Just Like the Flu

Lots of people say COVID-19 is just like the flu. There's a lot of evidence that it isn't. Here's a new one I came across today.

When people say something like "COVID-19 isn't that bad because the flu kills 60K Americans each year" they're missing a huge point: the 60K is an estimate based on excess deaths. Most of those 60K cases are not confirmed, they're suspected, and there may be comorbidities involved (like being old).

The story is almost the same for COVID-19. Almost. Here we have 100K deaths with comorbidities too. Except that a much higher fraction of COVID-19 deaths are confirmed with tests.

It turns out that if we compare confirmed deaths, then we have information from the worst week of COVID-19 vs. the worst week in each of the last 7 years of the flu. And what do we find? COVID-19 is ten times worse than the worst flu year, and 44 times as bad as the least bad flu year.

So, those naysayers are comparing apples to oranges. A better comparison would be estimated excess deaths this year (due to COVID-19 and the flu) with previous years (which are mostly flu, and already counted as excesses).

This is where it gets ugly. Here's a graph of deaths per week, by country:

The vertical scale is not logged (since these are not growing series), and may differ from panel to panel (reflecting different country sizes). The implication of the linked article above is that the figure of 60K annual flu deaths for the U.S. is calculated by measuring the distance of the gray curve above its minimum, and attributing some fraction of the difference to the flu. Then if COVID-19 were "just like the flu" the height of the pink peaks would be comparable to the distance between the top and bottom of the gray arcs. It isn't ... not by a long shot.

The New York Times has been making excess death data available for a variety of places. It's scary.