Month: December 2020

Fallacies about constraints

I am reading many people claim something like “production and distribution of the vaccine is the constraint, not FDA approval.”

There are multiple mistakes in such a view, and here I wish to focus on the logic of constraints rather than debate the FDA issue.

First, there are vaccines available right now, and it helps some people (and their contacts) to have those distributed sooner rather than later.

Second, easing the FDA constraint encourages the suppliers and distributors to hurry to a greater degree.  Just imagine if the FDA were to take a few months longer to approve.  The more general point is that citing “x is right now the main constraint right now” does not mean “the elasticity of x is zero.”

“Sure” wrote in the comments:

On the economics side, I am not convinced that production has ramped up as full and as fast as possible. After all there is some risk premium for expanding plants, running constant shifts, etc. and the danger of delayed approval, particularly if you are in some (mostly negligible) way to the other vaccines may not warrant the investment.

After all, approvals appear to move stocks. Do we really think the market is that dumb? If approval has an impact on market value, why exactly would it not also have an impact on the cost of borrowing, expanding, etc.? Surely somebody believes that approval will result in something different will happen than was happening the day before.

Third, “FDA vaccine approval” is a complementary good for the final vaccine service, strongly complementary in fact.  If the other complementary infrastructure goods have price/quality combinations that are “too disadvantageous,” the theory of the second best implies that approval processes should be speedier and more lax than you otherwise might have thought.  This is just the converse of the classic result that multiple medieval princes imposing multiple tolls on a river create negative externalities for both river users and each other.  Lower those tolls wherever you can.

Fourth, let’s say there were three constraints, each absolutely binding at the current margin.  Speeding FDA approval, taken alone, would have absolutely no effect.  We then ought to be obsessed with identifying and remedying the other two constraints (along with approval)!

But we are not.  Instead we keep on citing those (supposed) constraints in defeatist fashion.  This absence of obsession with easing constraints is in fact one of the biggest reasons for thinking we can do better.  We need to throw more money and talent at these problems, and we are not working hard enough on how to do that.  We are just citing the constraints back and forth to each other and pleading helplessness.

As a final note, I recall that my recently deceased colleague Walter E. Williams was especially good on these issues.  I recall him once saying he wanted to hire a helicopter to drop a cow into the campus central quad, just to show people that supply has positive elasticity.  “I’m going to call them up and say “Williams wants a cow!””


Applying insights from magic to improve deception in research


• Researchers generally receive little training in experimental deception.

• Drawing on the field of magic, we present a novel model of effective deception.

• First, deception should have many “layers” rather than a single cover story.

• Second, these layers should be subtle rather than explicitly stated.

• We provide strategies for improving deception and thus the reliability of research.

And this:

Magicians have theorised that if tricks are too smooth and perfect, they end up seeming less impressive than ones with minor flaws (Kuhn, 2019).

Here is the article, by Jay A. Olson and Amir Raz, via the excellent Kevin Lewis.

Monday assorted links

1. Podcast with Daniel Gross on finding undiscovered talent.

2. The economics of Christmas trees.

3. Are Q-tips OK after all?  I’ve been wondering about this for years, of course.

4. Does the Flaxman pro-lockdown paper make any sense?

5. A grumpy rant about the humanities.: “If we were living in a culture dominated by grown-ups, Martin Scorsese would be considered the purveyor of middle-brow forgettable fare rather than the gold standard of sophistication, and at least the childless among us would not even have to be aware of Spider-Man’s existence.”

6. ORSIFU: Let’s hope this changes, but currently the FAA is not allowing pilots to take the vaccine.

7. More on mobile tricorders.

The Stargate Project

The Stargate Project was a long-running program, funded under various names, by the CIA, Army, and Defense Intelligence Agency to investigate and use psychic powers to defeat enemies of the United States, foreign and domestic. The program can be dated back to the end of World War II but it picked up in the 1970s with rumors that the Russians had a lead in ESP and with the popularity of the “psychic” Uri Geller.

Geller in fact consulted for the program and his powers were investigated under a DIA grant by the Stanford Research Institute. SRI concluded that Geller had “demonstrated his paranormal perceptual ability in a convincing and unambiguous manner.”  The CIA agreed concluding in 1975 that:

“A large body of reliable experimental evidence points to the inescapable conclusion that extrasensory perception does exist as a real phenomenon….the work at SRI, using gifted individuals, has achieved some convincing and striking demonstrations of the existence of paranormal perception, and has demonstrated perhaps less convincingly the possible existence of psychokinetic influences upon sophisticated physical instrumentation.

In fact, as late as 2017 the physicist running the SRI program thinks Geller was “clearly gifted when it came to doing certain psychic tasks.”

Need I tell you that Johnny Carson did a much better job than the DIA of showing Geller was a fraud or that a later investigation suggested that “Geller was allowed to peek through a hole in the laboratory wall separating him from the drawings he was being invited to reproduce.”

Nevertheless, the Stargate Project continued for decades and not just investigations. So-called “remote viewers” were recruited and paid to try to locate hostages, missiles and other locations of military and domestic intelligence secrets:

As he later told the Washington Post, McMoneagle was involved in some 450 missions between 1978 to 1984, including helping the Army locate hostages in Iran and pointing CIA agents to the shortwave radio concealed in the pocket calculator of a suspected KGB agent captured in South Africa.

Another remote viewer, Angela Dellafiora Ford, was asked in 1989 to help track down a former customs agent who had gone on the run, she recounted recently on the CBS News program 48 Hours. She was able to pinpoint the man’s location as “Lowell, Wyoming,” even as U.S. Customs was apprehending him 100 miles west of a Wyoming town called Lovell.

Publicly, the Pentagon continued to deny it was spending money on any kind of psychic research, even as reports leaked out in the 1980s of the details of the government’s experiments. Finally, in 1995, the CIA released a report conducted by the independent American Institutes for Research, which acknowledged the U.S. government’s long-rumored work with remote viewing for military and intelligence purposes.

In other words, don’t believe in star gates.

How many lives will be saved if the FDA had moved faster?

Many people are asking me this question.  I don’t mean to relitigate the question of whether the FDA should be moving faster, rather consider this an exercise in how to think about the trade-offs.  I thus am going to hold the safety and quality of the vaccine constant.

To proceed, consider the distinction between processes defined by economic time and processes defined by calendar time.  In Virginia it may snow in February but not in October, and that is defined by calendar time, not caused by local gdp.  But for many inventory processes, they do not restock until the shelf is emptied by buying customers, and that is economic time.  They don’t check to see if it is June or July.

Let us say that only economic time matters, though I will drop that assumption shortly.

Now, given how late we are in “the season,” it is easier to think about pushing the approval date back rather than moving it forward.  Let’s say that the FDA postponed the December 10 meeting to January 10.  Some number of people would die of Covid during that month — the current clip being around 2600 a day but changing — and then around Jan.10 some kind of vaccine-related health and economic recovery would move into fuller gear.

If only economic time matters, it seems the Dec.10 recovery and the Jan.10 recovery run about the same.  The net difference between the two scenarios is the lives lost in the meantime, to oversimplify say 2000 x 30 days, or 60,000 lives plus accompanying lost jobs and gdp.

I do not think that losing those lives would somehow speed the later, Jan.10-starting recovery process, and it may in some ways render it more fractious.

On top of that, the postponed recovery period likely will imply some kind of grinding uncertainty in the meantime, and possibly intertemporal substitution from some agents (like me!) who are waiting for the change to come before going to the barber. The true net costs are thus higher than what I listed two paragraphs above.

Now, how might the introduction of calendar time alter those estimates?

First, the production of complementary goods for vaccines (say freezers, but the point is more general) may be on a clock of its own, more or less on automatic pilot and requiring time.  When approval comes later, more of those complementary goods are in place, and thus the later recovery is a more powerful one.  That factor tends to lower the cost of delaying approval.

(Of course to the extent those same complementary inputs depend upon economic time, that is reason not to delay approval!  The sooner you approve, the sooner they will get working on getting those freezers in place, which of course boosts recovery power.  Supply is elastic with respect to approval, as suggested for instance by stock market reactions to approval decisions.)

Second, the seasonal effects will differ.  Ideally you want the spread of vaccines to be covering some of the more infectious and thus more difficult winter months.  February is worse than March, and so on.  Given the current clock, this is a big reason to be hurrying.

You might think of other ways calendar time could matter.  You also might think of various non-linear effects and interactions, though I am not sure whether they would make delay more or less costly.

Overall it seems to me that the costs of approval delay are likely very high.  They are not obviously overturned or minimized by citing the relevance of complementary inputs.  The import of complementary inputs might be more ruled by “economic time,” or the seasonal effects may be a stronger quantitative magnitude, again favoring faster speed of approval.

I do understand this is far from a final analysis, rather it is a starting point for conceptualizing the problem.

India vaccine markets in everything

Doses of the coronavirus vaccine developed by AstraZeneca and the University of Oxford could be available for purchase in India as soon as March, according to one manufacturer, in the first sign that the sought-after jab will make its way on to the private market.

Serum Institute of India, the world’s largest vaccine manufacturer, has a licence to produce the shot and has already manufactured 40m doses. Once the job is approved for use, Serum will initially supply the Indian government but then expects to sell 20m-30m doses to private facilities, according to Adar Poonawalla, chief executive.

“Everybody’s asking ‘When can I access the vaccine?’ I’ve told those guys it’s probably going to be March or April,” Mr Poonawalla told the Financial Times.

…the future availability of vaccines for sale privately in other countries, such as India, increases the likelihood of a secondary market developing for vaccines where locals or foreign visitors could pay for a vaccination if not eligible to be inoculated under their own government’s scheme.

Here is the full FT article, via J., the price is expected to be around eight dollars.

Sunday assorted links

Why are women so prominent in vaccine development?

Here is my Bloomberg column arguing that they are prominent in vaccine development, excerpt:

Then there is the vaccine from Novovax, which is based in Gaithersburg, Maryland. The Novovax results are not yet published, but early word is that they are very promising. This vaccine also is based on new ideas, using an unusual moth cell system to crank out proteins in a highly innovative manner.

Novovax’s team is led by Nita Patel, an immigrant from Gujarat, India. Her vaccine team is identified as “all-female.” Patel is from a very poor family; her father almost died of tuberculosis when she was 4 years old, and she often had to beg for bus fare.

Immigrants too, and there is much more evidence at the link.  In fact women have been prominent in vaccine research for a long time.  But why vaccines?  What is the best hypothesis here?

New CRISPR-based COVID-19 test uses smartphone cameras to spot virus RNA

This one brings us closer to the Star Trek medical universe:

Scientists at UC Berkeley and Gladstone Institutes have developed a new CRISPR-based COVID-19 diagnostic test that, with the help of a smartphone camera, can provide a positive or negative result in 15 to 30 minutes. Unlike many other tests that are available, this test also gives an estimate of viral load, or the number of virus particles in a sample, which can help doctors monitor the progression of a COVID-19 infection and estimate how contagious a patient might be.

“Monitoring the course of a patient’s infection could help health care professionals estimate the stage of infection and predict, in real time, how long is likely needed for recovery and how long the individual should quarantine,” said Daniel Fletcher, a professor of bioengineering at Berkeley and one of the leaders of the study…

The new diagnostic test takes advantage of the CRISPR Cas13 protein, which directly binds and cleaves RNA segments. This eliminates the DNA conversion and amplification steps and greatly reduces the time needed to complete the analysis.

“One reason we’re excited about CRISPR-based diagnostics is the potential for quick, accurate results at the point of need,” [Jennifer] Doudna said. “This is especially helpful in places with limited access to testing or when frequent, rapid testing is needed. It could eliminate a lot of the bottlenecks we’ve seen with COVID-19.”

In the test, CRISPR Cas13 proteins are “programmed” to recognize segments of SARS-CoV-2 viral RNA and then combined with a probe that becomes fluorescent when cleaved. When the Cas13 proteins are activated by the viral RNA, they start to cleave the fluorescent probe. With the help of a handheld device, the resulting fluorescence can be measured by the smartphone camera. The rate at which the fluorescence becomes brighter is related to the number of virus particles in the sample.


Now that the CRISPR-based assay has been developed for SARS-CoV-2, it could be modified to detect RNA segments of other viral diseases, like the common cold, influenza or even human immunodeficiency virus. The team is currently working to package the test into a device that could be made available at clinics and other point-of-care settings and that one day could even be used in the home.

“The eventual goal is to have a personal device, like a mobile phone, that is able to detect a range of different viral infections and quickly determine whether you have a common cold or SARS-Cov-2 or influenza,” Fletcher said. “That possibility now exists, and further collaboration between engineers, biologists and clinicians is needed to make that a reality.”

I recall once asking Silvana Konermann: “What am I going to buy at the CRISPR store?”  Well, this is what you are going to buy at the CRISPR store.

Here is the article.  And funded by Fast Grants, I am happy to say.  Quite the week for science, yes?

What is the meta-rational thing to do here?

…the LessWrong community…just released our first book set “A Map that Reflects the Territory: Essays from the LessWrong Community“. It’s a collection of essays from 2018 by Scott Alexander, Eliezer Yudkowsky, and over twenty other writers on LessWrong. It’s a 5-book set, and we actually used quadratic voting to determine what went into the books and what didn’t.

We’re now offering the books on pre-order for $29. It turns out the demand is much higher than I expected, we only planned to print 500 sets but we already sold that many in the first 48 hours.

That is from my email…

Saturday assorted links

Don’t Delay a Vaccine to Allay Fear

I am getting very angry at people like Anthony Fauci who say that FDA delay is necessary or useful to alleviate vaccine hesitancy.

Fauci told Fox News that the FDA “really scrutinises the data very carefully to guarantee to the American public that this is a safe and efficacious vaccine. I think if we did any less, we would add to the already existing hesitancy on the part of many people because … they’re concerned that we went too quickly.”

The WSJ says much the same thing just with a slightly different flavor:

…this regulatory rigmarole is essentially a placebo to reassure the public it will be safe to get inoculated.

The ‘we must delay to allay’ argument is deadly and wrong.

First, we should not let public policy be guided by the most risk averse, fearful, and scientifically illiterate among us. Letting the fearful lead is a recipe for stagnation, mediocrity, and eventual collapse.

Second, there is no guarantee that the risk averse, fearful and scientifically illiterate will be convinced by extra FDA investigation and there is plenty of evidence that they won’t be. Dozens of well-done studies have found no link between vaccines and autism. The scientific evidence that vaccines don’t cause autism is very strong. Yet many people don’t care. Moreover, I bet there is a significant overlap between those who think or fear that vaccines cause autism and those who fear a COVID vaccine. Will a few weeks of extra FDA investigation win these people over? No. More science won’t end science denialism.

Third, rather than alleviating fear, FDA delay may increase fear. People may reason, if the FDA is taking this long to review the evidence when thousands of people are dying every day it must be a hard decision. Delay also makes the vaccine less useful and less obviously useful. Thus, if vaccines come too late people will say that we were reaching herd immunity anyway and that vaccines are useless.

Thus, when thinking about how much investigation the FDA should do before approving a vaccine, allaying fear shouldn’t on the benefit side of the ledger. Greater investigation does have other benefits but I think that the costs of delay exceed the benefits at this time.

The FDA is very unlikely to find reasons not to approve a vaccine later in December. But if that is the case, they should approve now. Note that in August, before the efficacy results were in, I ran the numbers and I said that the case for early vaccination wasn’t strong for most people. Now that we have the efficacy results and months of safety data and a higher death rate, I think quick approval is obviously justified.

How badly has the FDA been lagging?

As a Johns Hopkins scientist who has conducted more than 100 clinical studies and reviewed thousands more from the scientific community at large, I can assure you that the agency’s review can be done within 24 to 48 hours without cutting any corners. They just need to work harder.

Contrary to popular belief, the FDA process is not hands-on—it does not interview vaccine trial patients or look under a microscope at the immune cells. It’s doing a statistical analysis and looking at data. For the vaccine trial, the data set is small and straightforward. If my research team, normally tasked with analyzing data on millions of patients, was asked to review the smaller Pfizer vaccine study of 43,000 patients, it would take about one hour.

The FDA also reviews manufacturing data from Pfizer on how they made the drug. But not only can that data be reviewed in a few hours, it should have been done months ago when it was available. While the FDA was waiting for Pfizer’s long-term vaccine results to come in, the agency should have anticipated this step and done it early.

The final step of the FDA review is to look at the outcomes of the study volunteers, including rates and severity of infection and side effects in the vaccine and placebo groups. Again, there is no plausible reason why this basic analysis cannot be done in 24 hours. The FDA and external scientists have a simple task: confirm or reject  the review already conducted by the trial’s independent data safety monitoring board before FDA submission.

That is from Marty Makary, who also details an ongoing history of FDA delays during the pandemic, starting with the very first Covid testing attempts from the University of Washington, which the FDA tried to nix, but continuing throughout.  And:

FDA insiders say the agency and its approximately 17,000 employees were dark for the four-day Thanksgiving holiday, including those working on the vaccine approval.

For those of us who lived through the 2008 financial crisis (agencies other than the Fed were not on the ball in response), or who have studied (and indeed practiced) the economics of bureaucracy for forty years, or who know the extensive literature on how the FDA operates, will not be surprised by a lack of urgency.  Or from the NYT:

Dr. Fauci said the politicization of the pandemic in his own country had led regulators to move a little more cautiously than the British, to avoid losing public support.

Sorry people, but I read that as “for political reasons we did not go more quickly.”

Here from Statnews journalists Matthew Herper and Nicholas Florko defend the FDA, going into considerable detail, do read it.  Here is one excerpt, in direct contradiction to some of the above:

The agency’s staff “were eating turkey sandwiches on Thanksgiving while reviewing documents,” Peter Marks, who heads the FDA center conducting the vaccine reviews, said on a Thursday webcast run by the Journal of the American Medical Association.

Additionally, members of an FDA advisory committee that will convene Thursday to review the data and issue its recommendations, have expressed no desire to meet sooner. STAT spoke to four members of the panel and all said the agency should not try to move any faster.

My view is this: if your agency is saying “usually we move five to ten times more slowly,” it is highly unlikely their current procedures are optimized for speed.  It is fast organizations that are good at moving fast, right Usain Bolt?

We’re now at the point where Covid-19 is the single leading cause of death in this nation.

Computational sentences to ponder

From a very smart correspondent:

“incidentally forgot to say that this is
basically the death of the
“extended Church-Turing hypothesis”
which roughly states that all reasonable computational models can be simulated on a Turing machine with only polynomial cost.  but it now sure looks like quantum computers are physically realisable and are non-polynomially faster

– in other words we now almost certainly know that computational complexity does depend on laws of physics”

The indentation was his, and with reference to this earlier reported quantum computational result from China.