What should I ask Brian Armstrong?
I will be doing a Conversation with him, in case you do not know Brian is co-founder and CEO at Coinbase.
So what should I ask him? And to be clear, this is the conversation I want to have with him, namely one that maximizes my selfish learning, not your mood affiliation. Here is the Wikipedia page for Coinbase, here is Brian on Twitter, why does a major CEO and person with 410k Twitter followers have no Wikipedia page of his own?
The politics of Covid just got even more hellish
That is the title of my latest Bloomberg column, here is one excerpt:
Preliminary data indicate that the new strain in the U.K. allows the virus to spread from one person to another more easily. The practical upshot is that even the strict lockdowns of early 2020, such as the one just ordered in the U.K. by Prime Minister Boris Johnson, may not be enough to reverse the spread of the virus.
It is far from obvious that politicians will be able to sell voters on strict lockdowns if they still allow the virus to spread. Furthermore, vaccine distribution has been sufficiently slow that a full lockdown would have to last for many months, and that probably isn’t feasible or desirable. Yet not having lockdowns would lead to a much more rapid spread of the virus, overloading hospitals and public health facilities.
And:
The biggest moral dilemmas might come in those countries that to date have been fairly successful at containing the spread of the virus. Apart from restrictions on foreign travel, life in Taiwan has been normal for some time now, and Covid-related casualties have been miniscule. Other successful examples of virus containment can be found throughout Asia and the Pacific.
But how will those countries deal with the new strain? It has already appeared in both Taiwan and China. So far it has not taken over, but the previous tactics of quarantine and tracing may no longer suffice, should the new strain become more active. It is already spreading in Denmark, which did a good job against Covid-19 early on.
Imagine being a leader of a country that has successfully contained Covid, and now realizing that a single mistake could undo almost a year of very hard work. You also know that, precisely because your country has been so effective at fighting the virus, it is not on the verge of vaccinating your entire population. What if you let a single returning citizen pass through customs taking one Covid test rather than three? What if you then cannot control the subsequent spread of the strain that person is carrying?
When was the last time that stakes for such apparently minor decisions were so high? How will leaders deal with the extreme moral anxiety that their decisions will likely induce?
It is like we are living in a horror movie, and just when we think it’s over, the monster comes back, stronger than ever.
Important throughout.
Tuesday assorted links
1. Comment on the electoral fraud result.
2. Sadie Alexander in Teen Vogue.
3. Saudi Arabia lifts blockade of Qatar.
4. Virginia just barely reaching twenty percent distribution.
5. Good Ivermectin results? And more here. Speculative.
6. A modeling exercise favoring First Dose First, the most extensive treatment I have seen to date.
First Doses First? — show your work!
Alex has been arguing for a “First Doses First” policy, and I find his views persuasive (while agreeing that “halfsies” may be better yet, more on that soon). There are a number of numerical attempts to show the superiority of First Doses First, here is one example of a sketched-out argument, I have linked to a few others in recent days, or see this recent model, or here, here is an NYT survey of the broader debate. The simplest numerical case for the policy is that 2 x 0.8 > 0.95, noting that if you think complications overturn that comparison please show us how. (Addendum: here is now one effort by Joshua Gans).
On Twitter I have been asking people to provide comparable back-of-the-envelope calculations against First Doses First. What is remarkable is that I cannot find a single example of a person who has done so. Not one expert, and at this point I feel that if it happens it will come from an intelligent layperson. Nor does the new FDA statement add anything. As a rational Bayesian, I am (so far) inferring that the numerical, expected value case against First Doses First just isn’t that strong.
Show your work people!
One counter argument is that letting “half-vaccinated” people walk around will induce additional virus mutations. Florian Kramer raises this issue, as do a number of others.
Maybe, but again I wish to see your expected value calculations. And in doing these calculations, keep the following points in mind:
a. It is hard to find vaccines where there is a recommendation of “must give the second dose within 21 days” — are there any?
b. The 21-day (or 28-day) interval between doses was chosen to accelerate the completion of the trial, not because it has magical medical properties.
c. Way back when people were thrilled at the idea of Covid vaccines with possible 60% efficacy, few if any painted that scenario as a nightmare of mutations and otherwise giant monster swarms.
d. You get feedback along the way, including from the UK: “If it turns out that immunity wanes quickly with 1 dose, switch policies!” It is easy enough to apply serological testing to a control group to learn along the way. Yes I know this means egg on the face for public health types and the regulators.
e. Under the status quo, with basically p = 1 we have seen two mutations — the English and the South African — from currently unvaccinated populations. Those mutations are here, and they are likely to overwhelm U.S. health care systems within two months. That not only increases the need for a speedy response, it also indicates the chance of regular mutations from the currently “totally unvaccinated” population is really quite high and the results are really quite dire! If you are so worried about hypothetical mutations from the “half vaccinated” we do need a numerical, expected value calculation comparing it to something we already know has happened and may happen yet again. When doing your comparison, the hurdle you will have to clear here is very high.
When you offer your expected value calculation, or when you refuse to, here are a bunch of things you please should not tell me:
f. “There just isn’t any data!” Do read that excellent thread from Robert Wiblin. Similar points hold for “you just can’t calculate this.” A decision to stick with the status quo represents an implicit, non-transparent calculation of sorts, whether you admit it or not.
g. “This would risk public confidence in the vaccine process.” Question-begging, but even if true tell us how many expected lives you are sacrificing to satisfy that end of maintaining public confidence. This same point applies to many other rejoinders. It is fine to cite additional moral values, but then tell us the trade-offs with respect to lives. Note that egalitarianism also favors First Doses First.
h. “We shouldn’t be arguing about this, we should be getting more vaccines out the door!” Yes we should be getting more vaccines out the door, but the more we succeed at that, as likely we will, the more important this dosing issue will become. Please do not try to distract our attention, this one would fail in an undergraduate class in Philosophical Logic.
i. Other fallacies, including “the insiders at the FDA don’t feel comfortable about this.” Maybe so, but then it ought to be easy enough to sketch for us in numerical terms why their reasons are good ones.
j. All other fallacies and moral failings. The most evasive of those might be: “This is all the more reason why we need to protect everyone now.” Well, yes, but still show your work and base your calculations on the level of protection you can plausibly expect, not on the level of protection you are wishing for.
At the risk of venturing into psychoanalysis, it is hard for me to avoid the feeling that a lot of public health experts are very risk-averse and they are used to hiding behind RCT results to minimize the chance of blame. They fear committing sins of commission more than committing sins of omission because of their training, they are fairly conformist, they are used to holding entrenched positions of authority, and subconsciously they identify their status and protected positions with good public health outcomes (a correlation usually but not always true), and so they have self-deceived into pursuing their status and security rather than the actual outcomes. Doing a back of the envelope calculation to support their recommendation against First Doses First would expose that cognitive dissonance and thus it is an uncomfortable activity they shy away from. Instead, they prefer to dip their toes into the water by citing “a single argument” and running away from a full comparison.
It is downright bizarre to me — and yes scandalous — that a significant percentage of public health experts are not working day and night to produce and circulate such numerical expected value estimates, no matter which side of the debate they may be on.
How many times have I read Twitter threads where public health experts, at around tweet #11, make the cliched call for transparency in decision-making? If you wish to argue against First Doses First, now it is time to actually provide such transparency. Show your work people, we will gladly listen and change our minds if your arguments are good ones.
Does soil heterogeneity induce greater individualism?
Itzchak Tzachi Raz says maybe so:
This paper studies the impact of social learning on the formation of close-knit communities. It provides empirical support to the hypothesis, put forth by the historian Fred Shannon in 1945, that local soil heterogeneity limited the ability of American farmers to learn from the experience of their neighbors, and that this contributed to their “traditional individualism.” Consistent with this hypothesis, I establish that historically, U.S. counties with a higher degree of soil heterogeneity displayed weaker communal ties. I provide causal evidence on the formation of this pattern in a Difference-in-Differences framework, documenting a reduction in the strength of farmers’ communal ties following migration to a soil-heterogeneous county, relative to farmers that moved to a soil-homogeneous county. Using the same design, I also show that soil heterogeneity did not affect the social ties of non-farmers. The impact of soil heterogeneity is long-lasting, still affecting culture today. These findings suggest that, while understudied, social learning is an important determinant of culture.
Here is the full paper. See also his paper on homesteading: “…we find that areas with greater historical exposure to homesteading are poorer and more rural today.”
Monday assorted links
1. Did irrigation entrench the patriarchy? By Alice Evans.
2. Casey Mulligan measures deaths of despair.
3. “We create a novel reign-level dataset for European monarchs, covering all major European states between the 10th and 18th centuries. We first document a strong positive relationship between rulers’ intellectual capabilities and state-level outcomes…We also show that rulers mattered only where their power was largely unconstrained. In reigns where parliaments checked the power of monarchs, ruler ability no longer affected their state’s performance.” Link here. And Ian Bremmer’s Eurasia Group on the top risks for 2021.
4. The Military Health System: “We find evidence that off-base care is associated with slightly greater resource intensity, but also notably better outcomes, suggesting marginal efficiency gains from care privatization.” That’s from Jon Gruber and co-authors, not the Heritage Foundation.
How will we interpret data on the new strain?
Instead of exploding relative to a baseline of 0 cases (like in March), the new strain will be exploding relative to a baseline of around 200,000 cases per day. As a result, day-to-day random noise will completely mask any increase in infections from the new strain until it becomes dominant — around day 40 in the above chart. This means that if people use the overall numbers to guide their levels of precaution, our reaction time could lag by two or three weeks compared to March — as if we had locked down in early April instead of mid-March.
Now, this isn’t entirely a correct comparison because the rate of exponential growth will be much below 1.36; a reasonable guess might be 1.12. (You can get this by assuming R = 1.6 and assuming a generation time of 4 days.) But the overall point is the same: because any increase in the prevalence of the new strain will be masked by noise in current COVID levels, the new strain won’t be evident in overall numbers until it starts contributing hundreds of thousands of daily infections.
That is from Eric Neyman, good points thought note we will have independent measures of the spread of the new strain, with shorter lags than is currently the case.
Globalization is older than you think
Asian spices such as turmeric and fruits like the banana had already reached the Mediterranean more than 3000 years ago, much earlier than previously thought. A team of researchers has shown that even in the Bronze Age, long-distance trade in food was already connecting distant societies…
Working with an international team to analyze food residues in tooth tartar, the LMU archaeologist has found evidence that people in the Levant were already eating turmeric, bananas and even soy in the Bronze and Early Iron Ages. “Exotic spices, fruits and oils from Asia had thus reached the Mediterranean several centuries, in some cases even millennia, earlier than had been previously thought,” says Stockhammer. “This is the earliest direct evidence to date of turmeric, banana and soy outside of South and East Asia.” It is also direct evidence that as early as the second millennium BCE there was already a flourishing long-distance trade in exotic fruits, spices and oils, which is believed to have connected South Asia and the Levant via Mesopotamia or Egypt.
Here is the full account, I strongly suspect globalization is much older than is commonly believed. Via Bruno M.
Sunday assorted links
1. Scott Sumner movie reviews.
2. Eli Dourado on the future course of innovation, recommended.
3. Further claims about Oumuamua. And early NBA results not correlated with “how reality ought to be.”
4. Even in hard-hit Manaus, hospitalizations are reaching a new record high. And more (speculative) worries about the South African strain.
5. Quantitative analysis of Darwin’s reading choices.
6. A blind date.
Sudan estimate of the day
…researchers have just declared that there was a huge, hidden outbreak in the capital of Sudan. In the absence of a good death registration system, they used a molecular and serological survey and an online one distributed on Facebook, where people reported their symptoms and whether they’d had a test. The researchers calculated that Covid-19 killed 16,000 more people than the 477 deaths confirmed by mid-November in Khartoum, which has a population roughly the size of Wisconsin’s.
Here is the full NYT article by Ruth Maclean. The main theme of the piece is that Africa may not have escaped Covid by nearly as much as we had thought.
Virginia fact of the day
Total vaccine doses distributed: 388,100
Total vaccine doses administered: 75,288
Here is the link (which at some point will update), that is below 20 percent.
We need to do much better than this. As a nation, Israel is doing about 10x better than the United States. You might think for some intrinsic reasons Israel could do 2x or 3x better, but 10x? It is time to get our act together.
And here is (very poor) performance for the various German Bundesländer (in German).
Saturday assorted links
1. “West Coast wellness elites think kambo, an Amazonian frog poison drug, is helping them purge “toxins” from their lives.” What kind of sentence is that? (NYT)
2. Congress will apply banking regulations to the antiquities market (NYT).
3. Scott Aronson on what we could have done and should have done.
4. Arnold Kling on functions vs. project management. And more from Arnold.
5. African Continental Free Trade Area starts.
6. Antisemitism and Financial Markets in the Time of the Dreyfus Affair.
7. It seems the new (non-mink) variant is now spreading rapidly in Denmark.
8. Don’t really get how Cato can rank Hong Kong as #3 for freedom.
Amazon search is failing us
Maybe it is different for you, but when I search Amazon for “D.M. Knight, Science and Spirituality”, D.M. being commonly used in citations to the book, I get this mess — no listing of the book!
If I pop the same into Google, the Amazon listing for the book comes up third.
So more and more I am using Google to search on Amazon (heaven forbid that your Amazon author’s name is “Glass,” you will be offered all sorts of glassworks for sale and if you are lucky the book listing is on p.3).
But what can I use for doing a Google search? Surely not Google…Daniel Gross must know!
The *variation* of quality within a given state government
We are seeing our state governments doing a poor job — yes a very poor job –distributing the vaccine. You can take this as evidence for various theories of bureaucratic dysfunctionality and it is. But still at the end of the day, always ask about the cross-sectional variation!
Virginia runs prisons, schools, maintains roads, has a Medicaid program, and various state-level functions, such as hiring staff for the governor, some of those in conjunction with other levels of government. Maybe those services are not productivity marvels, but they work OK — I’ve lived here for a long time. So why the differences? Here are a few hypotheses, not all of which need be true:
1. Learning curves are steep. Most of what governments do is just terrible at the beginning, but eventually there is learning and improvement. What is different here is simply the hurry.
2. Interest groups make everything run. It is clear who benefits from state-level Medicaid programs, and those constituencies keep the programs on track. In contrast, the beneficiaries from rapid Covid vaccination are quite diffuse and are not represented by strong, exclusive organizations.
3. Too many layers of government (and society) are involved. The states are waiting for the local public health authorities, who are waiting for the counties, who are waiting for the Feds, and so on. The private sector is involved too, through CVS and the like. No one is picking up the ball and running with it. No one was told who moves first. In contrast, the lines of responsbility for running roads, schools, and the like are fairly clear.
4. The real problem is the citizenry. The lines to get these vaccines for the 1A group are not long. Government made one mistake of assuming the first round of take-up would be rapid, but the real problem is the sluggishness of the demanders. And things will be OK once we get past the 1A group and open up distribution more broadly.
5. Logistics mentality is lacking. Our state governments have specialized in Medicaid, while contracting our schools to the localities and road construction and repair to the private sector. There is perhaps not a strong enough core of logistic expertise and logistics culture in most state governments.
What else? And what are the relative weights on the truth of these hypotheses? To what extent can we use these and other hypotheses to explain cross-sectional variation across the states? Why are West Virginia and the Dakotas doing relatively well in vaccine distribution so far, when those are not typically considered the most effective state governments?
Again, always ask about the cross-sectional variation!
2021 assorted links
1.Why a longer dosing interval should be fine. And the UK case for first dose prioritisation.
2. “We find that [Chinese] police stations are more likely to be located within walking distance of foreign religious sites (churches) than other sites (temples), even after controlling for the estimated population within 1km of each site and a set of key site attributes.” Link here.
3. Some UK doctors will defy instructions on postponing the second shot (NYT).