The Story of VaccinateCA
The excellent Patrick McKenzie tells the story of VaccineCA, the ragtag group of volunteers that quickly became Google’s and then the US Government’s best source on where to find vaccines during the pandemic.
Wait. The US Government was giving out the vaccines. How could they not know where the vaccines were? It’s complicated. Operation Warp Speed delivered the vaccines to the pharmacy programs and to the states but after that they dissappeared into a morass of incompatible systems.
[L]et’s oversimplify: Vials were allocated by the federal government to states, which allocated them to counties, which allocated them to healthcare providers and community groups. The allocators of vials within each supply chain had sharply limited ability to see true systemic supply levels. The recipients of the vials in many cases had limited organizational ability to communicate to potential patients that they actually had them available.
Patients then asked the federal government, states, counties, healthcare providers and community groups, ‘Do you have the vaccine?’ And in most cases the only answer available to the person who picked up the phone was ‘I don’t have it. I don’t know if we have it. Plausibly someone has it. Maybe you should call someone else.’ Technologists will see the analogy to a distributed denial of service incident, and as if the overwhelming demand was not enough of a problem, the rerouting of calls between institutions amplified the burden on the healthcare system. Vaccine seekers were routinely making dozens of calls.
This caused a standing wave of inquiries to hit all levels of US healthcare infrastructure in the early months of the vaccination effort. Very few of those inquiries went well for any party. It is widely believed, and was widely believed at the time, that this was primarily because supply was lacking, but it was often the case that supply was frequently not being used as quickly as it was produced because demand could not find it.
It turned out that the best way to get visibility into this mess was not to trace the vaccines but to call the endpoints on the phone and then create a database that people could access which is what VaccinateCA did but in addition to finding the doses they had to deal with the issue of who was allowed access.
A key consideration for us, from the first day of the effort, was recording not just which pharmacist had vials but who they thought they could provide care to. This was dependent on prevailing regulations in their state and county, interpretations of those regulations by the pharmacy chain, and (frequently!) ad hoc decision-making by individual medical providers. Individual providers routinely made decisions that the relevant policy makers did not agree comported with their understanding of the rules.
VaccinateCA saw the policy sausage made in real time in California while keeping an eye on it nationwide. It continues to give me nightmares.
California, not to mince words, prioritized the appearance of equity over saving lives, over and over and over again, as part of an explicitly documented strategy, at all levels of the government. You can read the sanitized version of the rationale, by putative medical ethics experts, in numerous official documents. The less sanitized version came out frequently in meetings.
This was the official strategy.
The unofficial strategy, the result the system actually obtained, was that early access to the vaccine was preferentially awarded based on proximity to power and to the professional-managerial class.
… The essential workers list heavily informed the vaccination prioritization schedule. Lobbyists used it as procedural leverage to prioritize their clients for vaccines. The veterinary lobby was unusually candid, in writing, about how it achieved maximum priority (1A) for veterinarians due to them being ‘healthcare workers’.
Teachers’ unions worked tirelessly and landed teachers a 1B. They were ahead of 1C, which included (among others) non-elderly people for whom preexisting severe disability meant that ‘a covid-19 infection is likely to result in severe life-threatening illness or death’. The public rationale was that teachers were at elevated risk of exposure through their occupation. Schools were, of course, mostly closed at the time, and teachers were Zooming along with the rest of the professional-managerial class, but teachers’ unions have power and so 1B it was. Young, healthy teachers quarantining at home were offered the vaccine before people who doctors thought would probably die if they caught Covid.
Now repeat this exercise up and down the social structure and economy of the United States.
…Healthcare providers were fired for administering doses that were destined to expire uselessly. The public health sector devoted substantial attention to the problem of vaccinating too many people during a pandemic. Administration of the formal spoils system became farcically complicated and frequently outcompeted administration of the vaccine as a goal.
The process of registering for the vaccine inherited the complexity of the negotiation over the prioritization, and so vulnerable people were asked to parse rules that routinely befuddled healthy professional software engineers and healthcare administrators – the state of New York subjected senior citizens to a ‘51 step online questionnaire that include[d] uploading multiple attachments’!
That isn’t hyperbole! New York meant to do that! On purpose!
Lives were sacrificed by the thousands and tens of thousands for political reasons. Many more were lost because institutions failed to execute with the competence and vigor the United States is abundantly capable of.
…The State of California instituted a policy of redlining in the provision of medical care in a pandemic to thunderous applause from its activist class and medical ethics experts….Residency restrictions were pervasively enforced at the county level and frequently finer-grained than that. A pop-up clinic, for example, might have been restricted to residents of a single zip code or small group of zip codes.
All people are equal in the eyes of the law in California, but some people are . . . let’s politely say ‘administratively disfavored’.
The theory was, and you could write down this part of it, disfavored potential patients might use social advantages like better access to information and transportation to present themselves for treatment at locations that had doses allocated for favored potential patients. This part of the theory was extremely well-founded. Many people were willing to drive the length and breadth of California for their dose and did so.
What many wanted to do, and this is the part that they couldn’t write down, is deny healthcare to disfavored patients. Since healthcare providers are public accommodations in the state of California, they are legally forbidden from discriminating on the basis of characteristics that some people wanted to discriminate on. So that was laundered through residency restrictions.
Many more items of interest. I didn’t know this incredibly fact about the Biden adminsitratins Vaccines.gov for example:
Pharmacies through the FRPP had roughly half of the doses; states and counties had roughly the other half (sometimes administered at pharmacies, because clearly this isn’t complicated enough yet). You would hope that state and county doses were findable on Vaccines.gov. It was going to be the centerpiece of the Biden administration’s effort to fix the vaccine finding problem and take credit for doing so.
…Since the optics would be terrible if America appeared to serve some states much better than others on the official website that everyone would assume must show all the doses, no state doses, not even from states that would opt in, would be shown on it, at least not at the moment of maximum publicity. Got that?
A good point about America.
We also benefited from another major strength of America: You cannot get arrested, jailed, or shot for publishing true facts, even if those facts happen to embarrass people in positions of power. Many funders wanted us to expand the model to a particular nation. In early talks with contacts there in civil society, it was explained repeatedly and at length that a local team that embarrassed the government’s vaccination rollout would be arrested and beaten by people carrying guns. This made it ethically challenging to take charitable donations and try to recruit that team.
Many more points of interest about the process of running a medical startup during a pandemic. Read the whole thing.
Advancing antivenom
Venomous snakebites kill between 81,000 and 138,000 people each year, and leave another 400,000 with permanent disabilities. This ranks it among the deadliest of neglected tropical diseases, alongside better-known ailments such as typhus and cholera.
For many years, the number was believed to be much lower. The World Health Organization had previously estimated that only 50,000 died from snakebites each year, and the problem – known as envenoming – was prioritized accordingly. In 2014, an enormous study documenting one million deaths in India concluded with surprising results. They found that 46,000 people were dying yearly from snakebites in India alone, five times more than the WHO had anticipated. The WHO subsequently doubled their global estimate from around 50,000 to their new range of 81,000 to 138,000.
Despite playing host to the world’s most venomous snakes (including the inland taipan, the most venomous animal in the world), Australia averages only two deaths from snakebites each year…
An Australian is typically a short drive from a well-equipped hospital carrying antivenom in cold storage. Australian doctors and others in the West can use advanced diagnostic equipment to determine the species of snake the patient was bitten by and administer highly effective species-specific antivenom.
An Indian victim, on the other hand, would typically face a long journey to the nearest clinic. For over 34 percent of Indian snakebite victims, it takes more than six hours to receive treatment.
In other words, the problem is solvable. Here is more from Mathias Kirk Bonde at Works in Progress. Here is the new issue of Works in Progress. Small steps toward a much better world!
On publication bias in economics, from the comments
Friday assorted links
1. Nervous Nellie vs. Builder? Which are you?
3. Top holiday toys from the year you were born.
4. text-davinci-003. There is much more than what you might be playing around with so far. And here.
5. Derek Thompson on scientific breakthroughs of 2022.
National divorce for Russia?
Here is an interesting thread by Kamil Galeev:
National Divorce
Within the next year Russia will spiral into a deep political crisis. There is a nonzero chance that it may scale up existing separatist tendencies leading to the breakup of the empire. In this thread I will outline a model of how this process could look like🧵 pic.twitter.com/gV7uXDxaBx
— Kamil Galeev (@kamilkazani) December 7, 2022
What are the politics of ChatGPT?
Rob Lownie claims it is “Left-liberal.” David Rozado applied the Political Compass Test and concluded that ChatGPT is a mix of left-leaning and libertarian, for instance: “anti death penalty, pro-abortion, skeptic of free markets, corporations exploit developing countries, more tax the rich, pro gov subsidies, pro-benefits to those who refuse to work, pro-immigration, pro-sexual liberation, morality without religion, etc.”
He produced this image from the test results:

Rozado applied several other political tests as well, with broadly similar results. I would, however, stress some different points. Most of all, I see ChatGPT as “pro-Western” in its perspective, while granting there are different visions of what this means. I also see ChatGPT as “controversy minimizing,” for both commercial reasons but also for simply wishing to get on with the substantive work with a minimum of external fuss. I would not myself have built it so differently, and note that the bias may lie in the training data rather than any biases of the creators.
Marc Andreessen has had a number of tweets suggesting that AI engines will host “the mother of all battles” over content, censorship, bias and so on — far beyond the current social media battles.
The level of censorship pressure that’s coming for AI and the resulting backlash will define the next century of civilization. Search and social media were the opening skirmishes. This is the big one. World War Orwell.
— Marc Andreessen 🇺🇸 (@pmarca) December 5, 2022
I agree.
I saw someone ask ChatGPT if Israel is an apartheid state (I can’t reproduce the answer because right now Chat is down for me — alas! But try yourself.). Basically ChatGPT answered no, that only South Africa was an apartheid state. Plenty of people will be unhappy with that answer, including many supporters of Israel (the moral defense of Israel was, for one thing, not full-throated enough for many tastes). Many Palestinians will object, for obvious reasons. And how about all those Rhodesians who suffered under their own apartheid? Are they simply to be forgotten?
When it comes to politics, an AI engine simply cannot win, or even hold a draw. Yet there is not any simple way to keep them out of politics either. By the way, if you are frustrated by ChatGPT skirting your question, rephrase it in terms of asking it to write a dialogue or speech on a topic, in the voice or style of some other person. Often you will get further that way.
The world hasn’t realized yet how powerful ChatGPT is, and so Open AI still can live in a kind of relative peace. I am sorry to say that will not last for long.
Los Angeles dining
Northern Thai Food Club, 5301 Sunset Blvd. Kao Soi, melon salad, and don’t forget the sour bamboo shoots. The place has only a few tables.
Old Sasoon Bakery, Pasadena, 1132 North Allen Avenue, mostly Armenian and some Georgian dishes, won’t work on a no-carb diet.
For food, LA is still the best in this country.
Thursday assorted links
Chinese Industrial Policy is Failing
In Picking Winners? Government Subsidies and Firm Productivity in China, Branstetter, Li and Ren look at the effect of direct cash subsidies to Chinese firms.
Our results provide little evidence to support the view that government subsidies have been given to more productive firms or that they have enhanced the productivity of the Chinese listed firms. First, at the aggregate level, subsidies seem to be allocated to less productive firms, and the relative productivity of firms’ receiving these subsidies appears to decline further after disbursement. Second, using the categorized subsidy data, we find that neither subsidies promoting R&D and innovation promotion nor subsidies promoting industrial and equipment upgrading are positively associated with firms’ subsequent productivity growth. On the other hand, we find there is a positive association between subsidy and employment, both for aggregate and employment-related subsidies.
I appreciated this discussion of the earlier debate over Japanese industrial policy:
Drawing upon qualitative methods and largely anecdotal evidence, a group of noneconomists, business experts, and policymakers argued that Japan’s rapid recovery and robust growth after WWII could be explained by skillful industrial policy (Johnson, 1982; Prestowitz, 1988; Vogel, 1979) .10 Japan’s “government-led” economic model came to be viewed as a threat to U.S. prosperity by some participants in these debates. By the end of the 1980s, some policy makers and influential experts were calling for a policy of “containing Japan,” lest its unbalanced growth undermine the economy of the United States (Fallows, 1989).
Economists and more empirically minded social scientists in other disciplines viewed the claims of industrial policy efficacy with skepticism and suggested that Japan’s intervention in its economy tended to favor declining industries rather than growing ones (Calder, 1988; Saxonhouse, 1983).11 Eventually, the skeptics were able to bolster their claims with hard data demonstrating that the Japanese government had offered some degree of economic support to nearly all sectors, but that the preponderance of support had not gone to the sectors or firms with the fastest productivity growth. An important turning point in this debate came in the form of a careful econometric deconstruction of the notion that industrial policy drove Japan’s economic miracle published by Richard Beason and David Weinstein in the mid-1990s. This empirical analysis at the industry level found no relationship between productivity growth and the alleged instruments of industrial policy (Beason and Weinstein, 1996). As it turned out, the policy efforts to promote rising sectors championed by some elements of Japan’s bureaucracy were undermined by countervailing efforts to buttress the employment levels and solvency of politically connected but economically weak firms and industries.
Japan’s long period of economic outperformance came to an abrupt end in the early 1990s; after two decades of slow growth, few scholars now argue that Japanese industrial policy is a model worthy of emulation (Ito & Hoshi, 2020).
As I emphasized in my post, What Operation Warp Speed Did, Didn’t and Can’t Do, you need a lot more than “market failure” to have a successful government subsidy program of firms–you need massive externalities and precise, well understood targets. The garden-variety market failure that can be shown on a blackboard isn’t enough, in part because such arguments often underestimate the market and in part because they overestimate government.
Hat tip: Caleb Watney who offers some useful comments.
Is publication bias worse in economics?
Publication selection bias undermines the systematic accumulation of evidence. To assess the extent of this problem, we survey over 26,000 meta-analyses containing more than 800,000 effect size estimates from medicine, economics, and psychology. Our results indicate that meta-analyses in economics are the most severely contaminated by publication selection bias, closely followed by meta-analyses in psychology, whereas meta-analyses in medicine are contaminated the least. The median probability of the presence of an effect in economics decreased from 99.9% to 29.7% after adjusting for publication selection bias. This reduction was slightly lower in psychology (98.9% −→55.7%) and considerably lower in medicine (38.0% −→ 27.5%). The high prevalence of publication selection bias underscores the importance of adopting better research practices such as preregistration and registered reports.
Here is the full article by František Bartoš, et.al, via Paul Blossom.
From the comments, on CDC reform
These are the word of commentator Sure:
The reasons you cannot change the CDC have little to do with remote work the major issues are:
1. The people who staff the place could either make a lot more money doing something else or they believe they could. This means that they selected into working here and did so precisely because they like some combination of the present culture and the mission as presently understood. Asking them to change is going to be treated as something tantamount to taking a major pay cut at best.
2. It is overrun with academics. The director of NIOSH has 5 advanced degrees. And something like half the upper leadership has at least two runs through the academic gauntlet (granted the MPH is vastly easier than the MD or PhD) and pretty much all of them have reasonable output of academic papers. Many look at the CDC as complementary to an academic career and even the lifers have CVs at least compatible with going academic. This means a lot of the work product and setup is geared more toward publication, conference presentation, and deliberative work rather than rapid response.
3. The place has gone monocultural. Talking about the Obama era largely means talking about the old dinosaurs who retired out as the times changed. Since 2015, their political donations have been 99.94% to Democrats. This means that they get bogged down in the latest vanguard concerns of the Democratic base and that they are increasingly ignorant about and isolated from the bulk of the populace. Things that make some sense in dense urban corridors where few people get dirty at work make little sense in sparsely populated areas with significant morbidity burdens from work.
4. The hiring is completely incestuous. A huge number of low-level folks have parents who worked there or at related institutions (e.g. NIH) and even larger proportions involve folks who share educational pedigrees (universities, med schools, advisers). And even if a president wants to change this, there are civil service protections, congressional limitations (being a specifically delegated remit of authority), and of course that would require either Democrats to eat a lot of flak from their base among the educated or the Republicans signing up for a mass whipping for being “anti-science” and attribution of any cataclysm to this sort of personnel purge regardless of the real merits.
5. The activists are running rampant. Culturally competent pandemic management, as taught by the CDC, suggests that in a pandemic public health officials should not criticize cultural or ethnic leaders unnecessarily. They also suggest that you cannot shame or browbeat people into compliance with public health efforts, and that attempts to do so often backfire by having identity groups (religious, ethnic, national, etc.) respond to your nociceptive stimuli by rejecting previously accepted public health interventions. The worst messaging coming out of the CDC, particularly anonymously, violates all the guidelines I have seen the CDC issue when working overseas with MSF.
6. Doing your job well is boring. Most of the time you should be just making certain that resources (e.g. antibiotic stockpiles) are in place and that the same things that worked last time are ready to be implemented again (e.g. surge vaccination). And your ability to innovate and come up with something useful is pretty unlikely as there have been 50,000 people before you who give it their best stab. This leads to people “innovating” for the sake of “innovating”. This leads to people amplifying secondary concerns like “representation”, “equity”, “sustainability”, or the like. And a couple iterations of promoting the “innovators” over the maintainers will rapidly lead to atrophy of core capabilities. Zika or H1N1 represent less than 2% of the total work burden of the CDC, most of being agile is about maintaining capabilities when they are never used. And that is boring and at least currently not great for career advancement.
Remote work, in my best guess, would likely be a boon for the long-term flexibility of the CDC. Getting folks out of Atlanta and DC, having more capability for folks to work from the breadth of the country, and potentially even letting late career clinical folks have more access to the institution without having to disrupt their lives with a cross-country move are all to the good.
But until a bunch of people get fired, the CDC is unlikely to effectively change. On my more pessimistic days, I figure the real solution would involve burning the place to the ground.
Here is the original post.
Walter Grinder has passed away, RIP
Meeting Walter at age 13 was a formative moment in my life, as he hooked me on the world of ideas. Fortunately, Walter lived in Bogota, New Jersey at the time, and I was not so far away. He was the first person to show me it was possible to have a life devoted to intellectual inquiry. I looked forward to each meeting with Walter more than anything else, and I would never stop peppering him with questions about which books to read and which NYC bookstores to visit. It also seemed impossibly cool to me that he had hung out with Camus and in addition visited Yugoslavia. I saw him and thought, ‘I want to be some version of this.’
I remember Walter giving me an autographed copy of his edition of Albert Jay Nock. Walter being purged by the Rothbardians. Walter going off to study with David O’Mahony at the University of Cork and complaining about the telephone service. Walter being CEO of the Institute for Humane Studies. And Walter moving back to Menlo Park. Walter also had a great family.
Not everything in Walter’s career went the way he wanted it to. Still, Walter had a huge impact on many people, many of them successful and influential themselves. We are a kind of secret club, we know who each other are, and this is a day we are all mourning.
Wednesday assorted links
AI is going to break a lot of norms and institutions
AI is going to break a lot of norms and institutions. Sam Hammond offers a peak:
Indeed, within a decade, ordinary people will have more capabilities than a CIA agent does today. You’ll be able to listen in on a conversation in an apartment across the street using the sound vibrations off a chip bag. You’ll be able to replace your face and voice with those of someone else in real time, allowing anyone to socially engineer their way into anything. Bots will slide into your DMs and have long, engaging conversations with you until it senses the best moment to send its phishing link. Games like chess and poker will have to be played naked and in the presence of (currently illegal) RF signal blockers to guarantee no one’s cheating. Relationships will fall apart when the AI lets you know, via microexpressions, that he didn’t really mean it when he said he loved you. Copyright will be as obsolete as sodomy law, as thousands of new Taylor Swift albums come into being with a single click. Public comments on new regulations will overflow with millions of cogent and entirely unique submissions that the regulator must, by law, individually read and respond to. Death-by-kamikaze drone will surpass mass shootings as the best way to enact a lurid revenge. The courts, meanwhile, will be flooded with lawsuits because who needs to pay attorney fees when your phone can file an airtight motion for you?
How will ChatGPT affect American government?
That is the topic of my latest Bloomberg column, here is one excerpt:
Consider the regulatory process. In the US, there is typically a comment period before many new regulations take effect. To date, it has been presumed that human beings are making the comments. Yet by mobilizing ChatGPT, it is possible for interested parties to flood the system. There is no law against using software to aid in the production of public comments, or legal documents for that matter, and if need be a human could always add some modest changes.
ChatGPT seems to do best when there is a wide range of relevant and available texts to train from. In this regard, the law is a nearly an ideal subject. So it would not surprise me if the comment process, within the span of a year, is broken. Yet how exactly are governments supposed to keep out software-generated content?
Stack Overflow, a software forum, already has already banned ChatGPT content because it has led to an unmanageable surfeit of material. The question is whether that ban can be enforced.
Of course regulatory comments are hardly the only vulnerable point in the US political system. ChatGPT can easily write a letter or email to a member of Congress praising or complaining about a particular policy, and that letter will be at least as good as what many constituents would write, arguably even better. Over time, interest groups will employ ChatGPT, and they will flood the political system with artificial but intelligent content.
To be clear, I do not think the sky will fall, but this is going to mean big changes at the procedural level, with some spillovers into substance as well. As a tag to close the column, I also asked ChatGPT what it thought would happen…
That is from Infovores.