Relevant and even prescient commentary on news, politics and the economy.

The Stanford / Santa Clara county study

A bit of a post-mortem on my last post.  Andrew Gelman discusses the Santa Clara county study here.  He focuses on sample imbalances, selection bias, and especially the way the authors deal with the specificity of the test.  See his post for the gory details, but he is quite critical of the study, to the point of asking for an apology from the authors for wasting our time:

I think the authors of the above-linked paper owe us all an apology. We wasted time and effort discussing this paper whose main selling point was some numbers that were essentially the product of a statistical error.

I’m serious about the apology. Everyone makes mistakes. I don’t think they authors need to apologize just because they screwed up. I think they need to apologize because these were avoidable screw-ups. They’re the kind of screw-ups that happen if you want to leap out with an exciting finding and you don’t look too carefully at what you might have done wrong.

How should we update our priors on COVID-10?

I want to think about how two pieces of news should change my thinking about COVID-19.  (Warning:  I have no expertise in medicine or public health, and you have no reason to take my thoughts seriously – but you knew that already.)

A new serological study in Santa Clara county (discussed by Kevin Drum here) suggests that far more people have been infected with COVID-19 than researchers had previously believed.  This is only one study and full of uncertainties, but it suggests that the infection fatality rate of COVID-19 (the chance that you die if you get the virus) may be much lower than was previously believed.  The authors suggest an IFR of .12% to .2%.  The Imperial College study, in comparison, used a rate of .9%.

It is tempting to say that a lower infection fatality rate is a good thing:  COVID-19 is less deadly than we thought.  I’m not sure this is right.  The fact that many more people have had the disease than we previously believed will not only affect our estimate of the infection fatality rate, it will also lead us to revise our estimates of other critical parameters.  For example, it suggests that COVID-19 is much more contagious than we previously thought and that the rate of asymptomatic infections is higher than we believed.

Higher rates of contagion and asymptomatic infection are both troubling.  Think about it this way.  Your chance of dying from COVID-19 depends on the probability that you die if you get infected times the probability that you get infected.  The new study suggests that the first number is lower than we thought, but the second number is higher.  I don’t know enough about the dynamics of the SIR model to know if this is a good tradeoff, but it’s not clear that it is:  Ebola had a very high fatality rate but a very low transmission rate, and we would clearly be better off with an easily contained Ebola epidemic than with COVID-19.  Higher asymptomatic transmission will make it much more difficult to control the spread of the disease without prolonged social distancing or massive testing.

Second, a piece by Matt Stieb in New York Magazine reminds us that death is not the only way COVID-19 harms us.  Serious long-term health impacts from COVID-19 infection – kidney and liver disease, heart and lung disease – may turn out to be relatively common among those who develop serious illness.  In addition to the life long suffering this will cause, it is conceivable that the long-term loss of life-years among people who survive initial infection with COVID-19 will exceed the loss of life-years due to immediate fatalities.  There may also turn out to be negative health consequences for people with less severe illness.  And the immediate suffering caused by the disease is often severe as well.

I don’t have a lot of confidence in this analysis, but overall, I’d say the news of the day raises my estimate of the likely health impact of the epidemic.  Of course, your mileage could easily vary.  What this means for policy is a separate question.

Epidemiologists, government failure, and COVID-19

Jason Brennan has a new post up doubling down on his earlier criticism of epidemiologists and government policy in response to the COVID crisis.  I responded to his earlier blog posts here.  I am still not convinced, but there are useful lessons to be learned from going through his argument.

Brennan continues to claim that epidemiologists produced bad statistical analysis, and that we should not take their advice seriously (all bolding is mine):

I’ve been criticizing epidemiologists–including the ones publishing in JAMA, the Lancet, NEJM, etc., and the famous ones who were making apocalyptic predictions on TV last month–for doing what is clearly bad work. My main complaint is, again, that their estimates about the danger of the virus are based on the wrong data (current infections) collected the wrong way (non-random testing of people who present themselves as sick). We all know better than that. You don’t sample on the dependent variable. You don’t sample in ways that suffer from severe selection bias. If you mostly test people who show up saying they are sick, and 3.4% of them die, it doesn’t tell you how many people have the infection, nor does it tell you what percent of people who have the infection will die.
Now, while many economists and others trained in stats have been saying the same thing, it’s surprising how many untrained people say we should instead defer to epidemiologists. You can see some of their arguments on Facebook and others in the comments to previous posts.

He does not repeat the suggestion in his earlier post that decisions based on analysis from epidemiologists “presumptively lack authority and legitimacy”, but he doesn’t withdraw it either.

Here is one way to think about the claims Brennan is making.  Suppose the following statements are true:

  • Epidemiologists overestimated the risk of COVID-19 by applying bad statistical methods to poor data
  • Politicians adopted costly lockdown policies based on flawed risk estimates produced by epidemiologists
  • Lockdown policies were unjustified at the time they were adopted, and this could have been known if better statistical analysis had been undertaken
  • Lockdown policies are unjustified now, given currently available information, and should be replaced by a different policy

What does the COVID-19 epidemic teach us about the role of government?

What lessons should we draw about the role of government from the COVID-19 epidemic?  I want to address this question in a few posts.  I’m going to start by examining a blog post by the libertarian philosopher Jason Brennan.  Brennan makes the following claims:

  1. Many medical journals published misleading papers based on bad statistical analysis and incomplete data; in particular the case fatality rate may have been overestimated because the total number of people infected was not (and still is not) known
  2. Many of the models used by researchers are “quite poor”
  3. Intellectual standards are low in medical research, the flawed papers accepted by medical journals would have been immediately rejected at economics journals
  4. Government has acted on this bad research and closed down the economy, inflicting severe harm on many people for unclear gains
  5. Government is responsible for the fact that better data was not available, because government failed to increase testing capacity and failed to undertake the types of studies that would have enabled researchers to estimate the key parameters needed for rational policymaking (such as the case fatality rate)
  6. Government decisions that are not made competently and in good faith are “presumed to lack authority (there is no obligation to obey it) and legitimacy (there is no moral permission to enforce it).”
  7. Because governments have acted on bad data and bad models in implementing the shutdown, government actions lack authority and legitimacy.
  8. The fact that government had to act quickly on the basis of bad information is irrelevant, or at least not dispostive.

I want to respond to these claims, especially his normative claims (6, 7, and 8), but I start with his criticism of medical researchers.

First, I believe it has been well known since early in the epidemic that crude case fatality rates are biased up because we do not know the true prevalence of the disease due to inadequate testing.  It was also known that crude case fatality rates early in the epidemic are biased down because there are people who are currently infected but have yet not died.  I believe that many researchers tried to deal with these statistical issues as well as many, many others.  I believe I have seen papers that published a range of estimates for the number of deaths based on different assumptions about these biases, although I am not going to go back and look for them.  Of course, I am sure that some bad papers were published.  This happens even when researchers are not facing a public health emergency.  But Brennan does not show that most papers published were bad.  In fact, he does not cite a single paper that used naive statistical methods.

Second, his claim that many of the models used are “quite poor” is unsubstantiated.  Again, I am sure that people made debatable modeling choices, and some people no doubt made errors in their analysis or misleading claims about their results.  Brennan does not provide any evidence that this was common, or that government policy depended on this.

Brennan apparently believes that intellectual standards are low in medical journals, at least relative to economics journals.  Again, he offers no evidence.  Of course, no doubt there are problems in some fields of medical research and epidemiology (nutrition studies are arguably pretty close to worthless, for example).  Many medical researchers understand this and are trying to correct it.  And it is also true that there are problems with empirical work in economics.  See Gelman here; many, many other citations could be given.  I note that the IHME model, which has been criticized for its methods and seems to be highly inaccurate in its predictions, was apparently developed by economists.  The only specific methodological flaw that Brennan points to is a failure to account for heterogeneity in R0.  But economists estimate models without heterogeneity in R0.  See here.  The use of simplifying assumptions and imperfect statistical tools is inevitable, especially under the severe time constrains imposed by a looming epidemic.  Empirical work is hard, researchers aren’t super human.

As an economist, I would love to believe that economists are geniuses with high intellectual standards, perfect models, and excellent statistical techniques.  I think we do have our professional strengths.  Nonetheless, Brennan’s attack on medical/epidemiological research and glorification of economics seems to me to be unwarranted and unhelpful.

Moving on . . .

Brennan seems to believe that bad research exaggerating the danger from COVID-19 led governments to overreact and implement excessively restrictive policies.  Again, he presents no evidence for this.  Many people believe that governments under-reacted, at least in the sense that they took too long to respond to the epidemic, and that this allowed the disease to spread to the point that standard test/trace/isolate methods were ineffective.  

What about Brennan’s moral claims?  He acknowledges that “in principle, governments can restrict our freedom to stop the spread of disease.”  But then he adds that “appealing to abstract principles is not enough to justify their actions. We need to know whether they made these particular decisions competently and in good faith, on the basis of good information.”

Here is an alternative view (one that, incidentally, I believe most economists would subscribe to):  

The Democratic COVID-19 Testing Strategy

Senator Patty Murray has released a Democratic proposal to increase our capacity to test for the coronavirus.  The report quite rightly emphasizes that expanded testing is critical for containing the epidemic and re-opening the economy.  It criticizes the Trump administration for it’s failure to recognize the need for extensive testing:

Democrats demand a clear, detailed plan to rapidly scale and optimize COVID-19 testing in the United  States. This plan must take a “whole of society” approach that, led by the federal government, will quickly scale and optimize COVID-19 testing, mitigate additional community spread, and help get America back to work.

The proposal is summarized as follows:

Testing is one of the most critical elements to beating this virus. Congress should designate $30 billion in new  emergency funding to support the rapid expansion and optimization of testing for COVID-19 in the  United States. While the funds would be directed to a broad range of testing and related activities, including critical surveillance and contact tracing, a significant portion of this funding should be dedicated to:

  • Optimize the testing supply chain at a national level, addressing manufacturing shortages of testing kits and  supplies through increased production, including through matching excess supplies or personnel with demand in other  locations, centralized procurement of testing supplies, guaranteed contracts to manufacturers, and effective use of  the Defense Production Act, when necessary; and
  • Develop and validate different types of tests necessary, including rapid, molecular point-of-care and at-home  testing platforms, to effectively monitor and control COVID-19.

This is an important step.  My comments:

The plan is not nearly ambitious enough.  If the epidemic continues for months, the health, social and economic toll will be devastating.  The immediate economic cost will run into the tens of trillions of dollars, and the economic devastation may last for years.  The political consequences are unpredictable but could be severe; there is a small but real risk that our democracy will not survive.  Given the stakes, we should do everything we can to increase our ability to test, so that we can identify people with the virus and let everyone else go back to work, school, and life.

No one knows how many tests we need to do to prevent an ongoing disaster, or how fast we can ramp up testing capacity.  But the more money we throw at the problem, the better we will do.

The amount of money in the bill is clearly inadequate.  Suppose we want to be able to test 10 million people a day for 90 days.  If the fully loaded cost of a test including the cost of new testing equipment, test kits, protective gear, and labor is $100, then 10 million tests a day for 90 days will cost $90 billion, three times the amount contemplated in the Democratic plan.  Given the costs of an ongoing epidemic, why hold back?

There is great uncertainty about the costs of testing.  Siddarth and Weyl suggest the cost could be $15 due to falling costs.  However, demand is being restrained by the current restrictive testing guidelines and labs are currently struggling with test kits and swabs.  I suspect that the main bottleneck to testing will soon be the PCR machines that identify viral RNA, and it seems likely that costs will rise, not fall, once current capacity limits are reached.

To get manufacturers to scale up production of PCR machines, we need to give them firm contractual commitments NOW.  These commitments can take the form of cost-plus contracts or incentive contracts.  What is critical is that we give manufacturers reasonable assurance that their investments in capacity will not be lost if the epidemic wanes, and that the prices they ultimately get paid will be sufficient to compensate them for all the risks that they will face, including the risk that the government decides to play hardball with them after they have invested million of dollars increasing their productive capacity.  Giving companies contracts is the way to avoid this risk.

I can’t emphasize enough the need for speed in giving out contracts.  Expanding production capacity could prove to be very difficult.  It took over 7 months to build the first liberty ships during the World War II mobilization.  That timeline would push mass testing back to late fall.  We should have started this process at least a month ago.

News flash: Libertarian invents new reason not to help people!

Thinking about the coronavirus is bleak, so let’s do some political philosophy to cheer ourselves up.

Libertarian philosopher Jason Brennan has a new post up claiming that our obligations to help strangers are much weaker than we might think they are, and may not exist at all, because most people are “morally very bad”.

Brennan begins with this question:

To what degree are our moral obligations to provide help and assistance to strangers reduced because those strangers are likely to be morally bad people?

He answers this question in 3 steps.  First, he argues that you do not have any reason to keep Bob, a rapist or child molester, from starving, even if you can easily afford to help.  Second, he argues that even if Bob has not actually done anything wrong, if Bob has a bad character and would do something very bad if he could get away with it, then you don’t have any reason to help Bob:

Bob is an evil person, though he has not yet done anything evil. We owe evil people less. You are free to, say, spend the money on toys for yourself rather than keep him alive through your charity. Indeed, you probably are obligated to refrain from helping keep Bob alive.

Finally, Brennan argues that most people are in fact morally very bad:

. . . Most people are disposed to be utter conformists and to obey evil authority. Nearly all our neighbors are disposed to be obedient concentration camp guards; they only reason they haven’t done that is because, thanks to moral luck, they haven’t been in such a situation. Further, most of people’s apparently altruistic behavior is in fact motivated by self-interest . . . People in general have quite bad moral character, but most of them haven’t done anything particularly bad because they haven’t had the opportunity . . .
This seems like a good reason to discount our estimates of what kinds of assistance we owe them.

The question I want to ask is whether this argument gives us a (strong) reason to oppose social insurance and progressive taxation, the main institutions that capitalist democracies use to assist those who find themselves in difficult circumstances.  (If Brennan is merely claiming that our obligation to help others on an individual basis is not as strong as we tend to think, I also disagree, but this is a less interesting and less important claim.)

Brennan assumes that people have fixed, immutable moral characters, and that their characters are either good or bad.  To use his terminology, people are either “saints” or “scoundrels”.  Scoundrels will act terribly when circumstances permit; saints always act decently towards others.  Brennan assumes that most people (“nearly all”) are scoundrels, and that the obligation of saints to help scoundrels is limited.  Note that Brennan is viewing morality entirely from the point of view of saints, even though he claims that most of us are scoundrels.

We could certainly push back on Brennan’s model and his empirical claims, but let’s run with them and see what happens.

Imagine a group of scoundrels who live together in a representative democracy with a market-based economy.  Even though everyone is a scoundrel, the society functions reasonably well because informal norms, social pressure, and the threat of civil and criminal sanctions all encourage people to behave decently towards one another.  Most people are vaguely aware of the fact that they might commit atrocities under less favorable circumstances, when they bother to reflect on such things, but they nonetheless come to think of themselves as imperfect but morally decent people.  This kind of grandiose moral self-delusion is, sadly, all too typical of scoundrels.

Can the House Democrats drive a hard bargain on CARES 2?

Michael Grunwald argues in Politico that House Democrats have a lot of bargaining power in negotiations over the next coronavirus relief bill, but that they are not aggressively using their leverage.  He suggests that Democrats are holding back because they are worried about being labeled obstructionist and getting blamed if legislation does not pass.

I agree with some of his analysis and have doubts about other parts.  But here I want to make a suggestion for getting an aggressive but narrowly-tailored bill through Congress.

Suppose that House Democrats quickly put together a new Democratic CARES 2 bill that aggressively attacks the epidemic.  The CARES 2 bill should include more aid for small business, funding for state and local governments, hospitals, transit agencies, the Post Office, etc.  It should extend the unemployment insurance provisions until the unemployment rate falls below, say, 5%.  It should impose stricter oversight on the Trump administration.  It should repeal the tax cut for real estate developers that was included in the first CARES act.  It should include tens of billions in funding to increase testing capacity and to make masks widely available.  It should include reforms to ensure that the 2020 election is fair, such as national vote-by-mail.  The bill can and should be aggressive, but it must be narrowly-tailored to fighting the epidemic.  No Christmas-tree stuff that makes it look like the Democrats are exploiting the crisis for narrow partisan purposes.

Because this would be a Democratic bill, it cannot pass the House by unanimous consent.  So suppose House Democrats come back to Washington and pass the bill.  They would be risking their lives to do right by the American people.  Seeing the older – and even elderly – House Democratic leadership wearing face masks as they take an emergency vote would be inspiring to many Americans.  And then the House Democrats should go home to be safe.  It would be up to the Senate to take or leave the Democrats offer.  If Republican Senators whine about this or that provision, the Democrats can say there will be time to fix it later, but that right now what is needed is for the Senate to come back into session and pass the House bill.  The Democrats’ courage in returning to Washington would make it difficult for Republicans to blame them for obstruction, and, as Grunwald says, Trump and the Republicans need a bill just as much as Democrats do.

Where-oh-where would we be without Trump’s firm leadership in this crisis?

Kevin Drum reminds us that our fearless leader restricted air travel from China after the airlines had already done so.

And today’s New York Times:

New research indicates that the coronavirus began to circulate in the New York area by mid-February, weeks before the first confirmed case, and that travelers brought in the virus mainly from Europe, not Asia.

“The majority is clearly European,” said Harm van Bakel, a geneticist at Icahn School of Medicine at Mount Sinai, who co-wrote a study awaiting peer review.

A separate team at N.Y.U. Grossman School of Medicine came to strikingly similar conclusions, despite studying a different group of cases. Both teams analyzed genomes from coronaviruses taken from New Yorkers starting in mid-March.

The research revealed a previously hidden spread of the virus that might have been detected if aggressive testing programs had been put in place.

Trump’s blame-avoidance is politically shrewd

Trump’s handling of the Covid-19 pandemic is predictably chaotic, vengeful, irresponsible, and impulsive.  His actions have worsened the epidemic, they have led to unnecessary deaths and to a very painful economic lockdown.  Coming in the year before he is up for re-election, this seems self-defeating:  if Trump could re-run history I have little doubt he would take aggressive action to nip the epidemic in the bud.  That said, the political strategy that is emerging from his chaotic approach to emergency management is actually quite shrewd.

As President, Trump will naturally get credit if things go well.  If the epidemic ends quickly and the economy roars back to life he can take a victory lap even if his actions make this outcome less likely.  Trump understands this:  what he is worried about is being blamed if things go badly.

It might seem that the need to avoid blame would give him an incentive to competently manage the crisis – to coordinate the distribution of ventilators and personal protective equipment, for example, but he is not even trying to do this.  (Exhibit A:  putting Jared in charge.)  The reason is that he understands that any effort to actively manage the epidemic will make him the target of criticism for problems even if his actions actually make problems less likely or severe.  If he takes responsibility for distributing masks or ventilators, then every time a shipment gets lost in an airport, every time a patient dies for lack of a ventilator or a health care worker gets sick with Covid-19, critics will ask why he didn’t do more to prevent it.  His strategy is to simply deny that he is in charge, and to place as much blame as he can on governors and states.  Or at least to lay the groundwork for doing this if things go south.

When it comes to managing the economic damage his strategy is to advocate – irresponsibly – for an early end to the social distancing and a quick resumption of normal economic activity.  My prediction is that he won’t actually push very hard for this, because if there is an early end to social distancing and it goes badly he would be on the hook for it.  So he will be happy leaving this decision to governors.  He is setting himself up to blame others if the economic slowdown continues and becomes a political liability.
There is no guarantee that his blame-avoidance strategies will work, but Trump is playing a bad hand skillfully.

An interesting question is whether Trump is pursuing this blame-avoidance strategy intentionally, or if it is just a happy (for him) side-effect of his managerial incompetence and his intense personal desire to avoid criticism.  My guess is that it’s a bit of both.  He is a talented and experienced grifter, with an intuitive understanding of how to play an audience.  His impulsiveness and his thin skin may sometimes lead him to act in self-defeating ways, but he has a natural understanding of the politics of credit-claiming and blame-avoidance, and right now that ability is standing him (not the world) in good stead.

Mass testing for Covid-19: economics, politics, and policy options

The Covid-19 epidemic is creating a painful dilemma for policymakers.  On the one hand, we need to practice social distancing to keep people healthy and to prevent our hospitals from being overwhelmed.  Unfortunately, this strategy is causing a severe economic contraction as people avoid contact with others.

An ideal response to this dilemma would have three basic components.  First, we would implement a hard, nation-wide lockdown to slow the spread of Covid-19.  This would “flatten the curve” and save lives by preventing hospitals from being inundated with patients in the next few weeks.  It would also buy time to put in place the testing, prevention, and surveillance measures we will need to start cautiously re-opening our economy.  Putting these measures in place should be the second element of our strategy.  Finally, as Paul Romer and Alan Garber argue, we need a major effort to increase our capacity to test for Covid-19, and to produce masks, gloves, and other forms of personal protective equipment (PPE) by an order of magnitude or more.  The ability to do mass testing and to provide masks and other PPE to most Americans will substantially reduce the risk that the epidemic drags on for many months and leads to an economic catastrophe.  (For further discussions, see here, here, and here.)

In this essay I explain why it is important to massively increase our ability to produce Covid-19 tests and PPE.  I also discuss how this can be done, considering the apparent reluctance of the Trump administration to lead this effort.  I make four basic points.

First, the ability to test millions of people daily for Covid-19 and to produce PPE for millions of Americans will require a large up-front capital investment by manufacturers that may turn out to be unneeded, but this investment is socially justified to lessen the risk of a severe and protracted economic shutdown.

Second, without firm contractual commitments from the government, businesses will not invest at the scale required to ensure that we can avoid a disaster.  Several factors will deter adequate investment by industry; the most important is probably the risk that the epidemic will abate and they will not be able to recover their investment costs.  The government can overcome this problem by agreeing to pay companies for tests and PPE even if the epidemic abates, by subsidizing investment in the capacity to produce tests and PPE, etc.  The critical point is that the government needs to make binding commitments NOW, it cannot wait to see if the epidemic can be brought under control using other means.  Valuable time has already been lost.

Third, the powers that the President has under the Defense Production Act to directly control the use of resources are not particularly useful if our goal is to spur investment in the capacity to produce tests and PPE.  We need to give firms incentives to invest in new capacity using contracts, competitions, and similar tools.

Fourth, an ambitious effort to expand production of tests and PPE will inevitably lead to genuine contracting failures and to situations that create the perception of failure.  Trump is clearly anxious to avoid setting ambitious goals and taking actions that might later be used to criticize him.  In response, Congressional Democrats want to force Trump to exercise his powers under the Defense Production Act.

This is a mistake.  Trump would likely veto any bill that tried to force him to act, and, in any event, it is very difficult for Congress to force a reluctant President to act.  Fortunately, there is no need for contracting efforts to be directed by the President.  Rather than trying to force a reluctant Trump to exercise his contracting powers under the DPA, Congress should either delegate the power to an agency, or it should create incentives itself, directly.  I will sketch out how this can be done.  The same point applies to efforts to organize mass testing, the distribution of equipment, and other activities where direct commands are an effective means of achieving our goals:  Congress should accept that Trump is unwilling (and arguably unable) to lead these efforts and try to work around him in ways that he can accept.

Mass testing and distribution of PPE mitigate the risk of economic disaster