Hard choices

Original published 6 April 2020; updated 7 April 2020 to correct an error

Exit strategy or roadmap to recovery?

I wrote about exit strategies four days ago, on 2 April. In the last week, it seems like this has become a dominant topic of conversation in “quality” media. And no post I’ve written has generated more feedback from friends, many of whom felt that I was being too pessimistic.

Many commentators are suggesting potential exit strategies, or what the American Enterprise Institute (AEI) calls a “roadmap to reopening.”

I want to share a few arguments that have partially moderated my view about how hard / long / fraught such a roadmap might be (though not fully).

My view on the likelihood of a vaccines in the near-term may be too pessimistic.

In my post, I argued that a widely-deployed vaccine in 2021 or even 2022 was unlikely:

First, we’ve tried for many years, and spent a lot of money, to develop vaccines against other viruses (common cold, AIDS, SARS, MERS) without success.
Secondly, if the true IFR (infection fatality rate) really is 1% or less, as seems likely, we will need to have confidence that severe side effects occur in well below 1 in 100 people who are vaccinated. It will take human trials of significant scale and time to conclude that.
Third, we need to overcome the challenge of manufacturing, distributing, and administering billions of doses around the globe; and we also need to get high (probably >2/3) compliance; something that we struggle with for some vaccines.

The two best arguments I’ve heard against this are as follows:

1. We will deploy talent and resources on an unprecedented scale, and bulldoze any regulatory or even normal safety hurdles, to get a vaccine done — think Manhattan Project — because an effective vaccine is literally worth trillions of dollars of economic benefit (leaving aside the health benefit). Looking at the success of prior vaccine development attempts is irrelevant; it would be like trying to form your base rate of the chances of putting a man on the moon by looking at past attempts.
2. We don’t need a perfectly effective, perfectly safe vaccine to make a difference. Say a vaccine had a 2% severe side effect rate. It would be certainly rational for at-risk people (e.g., elderly, significant pre-existing conditions, health care workers) to take it, which could contribute significantly to building herd immunity. Also, with a good serological test, we don’t need everyone to take it; only those who don’t have acquired immunity.

I’m not completely won over by either argument, but they are both make me realise that the right question isn’t, “When will we have a vaccine that we know with near-certainty to be safe, that everyone in the world will have taken;” to, “When will we have vaccine that we have good evidence is safe enough for at-risk populations to take, and in sufficient quantities to begin to make a difference in reopening the economy?” The likely timeline to the latter is faster than that to the former.

The economic vs health tradeoffs will shift over time to a new equilibrium

I made this point partially in the last post, but the argument has become increasingly compelling in my mind.

Many countries today, when faced with the choice between terrible health outcomes and terrible economic impact, have chosen to improve health outcomes at the expense of economic outcomes. But that calculus is not static.

There are two components to the argument.

First, economic costs (and the associated political costs) are increasing over time; and the damage done to an economy by shutdown is not linear with time. This is obvious when considering the limit cases: an extra one-day national policy does little harm, and asking everyone to stay home for the next decade would be economic suicide. While there are many different guesses about how long a shutdown we can weather before long-term damage is done to the productive capacity of an economy (it surely varies widely under different situations), conversations with economics, policymakers, and business leaders all converge on anything much more three months at the outside, being severe and lasting.

Lost GDP, lost wages and profits, personal and business bankruptcies, and the like are not the only source of increasing damage over time. Many emerging markets, and some developed countries (Italy at almost 150% debt-to-GDP ratio? Greece at 180%?), face severe constraints and potential future default or devaluation.

So for these reasons and more, economic pain goes up over time, likely at an accelerating rate.

At the same time, the health consequences of at least partially reopening the economy go down over time. We’re building more health care capacity (beds, ventilators, professionals), replenishing PPE, building testing capacity, adding serological testing to PCR testing, and developing better treatment protocols. We’re better able to identify and isolate vulnerable populations. We’ll likely have better therapeutics quickly. The chance of a vaccine goes up over time. Herd immunity is building, albeit slowly; and if we can reliably identify those with acquired immunity, we can allow them to return to work or to the frontlines without significant restrictions. On top of that, we can build effective contact tracing and quarantine approaches.

As a result, it’s quite possible to imagine a time in the near future when a healthy, younger individual could rationally assess his or her chance of dying of COVID-19 if infected as being at the level of seasonal flu.

As health costs of reopening go down, and economic costs go up, we will surely reach a new equilibrium. The open questions in my mind are two:

1. How quickly?
2. Might we reopen too quickly, experience a serious second wave, and have to shut down again?

Faced with unaccepted health versus economic choices, we may be willing to sacrifice more privacy and liberties

My thinking is less developed here, but the general argument is that we have examples from China, Singapore, and South Korea of how a combination of (a) strong technological surveillance, control, and communications technologies (e.g., apps that notify everyone about nearby confirmed/suspected cases; apps that publicly share your green/yellow (amber)/red status), and (b) invasive policies (e.g., forced quarantine, removal of infected family members) might allow for an equilibrium of greater openness with acceptable health risks.

I want to do more work to understand exactly what these measures are and how they are working, and will reserve that for a future post.

---

Hard choices for policymakers

On this topic, The Economist has an outstanding Briefing in the 4 April edition: The hard choices covid policymakers face. I highly recommend reading the full article, which is not behind the paywall as of now.

The article takes head-on the question of the tradeoffs between health and economic considerations, and how these may evolve over time. A few of the most salient points:

1. Policymakers in practice had no choice but to put in place strict control measures. The examples of China and Italy, in addition to epidemiological models, made it clear that how bad the possible, or even likely, outcomes could be without action. It many cases, it would have been politically impossible to allow the epidemic to continue without significant action.
2. Epidemiological models have severe limitations in their ability to accurately forecast the future, but allow us to ask “what-if” questions. They can’t answer hard policy questions on their own.
3. Governments will increasingly come under pressure to balance the health benefits and the economic costs of strict controls. How should they do this?
4. Even without policy changes the economic impact would have been significant in the early stages, and taking action was the right thing to do:

“[I]in the acute phase of the epidemic, a comparison of costs and benefits comes down clearly on the side of action along the lines being taken in many countries. The economy takes a big hit—but it would take a hit from the disease too. What is more, saving lives is not just good for the people concerned, their friends and family, their employers and their compatriots’ sense of national worth. It has substantial economic benefits.

5. At the same time, the economic impact will be severe, leading to lost income, individual and business bankruptcies.
6. Some fear that the economic impact will have its own severe impact on health, and even cause a higher rate of death. This is not clear from past downturns: “mortality is procyclical: it rises in periods of economic growth and declines during downturns.” Overall mortality fell during the Great Depression.
7. But economic costs will increase over time and individuals’ may be less willing to accept the impact over time as well. So we should expect greater pressure to restart the economy over time.
8. Vaccines are not the only path to relaxing constraints:

The rudiments of such a plan would be to ease the pressure step by step, not all at once, and to put in place a programme for picking up new cases and people who have been in contact with them as quickly as possible.

9. Some evidence from the flu pandemic of 1918-19 suggests that cities that worked longer and harder to step the spread of the flu performed better economically.
10. But there are fears that longer shutdowns could do greater structural damage.

---

Immunity passports vs false positives

I’ve been guilty (though in good company) of discussing serological testing as though it would be a panacea; of suggesting that once someone tests positive for the antibodies that we think confer resistance, she could resume any activity as though an immune superhero.

Indeed, the UK’s plans apparently include the idea of “immune passports.“

A lot of the discussion around serological testing has focused on how quickly they can be deployed at scale, and whether or not they are accurate enough.

But I have bad news. Serological testing will likely not be as useful as I had thought in the near term, for obvious reasons that I’m ashamed not to have thought through.

The reason is the same reason that tests for breast and prostate cancer are often not regarded as that helpful: the ratio of the rate of false positives to the rate of the disease in the population.

As epidemiologist Zachery Binney explains here, even a test with high sensitivity (the percentage of positives that are reported as positive) and high specificity (the percentage of negatives that are reported as negatives), if the true rate of the positives in the population is low, the test may produce a high ratio of false positives and not be very predictive.

This is really counterintuitive for most people, even those who like math. It’s worth working through an example to make the problem clear; but if you prefer, you could read the Wikipedia article or (even better) Zachary’s powerpoint.

[Note: math error corrected in the example below on 7 April 2020]

To make the math easy, let’s say our serological test has 95% sensitivity, 95% specificity, and that 5% of the population (say, in the US) actually has had COVID-19. Now we test 1,000 people. What happens?

50 people actually have it; 950 do not.

The test has 95% sensitivity, meaning it correctly detects 48 of the 50 people who have had it (“true positives”), but misses 2 of them (“false negatives”).

The test has 95% specificity, so it correctly detects that 902 of the people have NOT had it (“true negatives”), but it incorrectly reports that 48 people have had it who did not (“false positives”).

So (48)/(48+48) = only 50% of the time does a positive result actually mean that you have had COVID-19!

Now, this gets better over time for several reasons. Serological test will improve; there may be multiple kinds of tests that give uncorrelated results, in which case we could administer more than one; and the actual prevalence of immunity in the population will. be increasing over time.

But the key point is that serological testing, while potentially useful, is far from a panacea today.