Pour ceux qui veulent comprendre l’arbitrage économique des mesures de lutte contre la pandémie , un bon article , en anglais
|Budish Covid-19 update|
Posted: 14 Nov 2020 01:58 PM
Eric Budish has an update to his excellent Covid-19 paper. Eric has a few deep central insights about pandemic management, which necessarily joins economics and epidemiology.
Eric has a few deep central insights about pandemic management, which necessarily joins economics and epidemiology. Keep your eyes on R<1.
The reproduction rate R — how many people the average person who gets the disease passes it on to — is really the only thing that matters.
When R>1 the disease grows, initially exponentially, then only tailing off when a large (half or more) of the population is either immune or dead.
When R<1, the disease tails off. The costs of the disease grow enormously when R>1. Once R<1, further reductions in R don’t really do much good.
From a public health perspective, you don’t have to stop all transmission. Just get R less than one.
Thus, The goal of pandemic policy must be to maximize the economy (maximize utility, if you’re an economies) while keeping R<1.
The costs of changing R are so smooth, and the benefits so nonlinear, we might as well treat R<1 as a constraint. ..the formulation provides economics language for a policy middle ground between society-wide lockdown and ignore-the-virus, and a new infectious threat response paradigm alongside “eradicate” and “minimize”.
Important simple insights: the R ≤ 1 constraint imposes a disease- transmission budget on society. Society should optimally spend this budget on the activities with the highest ratio of utility to disease-transmission risk, dropping activities with too low a ratio of utility to risk.
Contra most epidemiologists, you don’t shut down everything. You accept risk, and even some transmission, where it is important.
From my priorities, keeping business and school open is more important than bars nightclubs and parties, but gustibus do matter here. Market value is a good test however.
Second, masks, tests, and other simple interventions increase activities’ utility-to-risk ratios, and hence expand how much activity society can engage in and utility society can achieve while staying within the R ≤ 1 budget.
This is a deeply important point, which I really had not grasped: Do not evaluate the value of mask-wearing by how much it can reduce the spread of disease. Evaluate the value of mask-wearing by the vale of activities we can open up, while keeping the disease spread constant. That includes activities which can open safely if people wear masks while doing them, but also activities that can open if people wear masks elsewhere.
A simple numerical example, based on estimates from the medical literature for R0 and the efficacy of facemasks and complementary measures, suggests the potential gains are enormous.
Eric does not draw one conclusion, which I suggest he does: Policy should assiduously focus on measuring the reproduction rate, and policy initiatives should be keyed to that measure.
Right now, national, state, and local lockdown measures are keyed to the test positivity rate, which the media are also obsessed with.
The test positivity rate is about the dumbest number to look at and control. Using the test positivity rate or even the correct prevalence of infectious people to gauge policy guarantees covid cycles. The test positivity rate takes the people who happen to come in for any reason to get a test, and measures what fraction are positive. 10 in 100 is the same as 1000 in 100000.
As in that example, you can have the same test positivity rate with vastly different fractions of people in the community infectious, and thus vastly larger danger of going out. Even if we do measure the correct fraction of people infected, via random testing (a big improvement), it is a mistake to crack down when that number is large, and to ease up when that number is small.
Ease up when small leads to a high reproduction rate, and the cycle restarts. Measure, and respond to the reproduction rate.
The paper now includes a very good review of literature on simple interventions.
Do masks work? You can hear opinion on both sides. After all, aerosols go right through the space between mask and cheek.
Scandalously, There is not yet an RCT study of masks.
P. 16 ff, however, has a long list of references and a bottom line, the preponderance of evidence from various sorts of empirical studies, combined with common sense conceptual understanding based on how the virus is known to spread, all point to reductions that are significant, perhaps on the order of 50% or more in conjunction with complementary measures.
I wish that were a bit nuanced. After all, masks are clearly going to do a lot more good inside than 20 feet away from each other outside, and N95 masks are going to do a lot more good than a bandana. Masks have costs too, and blanket mask mandates impose costs.
A subtle cost is that since most people know that wearing a mask outdoors in a 20 mph wind is silly, if government mandates that, then that undermines authority of mask recommendation that help.
Random mass testing is a second easy intervention. Again, Eric has a novel insight: tests are therefore likely to be of especially high social value for activities that are both high pre-virus utility …and high risk … … Additionally, tests are likely to be especially socially valuable for activities where facemasks and other cheaper interventions either are not sufficiently effective…or are too harmful to utility …. Congregate settings such as nursing homes may be an example of the former … film and television production is an obvious example of the latter.
Eric’s conclusion is insightful: There are four features of Covid-19, relative to other past pandemics, that together make this formulation potentially appropriate:
1. Mortality / morbidity cost high: Covid-19 is sufficiently lethal and harmful that R ≤ 1 is a desirable policy goal even at meaningful expense.
2. Eradication likely not feasible: Covid-19 had already spread relatively widely by the time of policy intervention in many countries, making eradication an unrealistic goal for many countries.
3. R ≤ 1 feasible with modestly expensive measures: with an initial R0 in the ballpark of 2.0-4.0, and a fast understanding of how the virus spreads, medical experts quickly converged upon a suite of public-health responses that together could achieve R ≤ 1. As Atul Gawande put it: “we have learned in hospitals where we’ve been going to work every day in the pandemic and have avoided infections, that if you have hygiene, distancing, mandatory masks, and screen everybody for symptoms so that they stay home and get tested, that shuts the virus down.”..
.4. Minimize unboundedly expensive:… the minimize objective makes it difficult to think about tradeoffs if the interventions themselves are very expensive. (To see how clearly Eric is thinking here, keep reading for his analysis of cases that violate these assumptions, and other policies would be appropriate.) One fly in the ointment. Death rates are declining rapidly as better treatment comes about. Is this really sufficiently lethal, point 1?
A commenter on a previous blog post pointed to the declining
death rate in the US from all sources and points out that Covid-19 has put the US overall death rate at…where it was in 2006. Not good, but how much GDP and unemployment is that worth? If we want to spend that much GDP and unemployment to reduce death rates, are there not more efficient ways to do that? Ban sugary sodas rather than close bars?
The story of very large fractions of people with seemingly permanent after effects needs better quantification, as that seems now like the biggest unknown. But, paradoxically, the better treatment gets, the less the case for costly interventions.
Update: In response to comment. Yes, R is very different across activities, and most of the high average R is in a fat tail of super spreader events. That means that effort should be focused on super spreader events and not bother with small potatoes.
The grumpy economist.