The Non-Linear Loophole?

Last month, I asked readers for a “Great Reconciliation” of three popular beliefs:

1. Risk mitigation should be directly proportional to risk severity.

2. Medically speaking, COVID is 2-5x as bad as flu.

3. Our COVID mitigation efforts should be much more than 5x our flu mitigation efforts.

The most theoretically compelling resolution I’ve encountered maintains that contra (1), our response to risk should be strongly non-linear.

On the surface, this is a plausible story. Consider: How much money would someone have to pay you to endure a 1% chance of death today? Suppose your answer is $100,000. Does logic compel you, then, to accept a 100% chance of death today for 100*$100,000=$10M?

That’s plainly absurd. Indeed, unless your loved ones prefer fat stacks to your continued company, dying for cash is an exercise in futility. And since the value of risk is clearly non-linear, the value of risk mitigation must be non-linear, too.

Should we infer, then, that the War on COVID is prudent after all? Hardly. Sure, non-linearity makes sense when you raise a high risk. But approximate linearity still makes sense when you raise a low risk. If you disvalue a 1% risk of death at $100,000, would you really require far more than $110,000 for a 1.1% risk? Would you really require far less than $90,000 for a .9% risk? Remember, non-linearity is symmetric: If X increases faster than linearly, X should also fall faster than linearly.

Remember, moreover, that you face a long list of risks. They add up to a scary sum, but taken individually, even broad risks (e.g. “all accidents” or “all contagious disease”) are typically modest. So while it might be wise to take great efforts to halve your total risk, taking great efforts to halve any specific risk remains foolish.

Indeed, assigning non-linear weights to specific risks readily leads you to choose higher overall risk over lower overall risk. Suppose you face three risks: death by accident, death by contagious disease, and death by other. These initial annual risks are .1%, .1%, and .8%, initially disvalued at $10,000, $10,000, and $80,000. Now suppose that doubling the first two risks (to .2% and .2%) will reduce the latter risk by three-eights (to .5%), slashing your overall death risk from 1% to .9%. If you value each specific risk quadratically, you will disvalue these three risks at $40,000, $40,000, and $31,250. Upshot: Non-linearity counter-productively leads you to disvalue the safer package over the riskier package, merely because of the composition of danger.

Too abstract? What would you think about someone who categorically refused to drive on rainy days because inclement weather raises his accident risk by 25%? “Eccentric,” if you’re kind.

If non-linear risk valuation doesn’t explain the gargantuan global response to COVID, what does? I’ve invoked them before, and I’ll invoke them again: hysteria and herding. Novel, vivid risks lead to wildly innumerate overreactions, especially when all the other kids are overreacting too.

Alas, that’s human nature. Yet you can and should rise above such feelings and calmly tell the herd: You will not stampede me. And take comfort in the fact that ADHD shall save us – indeed, is saving us already.

READER COMMENTS

READ COMMENT POLICY

Steve

Oct 22 2020 at 9:45am

Non-linearity applies here too. If we can linearly increase our response to prevent an exponentially larger number of deaths, we should probably do that. COVID seems to have this property, whereas flu does not. In other words, while for any given individual the medical risk of COVID may be 2-5x the risk of flu, the total risk (risk to each individual * number of individuals afflicted) is likely to grow exponentially larger for COVID given a lack of non-pharmaceutical response.

jj

Oct 22 2020 at 10:55am

I can’t think you missed this, but did you ignore higher susceptibility and transmissibility of COVID? Getting the virus may be 2-5x as bad for an individual, but if you have 5x chance of getting it, the expected cost is 10-25x the flu.
2-5x is a decent estimate for the average effect (alpha) but it obscures the greater spread (beta). It’s probably 0.1x for kids and 10x for the elderly. People will pay to reduce beta.

robc

Oct 22 2020 at 1:59pm

Getting the virus may be 2-5x as bad for an individual

Actually, I think covid may be less bad for the individual, especially if young. I am pretty sure the IFR is higher for flu. The 2-5x bad is only due to the transmissibility.

Oct 22 2020 at 4:39pm

Here’s one data point supporting COVID severity <= Flu severity.

Death rate among COVID hospitalized = 7.6%

Death rate among Flu hospitalized = 8.9%

Mm

Oct 23 2020 at 8:05am

But COVID hospitalization rate much higher….

Thomas Hutcheson

Oct 22 2020 at 11:14am

A good analysis of why overreaction can happen. But surly another reason is not having available an “optimal reaction” alternative.

Chris Lawnsby

Oct 22 2020 at 12:56pm

In my view, the non-linearity isn’t captured just by the risk of COVID per se. So, for example:

200,000 COVID deaths are bad, but not apocalyptic.

1,000,000 COVID deaths aren’t necessarily only 5x worse.

It’s possible that after a certain threshold of COVID damage, OTHER bad things start happening. Maybe riots start in the street; maybe elections are cancelled; maybe public trust evaporates; maybe wars start.

I totally agree that some of the COVID hysteria is overdone– I personally know plenty of people who are simply uninformed about the actual personal risk of COVID.

Obviously, locking down too far could also cause the very problems I worry about. I’m just pointing out that it isn’t necessarily irrational to “over-worry” about the spread of COVID, even if it’s true that the actual biological risk isn’t as high as it seems.

Christopher A Lawnsby

Oct 22 2020 at 1:00pm

PS: I love your blog and your books so much. I teach high school math and I bring a lot of your ideas into the classroom– the kids love discussing them 🙂

mark

Oct 29 2020 at 7:41am

Then you may want the kids to spot a “mistake” in the math: 3 risks of 0.01 , 0.01 and 0.08 do not really sum up to 0.1 😉 With bigger numbers that gets obvious: One needs to survive 10 challenges, each has a chance of 10% to “kill” one (throw you out of the contest or al.). Are the chances of death 10*10%=100% and no one survives ? Nope. On average every third (ca. 34,9%) shall survive! (0.9 to the power of ten.) 0.99*0.99*0.92=0.901692 – almost a 10% for sure, (ca. 9.83), but still.

Oct 22 2020 at 2:00pm

I would say the lockdowns ARE an example of the other bad things happening.

Vivian Darkbloom

Oct 26 2020 at 3:40am

“Maybe riots start in the street; maybe elections are cancelled; maybe public trust evaporates; maybe wars start.”

I don’t think the recent riots in Rome were over “Covid damage”. They were about the recent introduction of curfews.

Scott Sumner

Oct 22 2020 at 12:59pm

You said:

“Medically speaking, COVID is 2-5x as bad as flu”

I strongly disagree here. The estimates I’ve seen are that Covid is much more than 5 times as severe. Perhaps you are comparing it to a severe flu, like 1957?

In a typical year, perhaps 30,000 die of the flu, although no one know for sure. We are rapidly heading toward 300,000 deaths, and that’s with massive social distancing and lots of mask wearing. If all of America were hit as hard as New Jersey, the death toll would end up north of 700,000. And even New Jersey had some social distancing.

Or perhaps you are comparing it to the flu if the US population were not being vaccinated. But that’s not relevant for our policy response.

JFA

Oct 22 2020 at 2:04pm

For the over 60 population, the IFR for Covid has been measured at 2% while the IFR for seasonal flu is about 0.8% (https://www.acpjournals.org/doi/10.7326/M20-5352) (the IFR is not dependent on whether there is a vaccine but is dependent on effective treatment). So conditional on getting it, you are about 2.5 times more likely to die if you get Covid relative to the flu (if this relative risk is constant across age groups).

But this is where Bryan doesn’t do a proper expected value calculation. To calculate your risk, you can’t just look at IFR, you have to look at IFR times the probability that you will become infected. Covid seems to be non-seasonal (it hit hard in the winter and hasn’t subsided in the warmer months), and the number of infections for Covid are under conditions of less interaction among the population.

The CDC estimates between 9 million and 45 million get the flu every year (it varies by year): https://www.cdc.gov/flu/about/burden/index.html. I would imagine the total number of Covid infections is towards the upper tercile or quartile of that range (and remember that’s with a lot less interaction among the population).

So there are many scenarios in which you can do your expected value calculation depending on how much the population is mixing, but you have to remember to multiply the IFR by the probability of becoming infected. That’s where I think the disagreement/misunderstanding is coming from: Bryan is not making that calculation.

On Bryan’s side though is that however you perform that calculation, the risk is lower than what most people think (though still higher than the flu). And given the large decrease in mortality due to Covid, this type of calculation is likely to result in quite different numbers as time progresses.

Oct 22 2020 at 2:14pm

if this relative risk is constant across age groups

The relative risk is clearly not constant across age groups. The distribution of deaths of covid and flu are pretty radically different. Assuming relative risk of contagion is constant (which seems more likely to me), then the IFR for the young is way lower for covid relatively. Possibly enough that the overall IFR is lower for covid (or at least very similar).

Oct 23 2020 at 8:08am

As many have said, the IFR for flu is more likely 0.4%- many cases aren’t reported/counted b/c (like COVID) they are mild and don’t come to attention of public health officials

Oct 22 2020 at 2:34pm

I just threw together some numbers real quick, I had a spreadsheet sitting around with covid vs flu death distribution by age from late April. Using it, and your estimate of 2.5 for >60, and the assumption of relative infection risk being the same across age groups, and some population distribution data that is probably a bit out of date but close enough, I cam up with an overall IFR of slightly less than 1.1 for covid relative to the flu.

With all the assumptions that went into that, and the drop in CFR from the spring to now, I think they are close to a wash. Covid is much more dangerous for the old. Flu is much more dangerous for the young. It about balances out, close enough I am not going to call it.

So I think the relative OVERALL risk between the two is that covid is more contagious, so Bryan’s original 2x-5x seems about right to me.

Mark Z

Oct 22 2020 at 3:05pm

“So I think the relative OVERALL risk between the two is that covid is more contagious, so Bryan’s original 2x-5x seems about right to me.”

You’re ignoring that we will only manage to have kept the death rate from covid to 10x that of the flu with the most draconian and costly public health intervention in American history (or inasmuch as lockdowns etc. are endogenous, by people taking unprecedentedly costly measures to avoid transmission).

The correct comparison is: imagine we behaved the same way with covid as we do with the flu, how many people would’ve died. Almost certainly many more. I’d also add that covid has become less fatal in large part precisely because we already spent lots of money to improve treatment and prevention.

Oct 22 2020 at 3:24pm

Or, possibly, the draconian actions had little effect as are results are the same order of magnitude as Sweden (although worse). So maybe, in fact, the actions have made things worse. Then again, we are still better than Belgium, so maybe there are other factors that are more important and social distancing efforts have a small, at best, effect.

Oct 22 2020 at 7:05pm

“Or, possibly, the draconian actions had little effect as are results are the same order of magnitude as Sweden (although worse).”

People in Sweden are not acting like it’s a normal flu season. Note that I’m not merely referring to government action, but to all the changes people have made to adjust to the pandemic: not spending time indoors with lots of people, wearing masks, standing 6 ft away from everyone, not visiting elderly relatives, etc. Swedes are doing most of these things as well. It strikes me as highly unlikely that if we had just continued to behave as we were in January, no face masks or distancing, continued to crowd into stadiums by the tens of thousands for sporting events, etc., that the deaths would not have been higher.

Oct 23 2020 at 7:01am

Nothing the Swedes did would be described as “draconian”. Your word, not mine.

Videos I have seen from Swedish malls look exactly like the US in January.

Mark Bahner

Oct 23 2020 at 11:20am

People in Sweden are not acting like it’s a normal flu season. Note that I’m not merely referring to government action, but to all the changes people have made to adjust to the pandemic: not spending time indoors with lots of people, wearing masks, standing 6 ft away from everyone, not visiting elderly relatives, etc.

My impression from numerous accounts (I have not been in Sweden) is that most people in Sweden do not wear masks or social distance…at least when outdoors:

“Surreal”

April 21, 2020: No masks in this picture

Oct 23 2020 at 12:11am

Thanks JFA. So even if the IFR were 2.5 times higher, his broader point is clearly wrong. Without measures to prevent Covid, it would be far worse than the flu is in a country where people are being vaccinated. I think that’s what Bryan misses. We have flu shots, but there is no vaccine for Covid (yet.) In my view, the death toll would have been more than 20X higher than a typical flu year, without preventive measures for Covid. Maybe 30X higher.

Oct 23 2020 at 7:03am

Why? What evidence is there of that? If that was the case, Sweden would be much worse than the US or the UK.

Dylan

Oct 23 2020 at 10:14am

I’m not sure of that. I think government action, while effective at the margin, is a lot less powerful than either side imagines. I know that there were all these pictures of New York right after lockdown that showed deserted streets and monuments, but that seemed to be a matter of being where people worked and selective timing. For the majority of lockdowns, my neighborhood seemed as busy, if not busier, than ever.

Behavior obviously changed, but I’m not convinced that New Yorkers have been more socially distant than Swedes over the last several months, no matter the difference in regulation.

Oct 23 2020 at 11:38am

Dylan,

I think we are in agreement on this one, I don’t think the regulations have done much (or the personal behavioral changes either) other than cause unemployment and tank the economy. I can’t, looking either at US states or western European countries, see any big difference in result that is due to policy or behavior. There may be some correlations but the size of the effect is small. Scott’s assumption that this would be much worse if we did nothing just doesn’t make any sense. Maybe a few more deaths earlier on.

I do think the original idea of flattening the curve to prevent hospitals from being overrun was a good idea, but that shouldn’t have been applied until hospitals starting feeling some pressure and loosened as hospitals were obviously not overrun. Accept that the area under the curve is going to be the same, the shape will just be a bit different.

Oct 23 2020 at 12:37pm

Rob,

I think the disagreement is that I think (and Scott too?) that the personal changes are what both saved a bunch of people’s lives and tanked the economy.

Where I think government policy might have helped, was by making the changes in behavior that were already happening more coordinated and acceptable. If I’m a business owner that wants my workers to be able to work from home but one of my competitors is still sending people out on sales calls, I might feel I need to do that too. If there weren’t rules keeping people from going to restaurants, but half my customers are voluntarily staying away. Waves of closures would happen anyway, but not in a coordinated fashion and without the political rationale for broad based support. That’s my speculation at least. Not highly confident on any of this one way or the other.

Oct 22 2020 at 2:03pm

This discussion reminds me of an interesting question I thought of a few days ago.

When, this winter, the # of cases of flu exceeds the # of cases of covid, can we at that time forget about covid and just go back to normal prevention.? As the IFR of flu is at least as bad as covid (especially for the young), I can’t imagine doing covid prevention when a bigger issue is going on.

SL

Oct 22 2020 at 2:31pm

The worst flu in the last decade killed 2,803 Americans aged 18-49. COVID has already killed 5,834 Americans aged 25-44.

Oct 22 2020 at 3:12pm

The estimated IFRs I’ve seen for seasonal influenza is in the .04%-.07% range when accounting for asymptomatic cases.

Oct 22 2020 at 3:13pm

Also, are you really expecting us to have a normal flu season? I expect we will see dramatically fewer flu cases and deaths than we do in normal years.

Oct 22 2020 at 3:21pm

I do too, I also expect most states (but not all yet) have blown thru most of the covid peak and while not at herd immunity levels yet, not that far away. So while flu may be lower than normal, covid will be even lower. So at some point, flu cases will overtake covid cases — at least in some areas.

Oct 22 2020 at 3:31pm

I’m not normally one to make wagers about real world outcomes, but care to bet on this?

Oct 22 2020 at 4:09pm

You realize “at some point” is entirely open ended so it would be impossible for you to ever claim victory?

I also don’t normally bet on these kind of things. If I did, I would have lost betting on covid never breaking 100k deaths in the US. However, I am 2-0 in internet bets…but both were about football.

But if you want to do a “told ya so” bet (I am petty enough to remember!), I would define the terms as (and this is to your advantage, I think):

Between November 1st, 2020 and April 30th, 2021, the 7 day average daily deaths in the US from flu will exceed that of covid as measured by the CDC.

That is far more to your advantage than my original statement. Which is another reason I won’t put money on it.

Oct 22 2020 at 4:29pm

Sure, I naturally expected that if we were to bet, that terms would have to be more carefully defined than a causal comment on a blog post.

Between November 1st, 2020 and April 30th, 2021, the 7 day average daily deaths in the US from flu will exceed that of covid as measured by the CDC.

That is far more to your advantage than my original statement. Which is another reason I won’t put money on it.

This is actually more to your advantage than how I’d interpreted your original meaning. I thought you were suggesting that flu deaths over the entire winter season were likely to surpass covid, not just a specific point estimate. Thank you for clarifying.

I still think you’re quite likely to be wrong on the new statement, although I don’t spend nearly enough time looking at weekly numbers and understanding delays and fluctuations in reporting periods to want to bet real money on a 7-day average either. But also, I don’t think that tells us all that much. If flu deaths for one week happen to be marginally higher than covid, but covid deaths for the entire season are 10x* the seasonal flu…does that evidence by itself tell us anything useful about our reaction to covid?

*NOT a prediction!

Oct 22 2020 at 2:46pm

Bryian isn’t thinking enough like an economist here, ironically (since I’m not an economist and he is). Even if our valuation of risk increases linearly, if cost per life (the amount we need to spend to save a marginal life) is a nonlinear function of lives saved, with increasing slope (pretty standard for a marginal cost curve), for both flu mitigation and covid mitigation, but the slope increases more slowly for covid, and the total value of mitigation is a linear function of x. For example:

y1 = (x^2)/3 (flu curve for cost of saving x lives)

y2 = (x^2)/4 (covid curve for cost of saving x lives)

z = 100x (cumulative worth of saving x lives)

Monetary unit is arbitrary of course. The optimal mitigation expense is the point where the cost curve intersects with the curve for z, the cumulative value of saving x lives. The flu cost curve, y1, intersects with z sooner than the covid cost curve, y2, does, because the former’s slope increases faster. Why might this scenario be realistic? For starters, covid is more contagious than the flu, so public health measures that reduce transmission save more people from covid at the margin than from the flu. We also already have a flu vaccine and effective treatments for the flu, further making the marginal value (in terms of lives) of spending still more to mitigate the flu lower than it is covid.

How much we ought to spend to mitigate a risk is not necessarily proportional to the risk, because the marginal cost of mitigating different risks differs from one risk to another, so obviously the optimal expenditure on risk mitigation – and the optimal number of lives saved from each risk – will not necessarily be the same, even if we are perfectly rational.

Darin

Oct 23 2020 at 11:31am

Amen! I’m sorry I missed the Twitter poll about this heuristic when it came out. It is a terrible rule-of-thumb, leading to all sorts of mischief. As you correctly state, the real criterion for evaluating interventions is benefit-cost > 0, where

Benefit = avoided risk, and

Cost = opportunity cost.

As a practical matter, the hardest term when it comes to COVID is avoided risk. It involves a counter-factual: how many people would have died if we hadn’t X, Y, or Z? Politics ensues.

The second-hardest part is figuring out opportunity cost. It makes no sense to save a $10-million-life from COVID, even at a cost of, say, $2 million, if there are cheaper options available. We are very focused on deaths from COVID, which is pretty silly.

Alan Goldhammer

Oct 22 2020 at 2:59pm

This is a nonsensical post and totally ignores the science and epidemiology about SARS-CoV-2. The simple equation that I posted back on one of David Henderson’s blog posts is this: population * herd immunity rate * case fatality rate = deaths

population is 330M; herd immunity rate is between 25 -75%; case fatality rate is between 0.3-0.6%. I’m sure that all those who post here can do the simple math. My estimation based on what I’ve read since March is the herd immunity rate is likely 50% (and note from the anti-vax efforts we know that herd immunity for whooping cough, measles, and chicken pox is between 80-90%) My estimation is that continuing hit or miss public health policies, mortality may be as high as 500K for the pandemic (if the herd immunity value is higher that number will be as well.

Assessing risk is always a personal calculation but there is probably 1/3 of the US population who are very cautious about COVID-19. Until a vaccine arrives they are not going to restaurants, flying on a plane, working in an office building, etc. Until this changes there is not going to be any more of a rebound than we have seen to date.

By my calculation back in late April, we would need to save over 2MM lives to justify the money that had been spent AT THAT POINT. Worst case scenario on your ranges is still under 1.5MM.

At the time I thought it was ridiculous. Less so now, but still doesn’t make any sense to me.

Eric

Oct 22 2020 at 5:39pm

I can give my statistician’s perspective. This is the difference between how the public and statisticians interpret the probability and risk.
When medical news says smoking increases the (relative) risk of having heart attack by 50%, to statisticians, that means among 100 non-smokers, 2 have heart attack, while among 100 smokers, 3 get heart attack. Most lay people interpret as if I don’t smoke I have 2 heart attacks per year, if I smoke it would be 3 per year.
Insurance companies use the statistician’s version to calculate the premium, and it is linear.

Jon Leonard

Oct 22 2020 at 6:31pm

On reflection, I object to the first statement: “Risk mitigation should be directly proportional to risk severity.” Surely it should be proportional to the marginal effectiveness of mitigating the risk? That is, we would ideally spend the most effort on the problems where effort helps, and less (or none) where the effort doesn’t help. In the context of cold calculation, it’s not “How many lives might be lost?”, but rather “How many lives might I save per dollar of effort?”, and they’re really not the same question.

Oct 23 2020 at 10:51am

On reflection, I object to the first statement: “Risk mitigation should be directly proportional to risk severity.” Surely it should be proportional to the marginal effectiveness of mitigating the risk? That is, we would ideally spend the most effort on the problems where effort helps, and less (or none) where the effort doesn’t help.

Excellent point! Even if the risk is severe, it makes no sense to spend money/time on mitigating the risk if the marginal effectiveness of that money/time spent is low.

Oct 23 2020 at 11:00am

One aspect of this pandemic I think emphasizes the point of focusing money/time on mitigation measures that are marginally effective. A tremendous amount of money/time was spent early in this pandemic in manufacturing new ventilators. This money/time was spent even though it seemed obvious (to me at least!) that money/time spent manufacturing new ventilators was not marginally effective, when compared to other measures. The ventilators were tremendously expensive, and they had very limited benefit, in terms of saving life-years.

Ron Browning

Oct 23 2020 at 7:59am

When “we” attempt to mitigate risk effectively and efficiently, “my” efforts to mitigate “my” risk, effectively and efficiently are hampered. My efforts to mitigate my risks have mostly been directed to counter “our” risk mitigation efforts.

Jim Hunziker

Oct 23 2020 at 10:00am

It’s in this sense that non-linearity is key: an infectious disease that’s 2-5x worse than the flu saturates any resource that’s planned to operate only at flu capacity. So the risk isn’t 2-5x worse than the flu, even though the disease is 2-5x worse. The risk is far greater because overwhelmed hospitals is the actual risk scenario we’re trying to prevent.

PaulS

Nov 1 2020 at 10:34am

Ummm…not exactly.

The goalposts shift with every passing news “report”. The pretense for wildly disproportionate “action” is “saving lives”. But hardly anyone will actually avoid the virus by dragging things out; the integrated dose, as it were, will not change appreciably. So then they switch to caterwauling about the hospitals. But even when it got out of hand in NYC,that hospital ship wasn’t really used. So then they switch back to “saving lives”.

No, it’s about virtue signaling and politicians hoarding power.

The Imperial College study that started the panic claimed an 0.6% risk of losing 10-11 years of life expectancy, both on average, or about 3 weeks, with the 3 week product largely independent of age. That quickly dropped by a factor of 3.

If the Devil asked me, or had asked me, if I would accept two years of partial house arrest in exchange for a week of life expectancy, I’d tell him where to go, in rude language.

Alas, people are easily frightened cowardly sheep. See the cover of The Myth of the Rational Voter.

Philo

Oct 23 2020 at 12:06pm

I remain puzzled by your endorsement of this principle: “Risk mitigation should be directly proportional to risk severity.” You ignore the question, How effective are the different available methods of mitigation? Take two new threats, each of which raises my probability of death in the next year by the same amount (1%, or whatever), if I take no countermeasures. I disvalue each of these threats in the same amount ($10,000, or whatever). The first threat I can completely neutralize—there is a 100% effective countermeasure—at a cost less than the threat’s disvalue. Then I should spend that amount, on (total) mitigation. Against the second threat I am helpless—there is no countermeasure that is at all effective. In that case, I should spend zero on mitigation.

Identical risk severities, widely divergent rational expenditures on risk mitigation.

Oct 23 2020 at 3:49pm

Yes, Jon Leonard basically made the same point (October 22nd, 6:31 PM). The true goal should be to save as many lives as possible per dollar spent, rather than to expend money related to risk severity.

Rob Rawlings

Oct 23 2020 at 10:11pm

This is an extremely good point (I would be curious how Bryan would respond).

Based on this insight one would predict that medical spending across all known diseases would tend towards an equilibrium where a dollar of spending on treatment for each disease equals the marginal benefits ( in lives saved and suffering alleviated) as adjudicated by the market or whoever is in charge of healthcare spending. Is there any evidence that this happens ?

However when a new disease come along that kills lots of people but that little is know about then the optimal medical spend for this disease will not be know until a period of trial and error and that is what I assume we are now going through.

I do think though that ‘save as many lives as possible per dollar spent’ should be the basis of policy (perhaps adjusted by some life-years criteria) .

Oct 23 2020 at 11:43pm

Yes, absolutely. Rather than writing, “The true goal should be to save as many lives as possible per dollar spent, rather than to expend money related to risk severity”…I should have written:

“The true goal should be to have as many years of healthy life per dollar spent, rather than to expend money related to risk severity.”

It’s very curious that Bryan should lead with “Risk mitigation should be directly proportional to risk severity”…without even seemingly considering Lionel Robbins’ classic definition of economics (which I learned of via writings of the economist Thomas Sowell):

Economics is the study of the use of scarce resources which have alternative uses.

Oct 23 2020 at 7:17pm

In 2019 the CDC reports there were 2,813,503 deaths in the USA. It seems very likely that there will be at least 280,000 covid deaths in 2020. I conclude that the average American has a roughly 10% greater chance of dying this year than last year. Of course for some the increase will be way more than 10% and for some far less. Nevertheless I’m sure that many people would be prepared to pay quite a high premium to stop the societal death rate by increasing by 10% (even if our own chance of death remained quite low).

It is kind of annoying though that most of us have paid that high premium and we have no idea if it made any difference or not to our (or anybody else’s) chances of surviving !

Oct 23 2020 at 11:54pm

In 2019 the CDC reports there were 2,813,503 deaths in the USA. It seems very likely that there will be at least 280,000 covid deaths in 2020. I conclude that the average American has a roughly 10% greater chance of dying this year than last year.

…

Nevertheless I’m sure that many people would be prepared to pay quite a high premium to stop the societal death rate by increasing by 10% (even if our own chance of death remained quite low).

However, you need to consider that all of the 280,000 people who die of COVID-19 were going to die anyway, at least some time. A very important question is “When were they going to die if they didn’t die of COVID-19 this year? For example, if (hypothetically) all of them were going to die in 2021 anyway, would many people be willing to pay that “high premium to stop the societal death rate by increasing by 10%…”? Certainly in that hypothetical, I don’t think so.

David

Oct 24 2020 at 5:48am

I think that the US is probably over preventing COVID, but I don’t understand why so many people who make this argument insist on lowball estimates of severity.

COVID is 10-20x worse than the flu. COVID IFR is around 0.5% and herd immunity ~ 50%. Flu IFR is about 0.05%, and attack rates are around 30%.

https://www.bloomberg.com/opinion/articles/2020-08-06/revisiting-how-covid-19-ranks-with-seasonal-flu-and-1918-pandemic

How much more is spent on covid prevention than flu prevention? It’s probably in the 100x range. Still much more than the ratio of severity, so you can still make your argument, and don’t have to rely on crankish estimates of severity.

Ray Lopez

Oct 26 2020 at 9:14am

Leaving aside that the probable chimeric virus causing Covid-19 leaves worse long-term problems than the common flu, and using only deaths, it’s on track to be 3x worse than the worse non-pandemic flu ever, the 1967 flu season. The only deadlier flu seasons were in 1967 when about 100,000 Americans died. Pop of 1970=203M, pop today = 332M, Covid-19 deaths after six months = 225k (projected 500k in one year), hence, Covid-19 is 3.1 times worse than the worse non-pandemic flu season ever, the 1967 flu season. Math: 500k / 100k*(332/203) = 500/163.5 = 3.1

It’s not clear that having 10 different models of flip flops and 15 different flavors of spaghetti sauce (the luxury of choice that is America) is worth having the SARS-CoV-2 virus become part of the four or five natural flu viruses that are endemic to human populations and cannot be eradicated. Hence, the USA should push for a non-linear response, a “Greek”, “New Zealand”, “South Africa”, “China” style lockdown, one more time. It probably however will fail, but worth a try.

PS– scientists say nobody has ever actually produced a medically efficient vaccine against a coronavirus. At best, as with existing flu shots, you have a 40-60% reduction in flu after taking the shot.

Jens

Oct 27 2020 at 5:52am

Medically speaking, COVID is 2-5x as bad as flu.

When i read the question for the first time, i paused because i got the impression that it also contains some “semantic non-linearity”.

A friend in school lost a finger while working on the farm with his father. Most of the people he met did not even notice, because he developed enormous skill to hide the stump (left ring finger). The missing finger didn’t impede his overall dexterity too much either. How would it have been if he lost 3 to 5 fingers? Absolutely not subsumable under “medically speaking”?

Does having Covid-19 feel 3 to 5 times worse than having the flu and is that medically speaking ?

I had appendicitis about ten years ago. It took almost 2 days to get the diagnosis. But after the appendix had been removed minimally invasively, i was actually quite fit again the next day. I also had the flu two years ago (presumably). It took me a week. So both cost me 4 to 5 days, in fact the appendicitis a bit less, so therefore appendicitis and flu are roughly equal medically speaking ?

Let’s say one illness costs me a month to cure it and another illness takes 5 months. Is the second disease, medically speaking, 5 times as bad as the first?

One disease reduces life expectancy by 1 year, another disease reduces life expectancy by 4 years. Medically speaking, is the second disease 4 times as bad as the first?

If two diseases differ in several of the above aspects, how can these differences be reduced to a scalar size so that one can say “disease A is x times worse than disease B” considering all the difference sets.

This comment is very good on the original post: https://www.econlib.org/the-great-reconciliation/#comment-257015

chrisare

Oct 28 2020 at 2:42am

What always seems to be missing in these exchanges is an attempt to account for the uncertain duration of the pandemic. This of course relates to how the pandemic will end.

Our appraisal of the right tradeoffs likely changes over time and depending on the outcome. For example, a lockdown for six months till a vaccine was developed with 100% certainty may have been tolerable. A lockdown which after two years will yield a 50% chance of a vaccine is far less tolerable.

Charlie

Nov 19 2020 at 12:15pm

We are certainly acting like this is 10x or more worse than the flu. Closing businesses, schools, etc.

So another way to ask your question is, if this is only 2-5x worse than the flu why such extreme reaction.

A lot of the reaction is coming from the Medical community. Hospitals are built to run at capacity and so a surge of new patients from a new disease pushes them beyond their limits. That is another aspect if this being non-linear.

So we need extreme measures until we incorporate this new disease into our planning for Medical services and capacity.

Comments are closed.