Covid-19, the Evil Genius, and How to Think about Societal Risks

How much risk of loss of life people are willing to accept likely depends on how one asks the question


When I was in law school, my Torts professor was Dean (now Judge) Guido Calabresi, affectionately known to everyone as Guido. Each year, Guido gave his class a hypothetical that ran something like this: Imagine an Evil Genius comes to you before the automobile is invented and offers you all the advantages of the automobile, in exchange for killing 50,000 people a year (the rate of deaths in car accidents at the time in the U.S.). If you turn down the offer, society will organize itself in a way that's not automobile dependent, and, perhaps counter-intuitively, Guido asked us to assume that the 50K lives were not being made up by hypothetical lives saved by the existence of cars. Basically, the problem asks whether one would be willing to kill off 50K lives a year in exchange for the conveniences and pleasures of the automobile.

I objected to the hypo on the grounds that automobiles do in fact save as well as cost lives, but something else bothered me about the problem, and I couldn't quite put my finger on it until I thought about it again many years later.

I realized that what bothered me about the hypo was that it assumed that the right way to look at the problem was to assume that one is a central planner, and gets to decide for everyone whether to kill off 50K people. When you put it that way, it's hard to justify killing 50K people for mere pleasure and convenience.

But there is another way to look at it, a more methodologically individualistic way. In a country of 300 million people, 50 thousand deaths a year means you have, on average, a one in six thousand chance of dying each year in a car accident. So an alternative way of approaching the problem is to ask, "would most people be willing to accept an additional one in six thousand risk of dying each year in exchange for the conveniences and pleasures of the automobile." That might give you a different answer than the Evil Genius hypo.

But wait, there's more. The risk of dying in a car accident is very much dependent on one's own behavior. Driving while intoxicated, speeding, driving in ice storms, driving more miles a year than average, driving a less-safe car, and so on, can dramatically increases one's risk of dying, while the opposite behaviors can dramatically decrease one's risk. So the question might instead be, "Would the average person, knowing that he can dramatically increase or decrease his own risk depending on his own behaviors, be willing to accept an average 1 in 6000 chance of dying in exchange for…." Framing it this way allows the risk-averse to take into account their own risk-averse behaviors, and the non-risk-averse to accept a greater risk.

Turning to Covid-19, let's say a particular public health intervention is predicted to save 100K lives. If you ask someone if they would be willing to sacrifice 100K people to avoid this intervention, people will be inclined to say no, "I would never put a price tag on human life, much less 100K human lives." But let's say you ask the question differently: "Would you be willing to accept a one in three thousand chance of dying this year to avoid this public health intervention?" or "Would you be willing to accept an average of a 3% or so increase in your risk of dying this year to avoid this public health intervention" (given that the background annual death rate in the U.S. approaches 3 million)?

Once you ask the question in one of the latter two ways, then instead of focuses on the raw number of deaths, we can focus on the tradeoffs, and of course the answer will vary both by an individual's own risk of being one of the victims (the ill and elderly have much greater risk), and the effect that the intervention will have on their own lives. Individualized behavior is going to make less difference than in the automobile example, but the principle of looking at individual or individualized risk, rather than raw numbers of victims, is the same.

So let's take a simplistic, intentionally skewed example: the government wishes to undertake an intervention that the best economists predict will lead to a permanent reduction in overall wealth of 10%, with vast unequal distribution of burdens, say shutting down all businesses for six months. The same person who would never acknowledge being willing to consign 100K people to death for "economic" reasons may very well be willing to accept, both personally and for the collective, a one in three thousand chance of dying (or a 3% average increase in the chance of dying) this year, or even a much higher risk, so that both society as a whole doesn't become significantly and permanently poorer, and some people (e.g., small business owners) don't lose everything.

I can't, as a philosophical matter, tell you that looking at average individual risk or individualized risk vs. average individual cost or individualized cost is a sounder way of looking at the problem then the more collectivist, central-planner-type approach that Guido assumed. But it's certainly an *alternative* way of looking at social trade-offs, and one that appeals more to my own mindset. And to my understanding, the more collectivist approach dominates public health education, and will therefore likely dominate medical advice about Covid-19.

NEXT: N.H. Court Rejects Challenge to Ban on Gatherings of 50 or More People

Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Report abuses.

  1. If cars had not been invented, a lot of people would have found it much more difficult to get laid.

    “Here’s to parties we tossed, to the games that we lost!
    We shall claim that we won them, one day;
    To the girls, young and sweet, to the spacious back seat
    Of our roommate’s beat-up Chevrolet…” –Tom Lehrer

  2. Let’s look at this a different way. In 2019, 69,000 people died of drug overdoses. We’ve effectively shut down the US economy, while suspending civil rights, over COVID death numbers that are perhaps 1/100th of that…

    1. You are ignoring the whole point of the shutdown. It’s not (primarily) to reduce the total number of deaths (although that would be nice); it’s to spread them out in time so that we don’t have, at any one time, more people needing care than we can provide. You know, “flatten the curve”, so we don’t overwhelm our health-care system.

      1. I think you missed the point of what “flattening the curve” is actually designed to do. It is actually designed to reduce the number of deaths.

        It’s designed to spread out the number of CASES over time, so that the number of severe cases don’t exceed the ability of medical care to care for those severe cases. As a case without care is far more likely to result in death.

        1. In 2019, 69,000 people died of drug overdoses
          COVID death numbers that are perhaps 1/100th of that…

          Your use of current deaths seems to be swallowing quite an assumption about the future.

          1. The deputy city manager of San Jose just stated that in an optimistic scenario Santa Clara County will see 2,000 deaths from COVID-19 in that county alone by the end of May. I hope he’s wrong; if he’s anywhere close to accurate, the national death toll will be way more than the number of people who die from drug overdoses. Personally, I don’t want to take the risk that he’s right.

            1. Anyone can make up numbers out of whole cloth. The county health department made a point of saying they haven’t reviewed his methodology.

              1. You’re right that anyone can make up numbers out of whole cloth. And I don’t know where he got his numbers. But this is an important public official from the third biggest city in California, in the heart of Silicon Valley. I think it’s unlikely that he’s a quack.

                1. His numbers are an obvious fantasy. The implied true CFR for the US would be much greater than any rational estimate, probably by an order of magnitude.

        2. Well, ok, that’s part of it, but, overwhelming the health care system with too many severe COVID cases at the same time will also prevent patients who need care for other indications besides COVID from getting the care they need.

          In other words, if we have B beds in intensive care units, it’s better to have 5B covid patients who need intensive care spread out over three years, than to have only 2B covid patients all of whom need care the same day. The total number of patients (the integral over time of the number of patients at any moment) is much less important than the maximum number who need care at any given time.

    2. Well, we are well over 690 deaths by now.

      Besides, you overlook a few things:

      1. Part of the reason the number of deaths is not much higher is the measures that have been put in place.

      2. The point is to prevent an explosion of cases. The standard of comparison is not the number of deaths through today under restrictions, it’s the number there would be over a year without them.

      3. I know a simple, guaranteed method to avoid dying of an opioid overdose. Send me $100 and I’ll share my secret. Covid is different.

      1. We are…

        1a. True, deaths aren’t higher. But, let’s put this in context. We shut down the US economy, suspended essential civil rights, and dropped a couple trillion dollars on the problem. This is for a number of covid deaths that are currently less than 2000.

        1b. Meanwhile, the number of deaths due to drug and opiods since 1999 is in excess of 750,000. Shouldn’t more aggressive means have been used to control those deaths? Why or why not?

        2. Ditto..

        3. I also know a simple, effective way to avoid dying of Covid, which you can send me $100 for. Granted, it’s not necessarily easy. But the point being made here is that all those opiod and drug overdose deaths. Hundreds of thousands of them. They are treated as “different”. A horrible tragedy. But we won’t spend a few trillion dollars to avoid those hundreds of thousands of deaths. We won’t suspend civil rights to avoid those hundreds of thousands of deaths. We won’t shut down the US economy to avoid those hundreds of thousands of deaths. Because…reasons.

        Perhaps you know those reasons. Why those deaths aren’t as important.

        1. Imagine NASA announced a comet was hurtling toward Earth and would likely kill 90% of civilization. If we began putting millions in resources to find a way to deflect it, would you be objecting because drugs have killed 750,000, but comets haven’t killed anyone?

          Epidemiologists, who know far more about this stuff than you or I, ran various scenarios and projected 4 million deaths in the U.S. in the next 3-6 months. That’s an average of about 22,000 people per day for 180 days – or more than seven 9/11 attacks ever day for six months. That’s more American deaths than in all previous wars combined.

          But, yes, our hearts go out to your 401(k) and its temporary dip.

          1. “Epidemiologists, who know far more about this stuff than you or I, ran various scenarios and projected 4 million deaths in the U.S. in the next 3-6 months.”

            Expert Epidemiologists have also made predictions on number of deaths from ebola outbreaks, Sars, Swine flu, etc which have been widely over estimated.

          2. I love how the predicted numbers keep going up. That’s the great thing about predictions. You pick the largest prediction, then magnify it, then magnify it again….
            200,000…2 million…4 million…40 million…

            But that’s besides the point. The point was, why not suspend civil rights to save the hundreds of thousands who have died of drug overdoses?

          3. And other epidemiologists have said those predictions are wild over-estimates… and they’re the ones making better arguments to the public.

      2. 1/ measures taken have no measurable effect yet. Covid-19 has a long incubation time before first symptoms (if they develop… it’s at least 2 weeks) during which you can still spread the disease. And it’s basically a month before first infection and outcome. The deaths we’re seeing now were locked in weeks before any state took real measures. The earliest we’ll see any impact of lockdowns is mid-april.

        Stop living in a fantasy world where the government does something and you immediately see the consequences of that action.

        2/ ‘Flatten the curve’ graphics are measured in days, not weeks or years. It was already too late to flatten the curve when lockdowns were put into place.

        3/ Meh. I know a guaranteed way to reduce risk of heart disease and diabetes, yet plenty of americans are over-weight. Guarantee heart disease kills more people than covid-19 ever does. Should we institute food controls?

  3. It actually is an interesting question when applied to very successful mass-murdering governments. Does the 1.6M body-count of Soviet citizens who died in the gulags appear different when you remember that the population of USSR during the gulag-years was a bit more than one hundred times greater? Is a one-percent risk of being sent to a gulag unacceptable or unacceptable (for yourself personally, and/or for you deciding as proxy for the community), in exchange for the improvement in living standards for ordinary people in USSR relative to the dreadful conditions so many of those people endured before the Revolution?

    1. First, you assume without evidence that the Communists improved life more than the Tsar or the (?) first revolutionists would have.

      Second, you omit all the others of the 40 million murdered by Stalin and the Communists.

      1. Dude, he isn’t arguing for Communism.

        1. He is apologizing for Communism. What else would you call someone who pretends they were much less murderous than in reality?

          Guess what I call someone who apologizes for a Communist apologist? A buttinsky for starters. Funny how you don’t think he can defend himself and need to jump in to protect the poor fool’s honor. A statist for seconds.

          1. LOL No I’m not apologizing for (and not arguing for) communism. I’m probably one of the more fervent anti-communists here (except for Eugene). I spent a few months in USSR. A dose of communist reality is the best medicine for converting a communist away from communism!

            But there’s no point in denying history and there’s very little question that life for ordinary Russians and ordinary denizens of the other countries which became Soviet republics when USSR came into being was much better after the Revolution than it had been under Czarist feudalism.

            It’s a RELATIVE judgement. My point is not an apology for Communism; it’s an INDICTMENT of what they had just before the Revolution.

            Here’s my favorite joke about communism, which was all the rage at the University of Leningrad when I was there:

            Q: Что лучше, коммунизм или онанизм?

            A: Онанизм лучше, потому что хуй в руке лучше, чем пизда на горизонте!

            1. There’s very little question that the Soviets lied a lot about their economy. The tens of millions that starved to death under their policies certainly didn’t see an improvement in living conditions. Aside from that, it’s probable that honest economic numbers would show an immense drop in industrial production in the Revolution, and a slow climb that didn’t match their WWI GDP until over 20 years later. It’s probable that when Robert A. Heinlein toured Russia in the 1960, the population of Moscow was only half the official figures, because they still needed to conceal the effects of Stalinism seven years after Stalin died.

  4. Unfortunately, people are notoriously bad at assessing probabilistic risk correctly, so the alternate formulation doesn’t help much either. This whole topic is wrong, because we don’t get to make such judgements ahead of time. We do stuff, and other stuff happens as a consequence. When we see bad stuff happen, we can mitigate it, say by inventing seat belts, making cars that protect their passengers in accidents, and so forth.

    1. I call it “the dice-player’s fallacy”: a dice-player bets big that he will throw double-sixes, but throws double-fours and loses big. Then he says: “From now on, I’m gonna learn from my experience, and only bet on double-fours, not on double-sixes!”

    2. Indeed – it’s more palatable, but part of why that’s the case is because we’re failing to appreciate what we’re saying.

  5. This alternate framing of the question is reasonable as far as it goes. The problem is that the probabilities are not spread evenly across society; we’ll see lots of nurses and service workers succumbing while law professors, professional athletes and lawmakers get off much easier.

  6. These sorts of hypos are always an interesting exercise.

    I won a debate competition, many years ago, when I pointed out that I (if I had the ability to wave a magic wand), could easily save, each year, thousands of lives, save most adults a $1000 bucks or so directly, and maybe another grand in incidental costs, greatly help the environment, greatly help our infrastructure, and many other benefits. The cost would be zero to mild for most people, with a small handful suffering mild to moderate costs (of a type unrelated to physical health).

    The action would be, of course, returning the national speed limit back to 55 mph. An act that has tons of positives, with the major drawback being that it will take almost all drivers a few more minutes to get to where they are going–and a small number of drivers/truckers a significantly longer drive (15-20% longer).

    Of course, this is so politically toxic that the mere suggestions gets you laughed out of the discussion.

    Doing the smart thing, or the wise thing, is a million miles from doing the poltically-paletable thing. Yet more evidence of why we, globally-speaking, sort of suck. 🙂

    1. Ah yes…just a few minutes.

      Of course, those few minutes, over the course of a year, add up. The average one way commute is 26.1 minutes. For the sake of argument, let’s say it’s all going at the 65 mph speed limit. Dropping it down to 55 mph is a drop of 16%. So, that one-way commute goes up to 31 minutes. Just 10 minutes a day. Just 50 minutes a week. Just 2,500 minutes a year. Or 41 hours.

      So, what’s 41 hours a year worth to you?

      Luckily there’s some good math here.

      The US Department of Energy estimates that for every 5 mph you increase your speed over 60 mph, you’re paying $0.21 extra per gallon of gas (based on a price of $3 per gallon).

      Let’s say you have a 25 mile commute. At 55 mph, that’s 27 minutes. At 65 mph, that’s 23 minutes. That’s about $0.42 to go 4 minutes faster. Or $6.30 an hour. Below minimum wage. So, the question is, would you pay $6.30 an hour to get an extra hour of productive/family time?

      Most people would say yes…absolutely. That is a deal that is very much worth it.

  7. The difficulty with both versions of Calabresi’s hypothetical is this:

    If there are no cars society looks completely different than it does now. Mass transit is orders of magnitude better, possibly in ways we don’t imagine. There are no interstate highways, and intercity travel is by rail or air, so rail service too is much better than today. Suburbia is much smaller, as people place a huge value on being in walking distance of food stores, banks, post offices, etc.

    IOW, in that alternative universe, the convenience of the automobile is vastly less than it is today.

    1. Don’t forget the horse drawn wagons for local supply and the horse related inconveniences.

      1. And don’t forget about the ostlers and blacksmiths who lost their livelihoods when cars replaced horses and horse-drawn buggies.

        (If Donald Trump had been around, he’d have promised to bring those guys’ jobs back. Make America Equestrian Again!

      2. That is a problem.

        To be fair, let’s assume the intention is to use automobiles only for short-haul travel as you describe. Note though, that some of the local supply could be handled via a mass transit system – one much more complex than what we have today.

    2. Why would you expect other travel to be better if it suffered *less* competition? Competition is what incentivizes improvements in service, not its lack.

  8. Professor Bernstein, have you ever discussed your new perspective with Judge Calabresi?

  9. The same person who would never acknowledge being willing to consign 100K people to death for “economic” reasons may very well be willing to accept, both personally and for the collective, a one in three thousand chance of dying (or a 3% average increase in the chance of dying) this year, or even a much higher risk, so that both society as a whole doesn’t become significantly and permanently poorer, and some people (e.g., small business owners) don’t lose everything.

    As always, variance matters. To ask this question without discussing the distribution of losses is to ignore a critical point.

    1. In the case of Covid variance seems to skew toward the sick and old in terms of the death rate. Speaking as a member of the elderly cohort we are not valuable to society or even to our family, and as to the former more of a drag.

      The Spanish flu disproportionately killed healthy adults. It killed my grandfather at age 39, making a mess of my mother’s life as well as the rest of her family’s.

      So taking a view of what is best for society the argument for business as usual with Covid seems to be strong. Also most older people, if they aren’t in some sort of assisted living facility, can self-isolate fairly easily. They don’t have to work and already have a lifestyle that is congenial to that.

  10. Bernstein’s conjecture fails to model the coronavirus crisis, because his model returns repeatedly to an assumption that risk and reward are shared alike by each person. But the coronavirus crisis presents an altogether different moral proposition—whether some people in power may morally decide to kill selectively members of one specific tiny group, in exchange for a benefit of relatively trivially advantaging members of a much larger group, who are unlikely to share alike in the loss which makes the advantages possible.

    Policy must always be morally questionable which says some specific folks must pay an ultimate penalty, so that other specific folks can enjoy a small-to-moderate financial improvement. That is not at all like saying to each person, “Do you accept this potential reward, in exchange for this percentage chance of potential loss.”

    1. Exactly! Saved me from having to say it. As someone at quite high risk, I don’t get to choose risky behavior, but others can choose it and maybe kill me as a result.

      In addition, in an epidemic of this kind, if enough people choose to ignore guidelines because they accept the risk, it raises the risk for everyone when the medical system becomes overwhelmed. That means our risk chooser just significantly raised his automobile death risk, since our health system lowers that risk quite a bit under normal conditions.

      So, the hypothetical is not appropriate. There still is a legitimate issue of how much do we harm X to save Y – but this doesn’t get at that effectively.

      1. I agree completely. Not only that, when the hospitals are overwhelmed, the people most likely to die are the doctors and nurses. From a societal standpoint, we should be willing to expend extra resources to protect them, because they protect us.

        1. How are they different from soldiers? Society expends them without much thought, or at least without transforming society on their behalf.

          1. I don’t think this it is really true that we don’t transform society for our soldier. The GI bill following WW2 dramatically changed the our society giving many young men and education they might never have gotten before. Think about businesses that exist primarily to provide items for the military. Think about how much effort is expended on protecting soldiers soldier in the field. In the civil war soldiers in a line into fire. Today’s soldier is heavily protected with helmets, armour, specialized clothing.

            1. Well we did expend the lives of some 400,000 young men, and while we viewed that as an existential conflict I doubt anyone would claim the same for Korea, Vietnam, Iraq war I & II, Afganistan, etc.

              1. I was trying to make a slightly different point. Doctors and nurses are a much more finite resource than soldiers; when there’s a shortage, more people die. Liberia and Sierra Leone had a major doctor shortage during the ebola epidemic, but Doctors Without Borders came in to help. Doctors Without Borders can’t help us here; we’re too big, and the whole world needs doctors. We’re nowhere near that point at the moment, but if we try to save the economy at the expense of health care professionals, the elderly, and the immunocompromised, we may get there.

    2. Except the politicians (who matter) are, by and large, part of that at risk group. Which gives the perverse incentive to impose harm on lots of people to advantage their age cohort.

      And claiming the harm to young working american’s is ‘mild’ is myopic at best. Nor would allowing the economy to function be equivalent to deciding to kill some people – inaction does not imply the culpability of having made a decision. The utilitarian train track switching problem is flawed.

    3. “his model returns repeatedly to an assumption that risk and reward are shared alike by each person”

      It’s seems very typical for you to ignore or intentionally misread what I write so you can be outraged as something I didn’t say. For example, in contrast to the quote from your comment above, I actually wrote, “we can focus on the tradeoffs, and of course the answer will vary both by an individual’s own risk of being one of the victims (the ill and elderly have much greater risk)”

      1. Yes DB, I did see that. Which is why I said that you, “returned,” repeatedly to what amounts to an opposite kind of analysis—that risk and reward be shared alike by each person.

        On balance and in context, I think your comment was structured as a rumination between a little of this and a little of that, framing alternatives largely as equally legitimate choices. I don’t find them equally legitimate. I find one of them to be morally monstrous.

        If you share that view, perhaps you should have mentioned that among the means of distinguishing them. If you do not share that view, then you can hardly expect someone who thinks one of your alternatives is a moral horror to ignore your silence on the moral question.

        Finally, the quote you offer in your defense—“we can focus on the tradeoffs, and of course the answer will vary both by an individual’s own risk of being one of the victims (the ill and elderly have much greater risk)”—seems morally flawed in a similar way. I read that as a suggestion that self-interest of the ill and elderly is the critique, without other commitment to larger moral context. It is as if you are accusing the ill and elderly of some kind of NIMBY response in the face of proposals for their personal demise.

  11. Apply the hypothetical to nuclear power production.

  12. Experts stated at the beginning of this an estimated 80% of the world population will get covid 19. That means you and I will most likely get it if true. Who will and who won’t survive is an unknown at the moment, and may be forever. The later you get it, when less people have it and more is known will probably improve your chances of survival. Any risks or precautions you take are up to you.

  13. “Would the average person, knowing that he can dramatically increase or decrease his own risk depending on his own behaviors, be willing to accept an average 1 in 6000 chance of dying in exchange for….” Framing it this way allows the risk-averse to take into account their own risk-averse behaviors, and the non-risk-averse to accept a greater risk.

    This is interesting. On the one hand the more risk-averse individual faces a smaller increase in the chance of dying. On the other, that same individual is less likely to accept an increase in risk.

    I have no idea whether these two effects cancel, but they certainly counteract one another.

  14. There is an obvious difference between the two. Auto accidents typically incl love only a few people, the one of two cars in the accident (occassionally more.) a model in which everyone is treated as an isolated individual with their behavior having no affect on other has some – not much but some – approximation to reality.

    But an epidemic is different. There is much more interaction and network effect that doesn’t exist in an auto accident model. People’s individual behavior affects large numbers of other people.

    In general, a model in which everyone is an isolated individual whose choices affect only themselves rarely affect social reality. As an example, most people can’t realistically decide not to have cars and still be able to function in society. The decision isn’t really an individual one. The approximation to reality of modeling choices as individual ones not affecting others is extremely weak. In reality there is a strong network effect. When most people decide to go with cars, a tipping point is reached and the costs of not going along increase to the point where most can’t afford them. In reality, society actually does make decisions like this as a whole.

    You may have liked your BetaMax VCR, but did you keep it after people stopped making videos for it? Probably not. This is no different. Your choices are in fact highly constrained by what others do. And this is so whether their is a central authority channeling those choices or not.

    This is true for most phenomena in society.

  15. What I was taught about risk is that our perception is strongly affected by whether we choose the risk or it is imposed upon us. So we downhill ski with no concerns, but worry that a middle eastern man in the aisle over may blow up our airplane. So we don’t worry about the risk if we drive a car, but worry that we may get the coronavirus. Actually numbers on the risk are not as important as whether we chose that risk.

  16. When I commute to my job, I accept a risk of death from collisions. I’d guess that, except for unusually short commutes, this daily risk is higher for most people than COVID-19.

  17. You are ignoring, among other things, the costs due to the increased rate of sick leave in a pandemic scenario.

Please to post comments