It is a great fault of symbolic pseudo-mathematical methods of formalizing a system of economic analysis . . . that they expressly assume strict independence between the factors involved and lose all their cogency and authority if this hypothesis is disallowed.

—John Maynard Keynes

I ended my last commentary by swearing to leave off thinking about the coronavirus. Alas, I am weak. The situation is bleak and depressing; it has affected nearly every aspect of my life, from my free-time to my work, my exercise routines and my relationships; but it is also, if one can be excused for saying so, quite morbidly absorbing.

What especially occupies me is how those in charge will weigh the costs and benefits of their policies. Because the threat posed by coronavirus is so novel, because these decisions involve human life, and because it is difficult not to feel afraid, I think there is a certain moral repugnance that many feel toward this kind of thinking. However, as I argued in my previous post, I think truly moral action will require a thorough appraisal of all of the many potential consequences of action and inaction. This will make any choice that much more difficult, and I do not envy those who will have to make it. 

As anyone familiar with the famous trolley problem knows, moral dilemmas often involve numbers. If the actor has to choose between a lower and a higher number of victims, one must choose the lower number. However, there are several refinements of the problem which show the limitations of our moral intuition. For example, respondents are willing to divert a runaway trolley onto a track where it will kill one person rather than five; but respondents are unwilling to push an enormously fat bystander onto the tracks to save five people. We seem to be willing to think in purely numerical terms only about those involved ‘in the situation’ and unwilling to do so with those we perceive as ‘outside the situation.’

Well, in case we are facing, virtually everyone is ‘inside the situation’; so this leads us to a numerical treatment. But of course this is not so simple. What should we be measuring and comparing, exactly? I raised the question in my last post about this calculus of harm, and how it seems impossible to compare different types and levels of harm. As hospitals get overwhelmed, however, and care begins to be rationed, doctors are forced to make difficult choices along these lines, giving treatment to patients with the highest chances of recovery. Politicians are now faced with a kind of society-level triage.

One obvious basis of comparison is the number of lives lost. This is how we think of the trolley problem. But I think there is a case for also considering the number of lived years lost. What is ethically preferable: allowing the death of one person, or allowing the lifespans of 10 people to be reduced by 20 years each? I cannot answer this question, but I do think that the answer is not easy or self-evident. Reducing somebody’s lifespan may not be ethically on a par with letting someone die, but it is still quite a heavy consideration.

Further down the line, ethically speaking, is quality of life. Though it seems egregious to weigh death against quality of life issues, in practice we do it all the time. Smoking, drinking, and driving carry a risk, and a certain number of people will die per year by engaging in these activities; but we accept the cost because, as a society, we apparently have decided that it is “worth it” in terms of our quality of life. But of course, this comparison is not exactly appropriate for the case of coronavirus, since we ourselves make the decision to smoke or drive, whereas the risk of coronavirus is not voluntary. Thus, to save lives we should be willing to accept a greater loss in quality of life in this case, since we cannot control our exposure to the risk.

How exactly we choose to weigh or balance these three levels of damage—lives lost, lives shortened, and lives made worse—is not something I am prepared to put into numbers. (I suppose some economist is already doing so.) But I think we are obligated to try to at least take all of them into account.

Now, the other set of variables we must consider are empirical. On the medical side, these are: the lethality of the virus and the percentage of the population likely to get infected. On the economic side there are obvious factors like unemployment and loss in GDP and so forth. There are also factors such as loss in standard of living, homelessness, and the poverty rate; and still more difficult to calculate variables like the rate of suicide and drug addiction likely to result.

One major problem is that we know all of these variables imperfectly, and in some cases very imperfectly. To take an obvious datum, there is the virus’s lethality rate. From the available numbers, in Italy the fatality rate appears as high as 8%, while in Germany it is as low as 0.5%. This huge range contains a great deal of uncertainty. On the one hand, there is a good case that Germany gives a more accurate picture of the virus’s lethality, since they have done the most testing, about 120,000 a day; and logically more testing gives a more accurate result. However, we should remember that the virus’s lethality rate is not a single, static number. It affects different demographics differently, and it also depends on the availability of treatment. All of these factors need to be taken into account to establish the virus’s risk.

Complicating the uncertainty is the fact that the virus can create mild or even no symptoms, thus leaving open the question of the total number of cases—a number that must be known to determine the lethality rate. Asked to offer an estimate of the total number of infected people in Spain (the registered number is about 45,000 as of now), mathematicians offered estimates ranging from 150,000 to 900,000—and, of course, these are little more than educated guesses. If the former figure is correct, it would put the lethality at around 2%, while if the latter is correct the lethality is about 0.4%: another big range. 

Now that Spain is receiving a massive shipment of tests from China, our picture of the virus will likely become much more accurate in the coming days and weeks. (Actually, many of these tests are apparently worthless, so nevermind.*) However, one crucial datum is still missing from our knowledge: the total number who have already had the virus. To ascertain this, we will need to test for antibodies. It appears we will begin to have information on this front soon, as well, since the UK has purchased a great deal of at-home antibody tests. I believe other countries are following suit. Not only is this data crucial to accurately estimating the virus’s threat, but it is also of practical value, since those with antibodies will be in far less danger either of catching or of spreading the disease. (In the movie Contagion, those with antibodies are given little bracelets and allowed to travel freely.)

The New York Times has created an interesting tool for roughly estimating the potential toll of the virus. By adjusting the infection and fatality rate, we can examine the likely death toll. Of course, these rough calculations are limited in that they make the mistake Keynes highlights above—they assume an independence of variables. For example, the calculator shows how the coronavirus would match up with expected cancer and heart disease deaths. But of course more coronavirus deaths would likely mean fewer deaths from other causes, since many who would have died from other causes would succumb to coronavirus. (Other causes of death like traffic accidents may also go down because of the lockdown.) The proper way to make a final estimate, I believe, would be to see how many total deaths we have had in a year, and then compare that total with what we would reasonably expect to have had without the coronavirus.

As you can see, the problem of coming up with a grand calculation is difficult in the extreme. Even if we can ultimately ascertain all of the information we need—medical, economic, sociological—we will still have only an imperfect grasp of the situation. Indeed, Keynes’s warning is quite pertinent here, since every factor will be influencing every other. Unemployment affects access to health care, an overwhelmed health care system will be less effective across the board, and the fear of the virus alone has economic consequences. This makes the ‘trolley problem’ model misleading, since there are no entirely independent tracks that the trolley can be moving on. Any decision will affect virtually everyone in many different ways; and this makes the arithmetical approach limited. 

Trump has said that the cure cannot be worse than the disease. Obviously, however, the decision is not a simple choice between economic and bodily well-being. This is what makes the decision so very subtle and complicated. Not only must we weigh sorts of damage in our ethical scales, but we also must be able to think synthetically about the whole society—the many ways in which its health and wealth are bound up together—in order to act appropriately.

Once again, I do not envy those who will have to make these choices.