Miles Kimball's Discussion of "When to Release the Lockdown: A Wellbeing Framework for Analysing Benefits and Costs," by Layard, Clark, De Neve, Krekel, Fancourt, Hey and O'Donnell

Richard Layard, Andrew Clark, Jan-Emmanuel De Neve, Christian Kekel, Daisy Fancourt, Nancy Hey and Gus O’Donnell are to be commended for doing a transparent cost-benefit analysis based on well-being of a key dimension of the policy choice we now: how long to continue the lockdowns. And Paul Frijters is to be commended for an excellent discussion of this paper, arguing that despite their bottom line that lockdowns should be ended soon, that Layard et al. are too kind to lockdowns. Let me offer some comments that might point out some aspects of the issue that might not be fully apparent from reading the Layard et al. paper and Paul Frijters’s discussion of it. I will not focus on their assumptions about what the consequences of different policies would be, except to say that their implicit epidemiological model seems too static when exponential dynamics are crucial. I will focus on the economics-of-happiness techniques used for evaluating the assumed consequences of policies.

The first issue to mention is that Layard et al. focus on life satisfaction, but there are many other dimensions of well-being. Indeed the UK Office of National Statistics now measures 4 aspects of well-being on a huge sample each month. No doubt the typical person would be willing to sacrifice some amount of being satisfied with their lives in order to gain in feeling that the things that they do in life are worthwhile. So life satisfaction isn’t everything. I could go on at greater length, but let me bring this up along the way where I can see how it might affect the calculations.

One issue I think is important methodologically—that could affect the calculations either way—is that Layard et al. implicitly assume that life satisfaction of 0 on a 0-10 scale is the level of life satisfaction at which—if that is what one would face form now on—one would be indifferent between living longer or dying immediately. This could be wrong in either direction. 0 is pretty low and most people report quite high numbers. Even a situation in which one reported a life satisfaction of 1 might be so intolerable that if doomed to that for the rest of one’s life, one might actively want to die. On the other hand, one might very much want to continue living even in a situation in which one barely reported a life satisfaction of zero even though one might wish to die in a situation where one had a latent desire to say -4 on the survey but was constrained to a nonnegative number. (Note also that some people who give a 0 for life satisfaction might, say, report a rating of 5 to “Overall, to what extent do you feel that the things you do in your life are worthwhile?” which is another question in the same UK Office for National Statistics survey.) Methodologically, the answer is that surveys need to be used to explicitly ask about tradeoffs between losing years of life and losing points of wellbeing. My understanding is that this is done a lot in the quality-adjusted life-years literature. The corresponding work needs to be done in the well-being adjusted life years literature.

A related methodological issue is that a point of life satisfaction might mean more near the bottom of the scale than at the top (or more at the top of the scale than near the bottom). Asking about tradeoffs with longevity given different starting levels of life satisfaction would be very useful. (The tradeoffs people are willing to make may also depend on the level of lifespan.)

Also, even once outcomes are measured in life-years of a benchmark level of well-being, just adding those life-years up may not be appropriate. One needs to decide how to deal with inequality. This is not addressed in the paper. Taking into account inequality probably makes the results more favorable to ending lockdowns if the economic harm is tilted more toward those of low well-being or low life-span than the health harm is, but makes the results more favorable to continuing lockdowns if the health harm is more toward those of low well-being or low-lifespan than the economic harm is.

Now, on to some considerations that, at least individually, have a clearer direction of effect on the bottom line. On how to value income, Layard et al. write:

There have been literally thousands of studies of the relation between wellbeing and income. They yield broadly similar results, which imply that a 1% gain in income increases wellbeing (measured 0-10) by around 0.002 points.

Layard et al. claim this number is uncontroversial, but Paul Frijters writes:

So they basically chose one of the lowest wellbeing effects of income one can find in the literature, which is the effect of hardly noticed changes in income on life satisfaction –  easily a factor 10 times less than normal for how people have been found to react to noticed losses of income …

Paul also makes another points in two different ways: government dollars can be used to save a lot of lives in ways that have nothing to do with the SARS-CoV2 and to increase well-being adjusted life years in other ways. Valuing all dollars or pounds at the best possible use the government could make of those dollars is probably valuing those dollars or pounds too highly. But valuing that fraction of extra dollars that would actually be extra tax revenue and actually be used by the government to do particular additional things at the value of those additional things seems fair. Even this latter is a big number, Paul argues.

On the valuation of dollars or pounds in life satisfaction or other well-being units, let me point out two relevant concerns.

First, dollars and pounds contribute to many, many other aspects of well-being that people care about besides life-satisfaction. Relative to many other aspects of well-being, dollars and pounds have a particularly low effect on life-satisfaction.

For what it’s worth, in data that Dan Benjamin, Kristen Cooper, Ori Heffetz and our team collected, we found a 1% increase in income raising a broader measure of well-being by .03 to .04 points on a 0-100 scale on which, to avoid top-coding and bottom-coding, we had urged people to use only 25-75, but allowed them to use the whole 0-100 range. That converts to somewhere between .003 and .008 on a 0-10 scale, depending on how you count the effects of our urging of people to use only 25-75.

Second, going the other direction, in this case, we are not talking about a pure loss of income: it is a loss of income associated with an increase in “leisure” broadly construed as the time away from paid work. Some of that extra time away from paid work is devoted to home production that yields reasonably close substitutes for market goods. And some of that extra time away from paid work is devoted to things like bonding time with family and (perhaps distant) friends for which there is no close substitute among market goods. The loss of income coupled with this increase in leisure shouldn’t be as bad for well-being as the loss of income from a lower wage with the same number of hours at work.

Combining the well-being value of a dollar or pound with the number of life-years at stake and the level of well-being at which one would be indifferent between living and dying implies a value of a statistical life or the closest thing to a value of a statistical life in a well-being-adjusted life years calculation. It is crucial to grasp philosophically with the differences between the numbers obtained in this way and the numbers obtained through other methods of getting a value of a statistical life. That is too big a topic for this post, but deserves a lot of thought and discussion.

Unemployment is awful, and an important part of the well-being calculation. But because it doesn’t seem as much like a personal failure, unemployment in this situation may not be as awful as unemployment usually is. Of course, the issue is not just unemployment now, but how long one will be unemployed. That depends a lot on monetary policy. (See and “Why We are Likely to Need Strong Aggregate Demand Stimulus after Tight Social Distancing Restrictions are Over” and “Narayana Kocherlakota Advocates Negative Interest Rates Now.”)

One contribution to the cost of lockdowns that Layard et al. give a relatively small number to is the cost of school closures. Here let me simply say that closing schools now doesn’t have to mean less school total. In the US at least, if this pandemic led to a policy shift toward somewhat more total months and hours per month of schooling, that policy improvement would be an important silver lining, and in any case could soon make up for less schooling or lower quality schooling right now. See “Magic Ingredient 1: More K-12 School.”

Conclusion: On the bottom line of whether lockdowns so far have been justified and whether it is justified to continue the lockdowns for somewhat longer, what I keep coming back to as the main justifications for the lockdowns is to give us time to figure out COVID-19 science. If in hindsight we consider the danger of the novel coronavirus to be low enough that we shouldn’t have done the lockdowns given that hindsight, that is great. It might not have been a mistake given our lack of knowledge at the time, which left open the possibility that the novel coronavirus was much worse. For example, the US seems to have a much greater prevalence of preexisting conditions that interact badly with COVID-19 than South Korea. Without substantial experience, we couldn’t depend on US mortality rates being much, much worse. Alternatively, maybe we needed time to figure out less costly partial lockdowns that could do much of the same job as more comprehensive lockdowns and the more comprehensive lockdown helps us figure out what dimensions of a lockdown are especially cost-effective and what are not so cost-effective. Or maybe some test-and-trace strategy will emerge that has a great benefit/cost profile, but requires that overall incidence not be too high, which it would have been with no lockdown.

Actually, figuring out and hopefully getting somewhat greater consensus on how to approach well-being cost-benefit analysis for our current situation is part of the COVID-19 science that we have bought time for with the lockdowns.

I honestly don’t know what the right policy is going forward. But unlike Paul Frijters, I don’t regret the lockdowns we have done so far as opposed to having done no relatively comprehensive lockdowns (though I regret many other aspects of our history of COVID-19 policy, especially the lack of an earlier response).

Don’t miss these other posts on the coronavirus pandemic:

Also, click on this link to see other posts tagged “happiness.”