PreprintPDF Available

Assessing the Wellbeing Impact of the COVID-19 Pandemic and Three Policy Types: Suppression, Control, and Uncontrolled Spread


Abstract and Figures

The COVID-19 crisis has forced a difficult trade-off between limiting the health impacts of the virus and maintaining economic activity. Welfare economics offers tools to conceptualize this trade-off so that policy-makers and the public can see clearly what is at stake. We review four such tools: the Value of Statistical Life (VSL); the Value of Statistical Life Years (VSLYs); Quality-Adjusted Life-Years (QALYs); and social welfare analysis, and argue that the latter are superior. We also discuss how to choose policies that differentially affect people’s wellbeing. We argue in favor of evaluating policies using a Social Welfare Function (SWF), which evaluates the possible distributions of wellbeing across individuals that may result from a policy. Such a function, we argue, should regard increases in the wellbeing of the less well-off as especially valuable. We then use a model to illustrate how such a framework can help evaluate two broad policy types in response to the pandemic: eradication of the virus, and more lenient control of the spread. Our model reveals how such evaluations depend on many empirical facts but also on key value judgments about the relative importance of health and on the extent of special concern for the worse off. The purpose of this brief is not to make precise recommendations, as conditions vary widely across countries and over time, but to provide a methodology.
Content may be subject to copyright.
Matthew Adler, Duke University
Richard Bradley, London School of Economics and Political Science
Maddalena Ferranna, Harvard School of Public Health
Marc Fleurbaey, Princeton University and Paris School of Economics
James Hammitt, Harvard University
Alex Voorhoeve, London School of Economics and Political Science
The COVID-19 crisis has forced a difficult trade-off between limiting the health impacts of the virus
and maintaining economic activity. Welfare economics offers tools to conceptualize this trade-off so
that policy-makers and the public can see clearly what is at stake. We review four such tools: the Value
of Statistical Life (VSL); the Value of Statistical Life Years (VSLYs); Quality-Adjusted Life-Years
(QALYs); and social welfare analysis, and argue that the latter are superior. We also discuss how to
choose policies that differentially affect people’s wellbeing. We argue in favor of evaluating policies
using a Social Welfare Function (SWF), which evaluates the possible distributions of wellbeing across
individuals that may result from a policy. Such a function, we argue, should regard increases in the
wellbeing of the less well-off as especially valuable. We then use a model to illustrate how such a
framework can help evaluate two broad policy types in response to the pandemic: eradication of the
virus, and more lenient control of the spread. Our model reveals how such evaluations depend on many
empirical facts but also on key value judgments about the relative importance of health and on the extent
of special concern for the worse off. The purpose of this brief is not to make precise recommendations,
as conditions vary widely across countries and over time, but to provide a methodology.
The Challenge
The COVID-19 crisis puts all governments in a difficult position. In the absence of extensive testing
capacities, they have to resort to near-universal lockdown and social distancing measures which exact a
severe economic toll. While developed countries have the ability to provide temporary support to avert
business collapse and worker hardship due to lack of liquidity, most developing countries do not. The
choice between lives and livelihoods is starker for the latter.
However, by relying on swift measures of quarantining travelers, testing, and labor-intensive contact
tracing before the number of infections rose to unmanageable levels, a few countries and states have
managed to ward off the first wave of the pandemic. And some, as in South-East Asia, managed to do
so even though their first cases appeared in January, before most other countries took serious measures.
They could also count on the cooperation of populations accustomed to public health campaigns and
protections against infectious diseases. Many developing countries have benefited from a longer
warning period, and might be able to emulate these strategies.
Even in countries where the virus has spread widely, it is technically possible to control the pandemic.
Lockdown measures observed in many countries reduce the reproduction rate of the pandemic to
numbers that would guarantee its local eradication in a few months if the lockdown were sustained. It is
also possible to keep the pandemic under control by a stop-and-go policy of periodic lockdown episodes
of a few weeks until a vaccine is found. Two pressing questions with regards to these policies of
suppression and control are: How should we conceptualize the benefits and burdens they produce as
compared to a policy of uncontrolled spread? And, once we have these benefits and burdens clearly in
view: how do we balance them in order to determine which policy is superior?
Our Proposal
We propose that governments rely on transparent evaluation methods in order to assess the wellbeing
impacts of the pandemic and of the policy response. There is considerable uncertainty around key
parameters of the pandemic as well as the reaction of the economy to exceptional measures.
Nevertheless, it is possible to determine an order of magnitude for the key elements of policy decisions.
No government can focus exclusively on epidemiological considerations or exclusively on economic
considerations; the wellbeing of a population depends on both health and wealth. Finding the right
balance requires relying on sound ethical principles and careful estimation of possible scenarios.
Obtaining the understanding and cooperation of the population, which may be crucial for successful
implementation, requires clarity about the objectives and the value judgments underlying the chosen
Evaluation Methods
How do we determine what economic cost is worth bearing in order to reduce the number of deaths and
other harms due to COVID-19? A well-known but often criticized method of policy evaluation is
benefit-cost analysis (BCA), which works by converting all the effects of a policy into monetary
equivalents and then summing up these equivalents. A different method, social welfare analysis,
proceeds by determining the effects of a policy on individual wellbeing and then applying an aggregation
formula to them to evaluate the overall effects of a policy. We briefly survey these methods and
emphasize the advantages of the latter.
The monetary measure of the value of saving lives most widely used in benefit-cost analyses is that of
the Value of a Statistical Life (VSL). When the individuals whose lives would be saved are not known,
each of the people at risk would be willing to pay some money to increase their chance of survival. VSLs
describe the monetary values that individuals attribute to a reduction in their own mortality risk. To be
precise, these monetary values are derived from the rate at which people are willing to trade off small
changes in their income against small changes in their risk of death. This, in turn, is estimated from
individuals’ reported preferences or from those that they reveal in workplace and consumption behavior,
such as the choices they make amongst jobs involving different levels of risk or their purchases of risk-
reducing equipment. For example, if someone would accept a pay cut of $1,000 per year to reduce their
annual risk of mortality by 0.1% (but would not accept a larger pay cut), then we say that the monetary
value of their statistical life is $1,000,000. Note that this is not the same as saying that they would be
willing to accept $1,000,000 in return for certain death or would pay this amount to guarantee their
survival. Rather, it means that each of 1,000 people, identical in all relevant ways, would, considering
their self-interest alone, be willing to pay an equal share of a $1,000,000 cost for something that reduces
the number of them expected to die in the year by one.
A person’s VSL can depend on their characteristics such as their age, their income and wealth, or the
overall level of risk they face. This can have unacceptable consequences for BCA. In particular, the fact
someone who is well-off is likely to place a higher monetary value on risk reduction than someone who
is less well-off implies that if individual-specific VSLs are used in a benefit-cost analysis of policies,
the interests of the well-off will count for more than those of the less well-off (because the monetary
value they place on reducing their risk of death is higher).
By using a single VSL, such as the population average, rather than individual-specific ones, this problem
is avoided. But others are then created. In particular, it seems reasonable to treat people in different age-
groups differently when assessing policies. Death is certain for all mortals and is commonly considered
a more serious loss from the societal and ethical perspective when it occurs earlier in life. Reasoning in
terms of life years preserved rather than lives saved appears to better take account of this widespread
sentiment. The skewed age distribution of COVID-19 fatalities makes this problem especially pressing.
A common solution is to use a different measure for policy evaluation: the Value of a Statistical Life
Year (VSLY). The VSLY is obtained by dividing the average VSL of the population by the average life
expectancy remaining (an individual’s current life expectancy remaining is the number of years she can
be expected to live, if she doesn’t die now). The value of saving the life of someone in any particular
age cohort is then given by the product of the VSLY and the life expectancy remaining for the cohort.
This yields a value of life saving that varies by age.
One criticism of both VSL and VSLY is that they do not take into account quality of life. Many people
would not regard a year of life spent bed-ridden as equivalent to a year of life in excellent health, for
instance (but see below). Quality Adjusted Life Years (QALYs) are a way of allowing the value
attributed to a life saved to depend on both its remaining length and its quality. The value of living in an
impaired health statesay, with diminished lung function due to COVID-19—is derived from people’s
preferences. These preferences may be elicited in a number of ways. Individuals may simply be asked
to assign a numerical value to life in a particular health state in comparison to both death and life in full
health. Alternatively, they may be asked how they would balance a longer life in an impaired state
against a shorter, healthier life, or asked what risk of death they would be willing to run in order to be
fully cured of their impaired health. These preference-based assessments can be questioned (Hausman
2015). For example, there is evidence that healthy people are poor predictors of what life would be like
in states of impaired health (Dolan and Kahneman 2008, Walasek et al. 2019). Nonetheless, even rough
indicators of the quality of life in impaired health states can be better than measures that neglect quality
altogether. This is particularly clear in the pandemic, in which it is important to take into account the
often substantial effect of contracting the illness on those who do not die from it. For this reason, the
use of QALYs in public health decision making is widespread.
It is also controversial. One key concern is that when it comes to life extension, the use of QALYs
regards as more valuable the life years gained by people who would, if saved, be in good health than the
life years gained by people who would, if saved, live with disabilities or in poor health because extending
the lives of the former would generate higher health-related quality of life (National Council on
Disability 2019). In our view, this objection is best addressed not by rejecting the use of QALYs, but by
assigning special value to improvements in the quality (and length) of life of those who are worse off
(John, Millum, and Wasserman 2017).
Estimates of the VSL and VSLY vary considerably between countries. Part of this variation is due to
differences in income per capita. For example, Robinson et al. (2019) recommend calculating the
average VSL for a country by multiplying the country’s per capita income by a factor proportional to
the square root of per capita income (but no smaller than 20), and a VSLY by dividing the VSL by the
average remaining years of life. For the USA, the typical VSL is around $10,000,000, and the VSLY a
little over $300,000 (see also Kniesner and Viscusi forthcoming); for a country with per capita income
of $10,000, a VSL of around $670,000 and a VSLY of around $20,000 (international dollars) can be
A similar variation is observed in the monetary costs that public actors regard as reasonable to incur to
gain one QALY. One approach is to estimate the value of a QALY by dividing the VSL by the average
remaining QALYs (Hirth et al. 2000). This produces values modestly larger than the VSLY. By contrast,
the World Health Organization (WHO) has suggested that interventions that generate a QALY for less
than 1 times per capita income are good value for money and that interventions that generate a QALY
for up to 3 times per capita income may be worth the cost (Bertram et al. 2016). In line with this formula,
for the USA, the Institute for Clinical and Economic Review suggests values between $100,000 and
$150,000 for one QALY (ICER 2018). In contrast, the British National Institute for Health and Care
Excellence applies figures in the $25,000 to $40,000 range for one QALY (NICE 2008). The fact that
these countries have different per capita incomes only partially explains these differences. As we show
below, the ranking of policies to deal with the pandemic based on benefit-cost analysis may well depend
on which values are adopted. It is therefore critical that attention be given to the justification of any
particular choice.
The alternative approach we emphasize here, social welfare analysis, proceeds by measuring the joint
health and economic impact of policies on individual wellbeing and then aggregating individual
wellbeing gains and losses to yield an overall measure of how beneficial a policy is. This method has a
singular advantage over BCA. Unlike population-average VSLs, the individual-specific wellbeing
values that social welfare analysis uses are sensitive to individualscharacteristics, such as their age and
income. And while the individual-specific VSLs, VSLYs and values of QALYs of the well-off are
inflated, relative to those of less well-off individuals, by the fact that money has relatively lower
marginal value for the well-off, this is not true of wellbeing values. So, the aforementioned bias in favor
of the well-off that the use of individual-specific monetary values introduces into BCA does not plague
social welfare analysis.
Social welfare analysis begins with a measure of the wellbeing levels associated with different possible
lives, represented by bundles of the goods that matter to individuals: income, health, longevity, and so
on. Many methods are used to obtain these measures (for a review, see Adler and Fleurbaey 2016). One
draws on reported levels of subjective wellbeing or life-time satisfaction scores to identify the
determinants of wellbeing (Clarke et al. 2018). Another derives a wellbeing measure from individuals’
preferences between probability distributions (lotteries) over alternative possible lives (Adler 2019). Yet
another relies on income, corrected for the value of non-market aspects of life, such as longevity, on the
basis of population preferences over these aspects (Blanchet and Fleurbaey 2013). Finally, the capability
approach measures opportunities in various aspects of life (Sen 1999).
The effect of a policy that reduces the risk of contracting COVID-19 by some percentage and at some
financial cost can be modelled by the shift in the population distribution of wellbeing levels generated
by the bundles of longevity, health, and income mentioned above. Since this shift captures not just the
impact on individuals’ longevity and health of the policy but also how these factors, together with
income, co-determine changes in individuals’ wellbeing, it provides all the information required for a
comprehensive analysis of the overall effect of implementing the policy.
To determine this overall effect, social welfare analysis proceeds by aggregating the set of individual
wellbeing values achieved by implementing the policy. It does so by means of a social welfare function
(SWF) that assigns to each distribution of individual wellbeing a measure of social value. A commonly
used SWF is the utilitarian one, which assigns to each set of individual wellbeing values the total (or the
average) of the values. This way of aggregating individuals’ wellbeing is insensitive to inequalities of
wellbeing in the population—in other words, it is indifferent to whether a given increment in wellbeing
accrues to a well off person or a badly off person. In this respect, it is in tension with the common
conception that a given improvement in the wellbeing of the worse off matters more than the same
improvement to the well off, because it comes to those who are in greatest need or because such
improvements reduce inequality (Adler 2019; Voorhoeve 2019). This problem can be addressed by
using distribution-sensitive SWFs that prefer policies that produce wellbeing gains for those with low
wellbeing over policies that produce the same wellbeing gains for those who are better off.
The choice of SWF is fundamentally an ethical one, as it requires balancing the wellbeing interests of
different individuals. This is a second important advantage of social welfare analysis: it allows for these
ethical choices to be made more explicitly and transparently than with BCA, which largely ignores the
distribution of benefits and burdens among individuals.
Reviewing Policy Options: An Illustration
To illustrate these approaches, a model simulating the pandemic as well as lockdown and testing policies
has been developed and adapted to several countries (USA, UK, France, Belgium, and Guinea).1 The
model takes account of inequalities in income and life expectancy across social groups, and allows for
various assumptions about the distribution of the economic cost and the fatality burden among these
groups. Such assumptions relate to policy choices about social protection and income support, as well
as access to health care. This model is not designed to make predictions or precise policy
recommendations, but rather to highlight the parameters relevant to sound decision making.
The illustration here focuses on the comparison between an suppression policy which implements a
lockdown order (“contact reduction” of a certain percentage) for as long as it takes to make the virus
almost completely disappear, on the one hand, and a control policy which limits the lockdown to periods
following weeks where fatalities are over a certain threshold. The latter policy may also eventually cause
the virus to nearly completely disappear, but over a longer period of time and with a succession of
shorter lockdown episodes rather than a single one. Supplementary policies (testing, mask-wearing,
income support) can also be taken into account and will modify the evaluation. The simulations we
present include testing and mask-wearing through the “contagion reduction” that they entail. It is
assumed that only contact reduction has a substantial economic cost, because testing and producing
protective equipment are relatively low cost and do not slow down economic activity—even if
displacing resources toward them has an opportunity cost. In this section, the discovery of a treatment
or a vaccine is assumed to come too late to affect the evaluation of these policies. How relaxing this
assumption would affect the analysis is briefly described at the end.
Fig. 1a illustrates the policy problem. The model starts with a single infection in week 1, which then
spreads over the population. The policy (suppression or control) starts on a particular week (from week
10 to week 23) and is continued until the pandemic virtually disappears. Either policy entails an
economic cost (due to lockdown)2 but also saves lives, compared to the absence of intervention. The
graphs do not show the outcomes over time, but the final outcome (economic cost and fatalities) as a
function of when the policy is initiated. When the policy starts on week 23, it is almost as if no
1 The model is contained in an Excel spreadsheet and can be downloaded freely from Users can change all parameters and assumptions and
determine the timing and intensity of contact reduction and testing policies. Refined versions of the model (in
Python code) are available upon request.
2 In these simulations, it is assumed that contact reduction by 70% reduces economic income during the
lockdown period by 35% (half the percentage).
intervention took place, because the first wave of the pandemic has almost fully passed. Therefore, our
analysis also covers the case of uncontrolled spread. The scissors pattern for both policies shows that an
early policy saves more lives (though a very early control policy is counterproductive in this respect
because it hinders the build-up of collective immunity) but has greater economic cost because it requires
longer lockdown episodes.3
Figure 1: The policy problem
(a) (b)
How to read the graphs: solid curves describe the outcomes of the suppression policy, dashed curves the control
policy; economic cost in red (left axis), total fatalities in blue (right axis). Note that the left axis has a different
scale in (a) and (b).
The contrast between Fig. 1a and 1b shows the complementarity of contagion reduction with contact
reduction. In the absence of contagion reduction (1a), the control policy is less costly but it also saves
fewer lives. With ambitious contagion reduction (1b), in contrast, control is dominated by suppression,
because the latter saves more lives at a smaller cost. This occurs because thanks to contagion reduction,
the suppression of the virus can be achieved in a shorter time.
The benefit-cost analysis can be illustrated by the contrast between the VSL and the VSLY approaches,
in the COVID-19 context where the victims are mostly the elderly (at least in developed countries). This
is shown on Fig. 2, which displays hypothetical simulations for the total cost (adding up the value of
lives lost to the crisis and the economic cost).
3 As shown on the graph, a very early suppression policy can have a smaller cost, though, because a quick
elimination of the virus is then possible before it spreads widely over the population.
Figure 2: The VSL and VSLY approaches
(a) (b) (c)
How to read the graphs: The vertical axis displays losses as negative numbers so that the higher on the axis,
the better. This makes the graphs directly comparable to Fig. 3 below.
In Fig. 2a-b, both approaches favor the eradication policy over the control policy, the VSL approach is
favorable to early eradication whereas the VSLY approach would also condone letting the virus spread
before eradication is implemented. This is because elderly victims of the virus do not lose many years
of life. These two graphs rely on a VSL equal to 150 times the GDP per capita and a VSLY equal to 3
times the GDP per capita. Fig. 2c relies on a lower VSLY equal to the GDP per capita, and there one
sees that, in spite of the staggering death toll, delaying policy adoption is acceptable and that very late
adoption of either policy is optimal. This illustrates how crucial the ethical parameter for the value of
life is.
The SWF approach is also implemented in the model. Individual wellbeing is computed for a whole life,
not just for a year, and depends on income and longevity. It is calibrated in a way that guarantees that
the willingness to pay of the average individual for a life year is equal to the same VSLY used in the
BCA method (3 times GDP per capita). In this way, in the absence of priority for the worse off, the
VSLY approach and the social wellbeing approach deliver similar assessments. But when a degree of
priority for the worse off is introduced, the evaluations can differ markedly. The social wellbeing
approach is then uniquely sensitive to three considerations. First, given the assumptions about the value
of longevity, the worse off include the victims of the virus, because their loss of longevity is a very
substantial wellbeing loss. Concern for the worse off therefore assigns greater significance to health
outcomes than economic outcomes. However, there are several elements to take into account. Many of
the victims have attained an old age, so that, compared to younger survivors, they have the advantage
of having avoided other fatality risks to reach that age, but also the disadvantage of having lived earlier,
in less affluent economic times. The very worst off tend to be among the middle-aged victims, whose
premature deaths are great losses.4
Second, inequalities in life expectancy and fatality rates across social groups reinforce the concern for
health, because the worse off in income incur a double penalty through a greater health toll. Unlike
4 In the version of the model with which the figures have been produced, these considerations are incorporated in
a simplified way. The victims are assumed to have had greater life expectancy than the average level in the
population (because most of them have already reached an old age), whereas the (younger survivors) are
assumed to face the prospect of a greater lifetime income. But no detailed depiction of inequalities in income and
fatality rates among age groups is made. A refined version of the model in construction will have this feature.
BCA, the SWF approach thus clearly identifies the value of policies which give greater access to health
care and reduce the correlation between health and income.
Third, inequalities in the economic cost of lockdowns may attenuate the previous considerations if the
disadvantaged social groups are more severely affected by the economic slowdown.5 This is especially
relevant for poor countries in which the most disadvantaged may fall into extreme poverty under these
circumstances. When priority is given to the worse off, the SWF approach favors strong social protection
measures that ensure a more equitable distribution of the economic cost. This point is illustrated in Fig.
3. It displays one particular example of a measure of societal wellbeing, the equivalent income, defined
as the level of income per capita which, equally distributed and associated with equal longevity for
everyone, would yield the same societal wellbeing as the contemplated (unequal) situation.
Figure 3: The SWF approach with priority to the worse off
(a) (b)
How to read these graphs: The vertical axis is measured in equivalent income per capita which reflects
societal wellbeing, taking account of inequalities. As in Fig. 2, the higher the better.
In Fig. 3, the elasticity of the economic cost to income is either 0.5 (Fig. 3a), meaning that the
distribution of the cost is regressive, or 1.5 (Fig. 3b), meaning that the distribution is progressive.6 The
priority for the worse off is substantial in Fig. 3, and implies that improving the wellbeing of an
individual who is half as rich is four times as important for the social evaluation.
Fig. 3a illustrates the fact that if the lower income groups in the population bear a disproportionate
economic burden, and if this economic burden is heavy due to a lack of contagion reduction, then late
may be preferred to mid-time adoption of a suppression policy, because the economic cost is substantial
when the policy starts in the middle of the time range. And late adoption of a control policy is preferred
to any other time. In contrast, Fig. 3b illustrates the situation in which, with a progressive distribution
of the economic, early policy adoption is preferred when eradication is considered, and as far as control
policy is concerned, very late policy adoption is not optimal at all.
5 This particular model does not include the long-term and indirect effects on people’s health and other long-term
outcomes, but it does include the additional deaths not due to the virus but caused by the disruption of health
care (either on the supply side or on the demand side, when patients for other conditions shun health care
facilities out of fear).
6 Inequalities of income are represented by distinguishing quintile groups. An elasticity of 0.5 means that a group
with an income level greater by 1% than another group bears a share of the economic cost that is greater by 0.5%
(therefore, the cost as a fraction of its baseline income is half a percentage point lower).
There are many parameters on which these simulations depend, but which cannot be discussed here in
detail. For instance, if contact reduction can be raised from 70% to 80%, early eradication is much
quicker and much less costly and does appear preferable even if one adopts a low figure for the value of
life, keeping the other parameters as in Fig. 2c. The model incorporates spontaneous contact reduction
by individuals when they witness a peak of mortality, or additional fatalities when hospitals are
overwhelmed. The results would also change with an early discovery of a treatment, which would lower
the fatality rate and enhance the relative value of controlling the pandemic compared to full suppression.
The demographic structure of the population and background health conditions may affect the lethality
of the virus; in higher-income countries, deaths are concentrated in the elderly but it appears that in
developing countries, worse background health makes younger patients more vulnerable than in rich
countries. There are important missing elements in these simulations. In particular, they do not assume
any spatial heterogeneity between regions in a country, they ignore possible contamination by travelers
coming from abroad, and they do not finely distinguish the situations of different age groups. They also
ignore possible longer-term economic and health consequences of the crisis.
The upshot of these remarks is that precise prediction with this, or any other model, is not possible at
the moment. The point of this policy brief is primarily to propose this methodology as a framework
within which the combined effect of different values for relevant empirical and ethical parameters on
policy conclusions can be studied. When available, more precise predictions can be introduced in this
framework to support decision-making that is both scientifically and ethically well-informed.
Our model shows how sensitive the path of the pandemic is to the various parameters, reflecting how
much uncertainty there is, which makes decisions quite hard for policy-makers. What is especially
difficult, from a policy point of view, is the following. Looking at the figures presented here, it is
tempting to conclude that near-total suppression of the virus is, in most cases, the dominant policy. But
the duration of the required lockdown period can stretch over 4 months at 70% contact reduction (3
months at 80% contact reduction), if the start date is around a time corresponding to week 15-16 in our
figures, as seems to be the case in many countries.7 This means that, even when public health authorities
are convinced that eradication is the best policy, implementing this policy may be politically difficult
and requires very strong support measures for the population suddenly deprived of work and income. If,
in the middle of the way towards near-eradication, the authorities revert to a more modest control
strategy, then much of the effort has been in vain, because they have only pushed the infection wave
into the future (see also Gollier 2020 and Kissler et al. 2020). This is why a clear communication on the
strategy and the ethical choices, based on rigorous modelling as proposed here, may be crucial to
convince the population of the need to stay the course.
In conclusion, although the current crisis presents a difficult trade-off between lives and livelihoods,
especially for governments in fragile states or with frail leadership and a low degree of cooperation in
the population, it is possible to lay out the main considerations that should guide policy, including the
key normative issues about valuing lives and giving priority to the worse off. Our quantitative analysis
illustrates how the relevant empirical and normative elements of sound policy-making can be put
together into a rigorous framework, and why the SWF approach which takes account of the distribution
of impacts and of background inequalities is more attractive than the BCA approach and, in particular,
more consistent with widely shared ethical views regarding life-saving by income and age.
7 Most countries which have been successful in eradicating the virus (such as Vietnam, New Zealand, or South
Korea) have adopted very early lockdown and testing policies and have endured little damage compared to the
other countries. China appears to have implemented a very strict lockdown policy (perhaps more than 90%
contact reduction in the Wuhan area).
As illustrated here, modelling costs and health impacts may reveal dominant policies which would not
have been obvious otherwise. In particular, it is crucial to check if the less politically difficult (in the
short run) strategy may turn out to be worse, all things considered. However, there are many aspects to
policy that cannot be resolved by dominance considerations, and this is where weighing the relative
importance of fatalities and economic costs is required.
Adler, Matthew D. (2019). Measuring Social Welfare: An Introduction. Oxford University Press.
Adler, Matthew D. and Ferranna, Maddalena and Hammitt, James K. and Treich, Nicolas (2019) “Fair
Innings: The Utilitarian and Prioritarian Value of Risk Reduction over a Whole Lifetime” (November
16, 2019). Duke Law School Public Law & Legal Theory Series No. 2019-79.
Adler, Matthew D. and Fleurbaey, Marc (2016). Oxford Handbook of Wellbeing and Public Policy.
Oxford University Press.
Adler, Matthew D., Hammitt, James K. and Treich, Nicolas (2014). “The Social Value of Mortality Risk
Reduction: VSL versus the Social Welfare Function Approach.” Journal of Health Economics 35: 82-
Bertram, Melanie, Lauer, Jeremy, De Joncheere, Kees, Edejer, Tessa, Hutubessy, Raymond, Kieny,
Marie-Paule, Hill, Suzanne (2016).CostEffectiveness Thresholds: Pros and Cons.” Bulletin of the
World Health Organization 94:925-930. doi:
Clark, Andrew, Flèche, Sarah, Layard, Richard, Powdthavee, Nattavudh, and Ward, George (2018). The
Origins of Happiness: The Science of Wellbeing over the Life-Course. Princeton NJ: Princeton
University Press.
Dolan, Paul and Kahneman, Daniel (2008) “Interpretations of Wellbeing and Their Implications for the
Valuation of Health.” The Economic Journal, Vol. 118(525): 215-234.
Gollier, Christian (2020). “Cost-benefit analysis of age‑specific deconfinement strategies,” Covid
Economics 24: 1-31.
Hirth Richard, Chernew Michael, Miller Edward, Fendrick Mark, Weissert William. Willingness to pay
for a quality-adjusted life year: in search of a standard. Med Decis Making. 2000;20(3):332‐342.
ICER (Institute for Clinical and Economic Review) (2018). “ICER Value Framework”. https://icer- [accessed 23 May
John, Tyler, Millum, Joseph, and Wasserman, David (2017). “How to Allocate Scarce Health Resources
without Discriminating against People with Disabilities,” Economics and Philosophy 33 (2): 161-186
Kissler, Stephen, Tedijanto, Christine, Goldstein, Edward, Grad, Yonathan, and Lipsitch, Marc (2020).
“Projecting the Transmission Dynamics of SARS-CoV-2 through the Postpandemic Period. Science 368
(6493): 860-868. DOI: 10.1126/science.abb5793
Kniesner, Thomas J. and Viscusi, W. Kip (forthcoming) “The Value of a Statistical LifeOxford
Research Encyclopedia of Economics and Finance. Vanderbilt Law Research Paper No. 19-15.
Layard, R., A. Clark, J.-E. De Neve, C. Krekel, D. Fancourt, N. Hey, and G. O’Donnell, “When to
release the lockdown: A wellbeing framework for analysing costs and benefits,” CEP Occasional Paper,
49, 2020.
Love-Koh J., M. Asaria, R. Cookson, S. Griffin (2015). The Social Distribution of Health: Estimating
Quality-Adjusted Life Expectancy in England. Value in Health 18: 655-662.
NICE (National Institute for Health and Care Excellence). (2008). Social Value Judgments, Principles
for the development of NICE guidance 2nd edition.
principles [accessed 23 May 2020].
Robinson, L., Hammitt, J., & O’Keeffe, L. (2019). Valuing Mortality Risk Reductions in Global Benefit-
Cost Analysis. Journal of Benefit-Cost Analysis, 10(S1), 15-50. doi:10.1017/bca.2018.26
Salje H., C. Tran Kiem, N. Lefrancq, N. Courtejoie, P. Bosetti et al.. 2020, Estimating the burden of
SARS-CoV-2 in France. pasteur-02548181.
Schelling, T.C., “The Life You Save May Be Your Own,” in S.B. Chase (ed.), Problems in Public
Expenditure Analysis, Brookings, Washington, D.C., 1968.
Sen, Amartya.K. (1999). Commodities and Capabilities. Oxford University Press.
Verity R., L. C. Okell, I. Dorigatti, P. Winskill et al. (2020). Estimates of the severity of coronavirus
disease 2019: a model-based analysis, The Lancet.
Voorhoeve, Alex (2019). ‘Why Health-Related Inequalities Matter and Which Ones Do.’ In. O. F.
Norheim, E. Emanuel and J. Millum (eds.) Global Health Priority Setting: Beyond Cost-Effectiveness.
Oxford University Press, pp. 145-62.
Walasek, Lukasz, Brown, Gordon D. A. and Ovens, Gordon D. (2019). “Subjective Wellbeing and
Valuation of Future Health States: Discrepancies between Anticipated and Experienced Life
Satisfaction.” Journal of Applied Social Psychology 49 (12): 746-754.
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
France has been heavily affected by the SARS-CoV-2 epidemic and went into lockdown on the 17 March 2020. Using models applied to hospital and death data, we estimate the impact of the lockdown and current population immunity. We find 3.6% of infected individuals are hospitalized and 0.7% die, ranging from 0.001% in those <20 years of age (ya) to 10.1% in those >80ya. Across all ages, men are more likely to be hospitalized, enter intensive care, and die than women. The lockdown reduced the reproductive number from 2.90 to 0.67 (77% reduction). By 11 May 2020, when interventions are scheduled to be eased, we project 2.8 million (range: 1.8–4.7) people, or 4.4% (range: 2.8–7.2) of the population, will have been infected. Population immunity appears insufficient to avoid a second wave if all control measures are released at the end of the lockdown.
Full-text available
It is urgent to understand the future of severe acute respiratory syndrome–coronavirus 2 (SARS-CoV-2) transmission. We used estimates of seasonality, immunity, and cross-immunity for betacoronaviruses OC43 and HKU1 from time series data from the USA to inform a model of SARS-CoV-2 transmission. We projected that recurrent wintertime outbreaks of SARS-CoV-2 will probably occur after the initial, most severe pandemic wave. Absent other interventions, a key metric for the success of social distancing is whether critical care capacities are exceeded. To avoid this, prolonged or intermittent social distancing may be necessary into 2022. Additional interventions, including expanded critical care capacity and an effective therapeutic, would improve the success of intermittent distancing and hasten the acquisition of herd immunity. Longitudinal serological studies are urgently needed to determine the extent and duration of immunity to SARS-CoV-2. Even in the event of apparent elimination, SARS-CoV-2 surveillance should be maintained since a resurgence in contagion could be possible as late as 2024.
Full-text available
Background In the face of rapidly changing data, a range of case fatality ratio estimates for coronavirus disease 2019 (COVID-19) have been produced that differ substantially in magnitude. We aimed to provide robust estimates, accounting for censoring and ascertainment biases. Methods We collected individual-case data for patients who died from COVID-19 in Hubei, mainland China (reported by national and provincial health commissions to Feb 8, 2020), and for cases outside of mainland China (from government or ministry of health websites and media reports for 37 countries, as well as Hong Kong and Macau, until Feb 25, 2020). These individual-case data were used to estimate the time between onset of symptoms and outcome (death or discharge from hospital). We next obtained age-stratified estimates of the case fatality ratio by relating the aggregate distribution of cases to the observed cumulative deaths in China, assuming a constant attack rate by age and adjusting for demography and age-based and location-based under-ascertainment. We also estimated the case fatality ratio from individual line-list data on 1334 cases identified outside of mainland China. Using data on the prevalence of PCR-confirmed cases in international residents repatriated from China, we obtained age-stratified estimates of the infection fatality ratio. Furthermore, data on age-stratified severity in a subset of 3665 cases from China were used to estimate the proportion of infected individuals who are likely to require hospitalisation. Findings Using data on 24 deaths that occurred in mainland China and 165 recoveries outside of China, we estimated the mean duration from onset of symptoms to death to be 17·8 days (95% credible interval [CrI] 16·9–19·2) and to hospital discharge to be 24·7 days (22·9–28·1). In all laboratory confirmed and clinically diagnosed cases from mainland China (n=70 117), we estimated a crude case fatality ratio (adjusted for censoring) of 3·67% (95% CrI 3·56–3·80). However, after further adjusting for demography and under-ascertainment, we obtained a best estimate of the case fatality ratio in China of 1·38% (1·23–1·53), with substantially higher ratios in older age groups (0·32% [0·27–0·38] in those aged <60 years vs 6·4% [5·7–7·2] in those aged ≥60 years), up to 13·4% (11·2–15·9) in those aged 80 years or older. Estimates of case fatality ratio from international cases stratified by age were consistent with those from China (parametric estimate 1·4% [0·4–3·5] in those aged <60 years [n=360] and 4·5% [1·8–11·1] in those aged ≥60 years [n=151]). Our estimated overall infection fatality ratio for China was 0·66% (0·39–1·33), with an increasing profile with age. Similarly, estimates of the proportion of infected individuals likely to be hospitalised increased with age up to a maximum of 18·4% (11·0–7·6) in those aged 80 years or older. Interpretation These early estimates give an indication of the fatality ratio across the spectrum of COVID-19 disease and show a strong age gradient in risk of death. Funding UK Medical Research Council.
Full-text available
One way of informing health policy decisions is to ask people about the impact that different health states would have on their future subjective well‐being. The present research explored the relation between anticipated and experienced changes in health‐related subjective well‐being, and examined whether affective forecasting errors could be reduced by psychological distancing manipulations. Using survey methodology, we tested whether people can accurately estimate the impact of different possible health states on their subjective well‐being. We also manipulated psychological distance: Forecasts were made about present self, future self, or others. Based on construal level theory and past work on affective forecasting errors, our prediction was that increasing psychological distance may reduce the mismatch between anticipated and experienced impact of health states on subjective well‐being. We found that the impact of ill health on subjective well‐being was greatly overpredicted and that this overprediction was not eliminated when participants were asked to make predictions about themselves in the future or about other people. Consistent with past work, we found that our participants correctly expected that their subjective well‐being would deteriorate more if they experienced the highest levels of mental illness as compared to the highest intensities of pain or most severe limitations to physical functioning.
Full-text available
The author outlines and defends two egalitarian theories, which yield distinctive and complementary answers to why health-related inequalities matter. The first is a brute luck egalitarian view, according to which inequalities due to unchosen, differential luck are bad because unfair. The second is a social egalitarian view, according to which inequalities are bad when and because they undermine people’s status as equal citizens. These views identify different objects of egalitarian concern: the brute luck egalitarian view directs attention to health-related well-being, while social egalitarianism focuses on health-related capabilities that are central to a person’s status as a citizen. The author argues that both views are correct and should jointly guide priority-setting in health.
I calibrate a Multiple‐Risk Susceptible–Infected–Recovered model on the covid pandemic to analyze the impact of the age‐specific confinement and polymerase chain reaction (PCR) testing policies on incomes and mortality. Two polar strategies emerge as potentially optimal. The suppression policy would crush the curve by confining 90% of the population for 4 months to eradicate the virus. The flatten‐the‐curve policy would reduce the confinement to 30% of the population for 5 months, followed by almost 1 year of free circulation of the virus to attain herd immunity without overwhelming hospitals. Both strategies yield a total cost of around 15% of annual gross domestic product (GDP) when combining the economic cost of confinement with the value of lives lost. I show that hesitating between the two strategies can have a huge societal cost, in particular if the suppression policy is stopped too early. Because seniors are much more vulnerable, a simple recommendation emerges to shelter them as one deconfines young and middle‐aged people to build our collective herd immunity. By doing so, one reduces the death toll of the pandemic together with the economic cost of the confinement, and the total cost is divided by a factor 2. I also show that expanding the mass testing capacity to screen people sent back to work has a large benefit under various scenarios. This analysis is highly dependent upon deeply uncertain epidemiologic, sociological, economic, and ethical parameters.
The social value of risk reduction (SVRR) is the marginal social value of reducing an individual’s fatality risk, as measured by some social welfare function (SWF). This Article investigates SVRR, using a lifetime utility model in which individuals are differentiated by age, lifetime income profile, and lifetime risk profile. We consider both the utilitarian SWF and a “prioritarian” SWF, which applies a strictly increasing and strictly concave transformation to individual utility. We show that the prioritarian SVRR provides a rigorous basis in economic theory for the “fair innings” concept, proposed in the public health literature: as between an older individual and a similarly situated younger individual (one with the same income and risk profile), a risk reduction for the younger individual is accorded greater social weight even if the gains to expected lifetime utility are equal. The comparative statics of prioritarian and utilitarian SVRRs with respect to age, and to (past, present, and future) income and baseline survival probability, are significantly different from the conventional value per statistical life (VSL). Our empirical simulation based upon the U.S. population survival curve and income distribution shows that prioritarian SVRRs with a moderate degree of concavity in the transformation function conform to widely held views regarding lifesaving policies: the young should take priority but income should make no difference.
The social welfare function (SWF) framework is a powerful tool for evaluating governmental policies in light of human well-being. The framework originates in theoretical welfare economics and is widely used in contemporary economic scholarship, although not (yet) in governmental practice. This book is intended to provide an accessible, yet reasonably rigorous overview of the SWF approach. The framework has three components: an interpersonally comparable measure of well-being, which functions to translate outcomes into lists (“vectors”) of well-being numbers, one for each person in the population; a rule (the SWF) for ranking well-being vectors, such as the utilitarian SWF (which simply adds up well-being numbers), a continuous-prioritarian SWF (which gives greater weight to the worse off), or some other; and a procedure for ranking policies, understood as probability distributions across outcomes. Each component of the SWF framework is reviewed in detail; in doing so, the book engages both the economic literature on SWFs and philosophical scholarship regarding individual well-being, ethics, and distributive justice. The book also clarifies the difference between the SWF approach and cost-benefit analysis (CBA), which uses money rather than an interpersonally well-being measure as the scale for quantifying policy impacts. The book includes a detailed case study of risk regulation—illustrating how the SWF framework can be used in practice and how it contrasts with CBA. The book is written to be accessible to readers without much mathematical training, but is backed up by an extensive mathematical appendix.