Prioritarianism in Pandemic Policy

Health vs. Economy in the Pandemic Cornell Program on Ethics and Public Life Webinar October 1, 2020 Matthew D. Adler, Duke University

Health vs. Economy in Pandemic Control: What is the Right Balance? My scholarship focuses on prioritarianism: both the philosophical theory, and its application to various policy domains (climate policy, risk regulation). Adler, Measuring Social Welfare (2019) In this webinar, I ll discuss how prioritarianism can be used to make tradeoffs between lifesaving and economic impacts in setting Covid-19 policy. A disclaimer! I am not an epidemiologist, and am not making specific recommendations here about social distancing, masking, shutdowns, or other aspects of Covid-19 policy. Rather, I ll show how prioritarianism provides a systematic framework for doing the difficult balancing involved in setting such policies. For simplicity, I ll focus on tradeoffs between lifesaving and income, although Covid-19 policy has other impacts too (e.g., nonfatal health impacts). First, I ll (briefly!) summarize prioritarianism. I ll then illustrate the application of prioritarianism to Covid-19 policy via a simulation model (see Adler, What Should We Spend to Save Lives in a Pandemic? CovidEconomics [2020]).

Prioritarianism Prioritarianism is a version of welfarism. Welfarism: the ethical assessment of choices depends upon the possible outcomes of those choices; and the ethical comparison of outcomes in turn hinges on individuals well-being. Utilitarianism: the most familiar version of welfarism (Jeremy Bentham). Outcomes are ranked by summing up individual well-being numbers. Requires a measure of individual well-being. Prioritarianism: pioneered by the philosopher Derek Parfit. Gives greater weight (priority) to the well-being of the worse off. Operationalized by plugging individual well-being numbers into a concave transformation function (see next slide). The sum of transformed well-being, as opposed to the utilitarian s simple sum of well-being. I specifically advocate lifetime prioritarianism. In a given possible outcome, each individual has a lifetime well-being number, as a function of her life-history (her attributes in that outcome over her entire life). Outcomes are ranked by summing up individuals transformed lifetime well-being numbers. Requires (1) a measure of lifetime well-being and (2) the choice of a transformation function. The argument for a lifetime approach: the strength of someone s ethical claim to an improvement in well- being at time T depends not merely on how well she is doing at T, but also how she has fared at previous times. The fair innings idea in the public health literature.

A Prioritarian Transformation Function g(wH) g(wH w) Transformed well- being, g(w) g(wL+ w) g(wL) wH w wL wH wL + w Well- being, w

Lifetime Well-Being Philosophers debate the nature of well-being: hedonic accounts, preference accounts, objective- good accounts. Lifetime prioritarianism is compatible with any account of well-being (as long as lifetime well-being is measurable). Economic modelers typically adopt a preference view of well-being, and measure well-being with utility numbers (the approach I ve followed in my own work); but prioritarianism itself is not wedded to a specific view of well-being. Income-longevity tradeoffs at the level of individual well-being. (1) Holding fixed the duration of a life at longevity L, income is instrumental to the quality of life (more income tends to make life better, whether the view of well-being is hedonic/preference-based/objective). (2) It is possible to have a conception of lifetime well-being that accords absolute priority to longevity over quality of life. For any two lives (Q+, L) and (Q, L*), if L* > L, then (Q, L*) better than (Q+, L) regardless of how Q and Q+ compare. But this seems very implausible. (3) Thus, on any plausible conception of well-being, there will be tradeoffs between quality and longevity as contributors to lifetime well-being, and therefore between income (instrumental to quality) and longevity as contributors to lifetime well-being. To illustrate: in my simulation exercise, I use a (highly simplified!) model of life histories. An individual s lifetime well-being depends upon her longevity plus her income each year alive. Lifetime well-being is the sum of the logarithm of annual income. Thus, e.g., living 60 years with an annual income of $100,000 (60 5) is better than living 70 years with an annual income of $10,000 (70 4).

Prioritarianism under Uncertainty Policymakers operate under uncertainty. The policymaker doesn t know, for certain, what outcome will result from a given policy choice. Instead, each policy choice (including both the choice of leaving in place the status quo, and policy interventions relative to the status quo) is seen as a probability distribution ( lottery ) across outcomes. More specifically, prioritarianism (as most plausibly outfitted for uncertainty) does this. In the status quo, each individual (or each cohort of similarly situated individuals) faces a lottery over possible life-histories. The value of that lottery is the individual s expected transformed lifetime well-being. A policy intervention changes the individual lottery (e.g. by changing survival probabilities or income) and thus yields a different level of expected transformed lifetime well- being. The best policy maximizes the sum of expected transformed lifetime well-being (ETLW) across individuals. Let s now see how this approach plays out in a simulation model of Covid-19 policy.

A Covid-19 Simulation Population assumed to consist of individuals aged 20, 30, 40, 50, 60, 70, or 80, divided into 5 income groups corresponding to 10th, 30th, 50th, 70th, and 90th percentiles of the U.S. income distribution. So 35 cohorts. Survival curves (probability of surviving each year of life) and lifetime income profiles for each group based upon US population survival curve and income distribution. Percentages of different ages also from US data. I assume that, without intervention, 81% of the population becomes infected and the various groups face the infection fatality risks estimated by the influential Imperial College report of March 2020. For simplicity, in this model, the increased Covid-19 risk is borne in a single year. Govt can wholly eliminate the risk by an 80% reduction in GDP in that year. There is a (nonlinear) function from GDP reduction to reduction in infection risk. (For example, a 20% reduction in GDP leads to a 50% reduction in each group s infection risk.) Losses in income are borne proportionally by the different groups. (If GDP is reduced by p%, then the income of each of the 35 cohorts goes down by p%.) Lifetime well-being is sum of log income. I use prioritarianism with a moderate degree of priority to the worse off (Atkinson = 1.5) Right away, we can see that the nature of Covid-19 will create a divergence of interests between age groups. Since older groups have much higher fatality rates when infected, they tend to prefer large reductions in GDP (more stringent shutdowns). A given reduction in everyone s infection risk reduces the fatality risk of older groups by more. Age 20 30 40 50 60 70 80 .03% .08% .15% .60% 2.20% 5.10% 9.30% Fatality Risk if Infected

Optimal Policies -- Each group makes intrapersonal tradeoffs between risk and income. All groups want some degree of risk reduction, but none want to spend the amount of income required to wholly eliminate the risk. -- Richer individuals are willing to spend more for a given risk reduction, hence prefer longer shutdowns. -- The younger and poorer groups prefer shorter shutdowns. -- Prioritarianism adjudicates between the conflicting policy preferences of the different groups. It does so by maximizing the sum of ETLW across groups. ETLW is the ethical currency by means of which risk reductions and income losses for different groups are commensurated. In this model, prioritarianism finds that a 14% reduction in GDP is ethically optimal. This yields a significant reduction in each group s fatality risk (42%) but falls well short of completely eliminating the Covid-19 risk or reducing it to the level preferred by older groups. On balance, further reduction is not worth it. Preferred Percentage Reduction (Lockdown Stringency) by Cohort (Maximizes Cohort s Expected Well-Being) Income: Low Moderate Middle High Top Age: 20 0 0 0 0 0 30 1 1 1 1 1 40 1 1 1 1 2 50 5 6 8 10 15 60 21 27 31 35 44 70 32 39 45 49 58 80 31 39 46 51 61 Fatality Risk 20 30 40 50 60 70 80 Before .02% .06% .12% .49% 1.78% 4.13% 7.53% After .01% .04% .07% .28% 1.04% 2.4% 4.38%

Prioritarian Tradeoffs Value of a reduction in fatality risk by 1-in-100 for each group as a multiple of value of $5000 for 50 year old, middle income group Value of $5000 for each group as a fraction of value of 1-in-100 fatality risk reduction for 50 year old, middle income group Income: Low Income: Low Moderate Middle High Top Moderate Middle High Top Age: 20 76 71 67 63 56 Age: 20 105% 52% 28% 16% 4% 47 45 42 40 36 27% 15% 9% 5% 2% 30 30 29 27 26 25 23 19% 11% 6% 4% 1% 40 40 17 17 16 16 14 18% 10% 6% 4% 1% 50 50 10 10 10 10 9 19% 11% 7% 4% 1% 60 60 5 5 5 5 5 23% 13% 8% 5% 1% 70 70 2 3 3 3 3 28% 17% 10% 6% 2% 80 80

Utilitarian Tradeoffs The utilitarian optimum (13% GDP reduction) is close to the prioritarian, but the underlying intergroup tradeoffs are quite different from the prioritarian tradeoffs Value of a reduction in fatality risk by 1-in-100 for each group as a multiple of value of $5000 for 50 year old, middle income group Value of $5000 for each group as a fraction of value of 1-in-100 fatality risk reduction for 50 year old, middle income group Income: Low Income: Low Moderate Middle High Top Moderate Middle High Top Age: 20 17 20 22 25 31 Age: 20 80% 50% 33% 23% 9% 15 17 19 21 27 22% 16% 11% 8% 3% 30 30 12 14 16 17 22 17% 12% 8% 6% 3% 40 40 9 10 12 13 17 17% 12% 8% 6% 2% 50 50 6 7 8 10 12 19% 14% 10% 7% 3% 60 60 4 5 5 6 8 25% 17% 12% 9% 4% 70 70 2 3 3 4 5 33% 23% 16% 12% 5% 80 80

Some Model Assumptions Gain to life expectancy from preventing death of cohort member in current year Group Incomes Income: Low Income: Low Moderate Middle High Top Moderate Middle High Top Age: 20 $8,331 $11,425 $15,686 $21,827 $51,152 Age: 20 54.3 56.8 58.9 60.1 62.1 $22,098 $30,306 $41,607 $57,896 $135,680 45.1 47.5 49.5 50.6 52.5 30 30 $28,426 $38,984 $53,522 $74,476 $174,536 36.1 38.4 40.2 41.3 43.1 40 40 $28,681 $39,334 $54,003 $75,145 $176,103 27.4 29.5 31.2 32.2 33.9 50 50 $24,930 $34,189 $46,939 $65,316 $153,069 19.6 21.4 23.0 23.8 25.3 60 60 $19,719 $27,043 $37,128 $51,664 $121,075 12.8 14.2 15.4 16.2 17.4 70 70 $14,757 $20,238 $27,784 $38,662 $90,605 7.1 8.2 9.1 9.6 10.5 80 80

ETLW (Ex Post Prioritarianism) versus TELW (Ex Ante Prioritarianism ) p = .5 p = .5 Expected wellbeing p=.5 p = .5 Expected wellbeing Imani 50- 50- 50- 90 10 50 Jake 50- 50- 50- 10 90 50