A better way to grapple with benefit-cost trade-offs in a pandemic

The coronavirus pandemic has forced governments to face excruciating trade-offs. Benefit-cost analysis is a standard framework for making policy trade-offs, and some have suggested that it be used to make policy trade-offs in this pandemic. But benefit-cost analysis is flawed. We can do better.

The spread of this horrific virus has put front and center two types of trade-offs: Risk-income and risk-risk (risk here meaning fatality risk).

Shutting down businesses to enforce social distancing helps to “flatten the curve,” reducing individuals’ risks of dying from COVID-19. A flatter curve gives more time for the development of antivirals, the production of needed protective gear and medical devices and ultimately a vaccine. But a shutdown lowers individuals’ incomes. Income is reduced dramatically during the shutdown itself, and then later too if the shutdown causes a recession. How much should we as a society “pay” in lower incomes for a given flattening of the curve? This is a risk-income trade-off.

Looming shortages of gloves, masks and ventilators force us to think about rationing. Protective gear reduces fatality risk by lowering infection risk. Should police officers take priority over grocery store clerks in receiving N95 masks? This is a risk-risk trade-off. Medical devices cut down fatality risk among the infected. Which seriously ill patients should be put on ventilators if there aren’t enough ventilators for all? Should younger patients take priority over older ones? Again, a risk-risk trade-off.

Benefit-cost analysis is widely used by economists, and is now the dominant policy-analysis methodology in the federal government. It works as follows. Each positive or negative impact of a policy is converted into a monetary equivalent by asking how much individuals are willing to pay (for a positive impact) or willing to accept (in exchange for a negative one). The social value of a policy is calculated as the sum of the monetary equivalents for its positive impacts minus the sum of the monetary equivalents for its negative ones.

Valuing fatality risk reduction is nothing new for benefit-cost analysis. The linchpin is the so-called “value per statistical life” (VSL). An individual’s willingness to pay for a risk reduction is just the risk reduction multiplied by VSL. Imagine that Felicia’s VSL is $6 million. This means that Felicia is willing to pay $6 for a 1-in-1 million risk reduction, $60 for a 1-in-100,000 risk reduction and $600 for a 1-in-10,000 risk reduction.

In principle, VSL varies among individuals. Felicia’s willingness to pay for risk reduction need not be the same as Victor’s. Textbook benefit-cost analysis says to convert individuals’ risk reductions into monetary equivalents using individual-specific VSLs. But this approach has dramatically counterintuitive implications when it comes to risk-risk trade-offs. Because richer individuals tend to have higher VSLs, textbook benefit-cost analysis gives them priority in risk reduction. It implies that richer patients should get priority in receiving ventilators. 

The U.S. government in practice deviates from textbook benefit-cost analysis by using a single VSL for everyone (a population average). The number used is generally around $10 million. Benefit-cost analysis with a single VSL avoids giving priority to the rich in risk-risk trade-offs, but it has other difficulties.

First, it fails to give priority to the young in risk-risk tradeoffs. Do we really think that a rationing scheme for ventilators should not differentiate between 30 year-olds and 70 year-olds?

Second, benefit-cost analysis (whether we use individual-specific VSLs or a population average) is completely insensitive to the distribution of income. It surely matters how the costs of a shutdown are distributed across economic groups — which in turn depends on the details of the fiscal policies that government puts in place to mitigate those costs. But benefit-cost analysis ignores income distribution; it says that society should be indifferent to whether the costs of a shutdown are borne by the poor, the middle class or the rich.

We can do better. A different methodology is sometimes used in economics, especially for tax policy and climate change. This methodology is called the “social welfare function.” Rather than translate policy impacts into monetary equivalents, this framework translates them into “utilities.” An individual’s “utility” is a measure of her strength of preference for various goods (longevity, income, health and so forth). If Abigail prefers one bundle of goods to a second, the first bundle gets a higher utility.

We’ve written about how the social-welfare-function approach can be applied to risk policies. In a nutshell, each cohort of similarly situated individuals (for example, age groups subdivided by income) can be seen as facing a lottery over longevity-income bundles. A given governmental policy shifts the lottery that each cohort faces. The simplest, utilitarian, version of this framework assigns a social value to a policy by summing expected utilities across cohorts. A different version, “prioritarianism,” gives extra weight to the worse off. As we’ve demonstrated, this methodology has important advantages over benefit-cost analysis when it comes to risk-income and risk-risk trade-offs. It gives preference to the young in risk-risk tradeoffs, but either mitigates (utilitarian) or eliminates (prioritarian) the preference for the rich. Moreover, it is sensitive to the distribution of costs — preferring that income losses be borne by those higher up the socioeconomic ladder.

Benefit-cost analysis is a serviceable tool for grappling with trade-offs, but it can be improved on — and the social-welfare-function framework shows how.

Matthew Adler is a professor of law at Duke Law School. James Hammitt is a professor of economics and decision sciences at the Harvard T.H. Chan School of Public Health.