3.1 The value-of-a-statistical life measure

The dominant role of the U.S. in contributing to the global mortality cost of COVID becomes more striking after coupling the country’s leading mortality total with a high monetized value of the risk death. To provide an assessment of the magnitude of the economic cost of COVID-related mortality, one must monetize the costs of these deaths in a manner that reflects each country’s valuation of the risk. A reduction in the risk of death does not have an infinite value. Societies have limited financial resources and have competing uses for these funds, leading governments to explicitly or implicitly place a finite value on reducing different types of mortality risks. After recognizing that it is not feasible to make an unbounded financial commitment to policies that promote individual health, the task is then to determine the best approach for making this assessment. How to value these deaths depends on the underlying conceptualization of what is being valued. If the focus is on compensating the family for the financial loss associated with the death, then the present value of the lost earnings and medical costs is usually the appropriate measure for wrongful death awards. However, applying that measure in this context does not incorporate the welfare loss to those who have lost their lives because of COVID. More generally, societies’ willingness to pay to reduce risks of death goes beyond accounting measures such as the value of the income that the deceased will no longer be able to earn.

The principal economic approach used to value mortality risks takes a prospective perspective with respect to the willingness to pay to reduce mortality risks. Thus, rather than asking what the value ex post would be of compensating survivors for the certainty of a particular death, the focus instead is on the ex ante value per expected death that is prevented in situations involving small mortality risk reductions. Consider the following example. Suppose that there is a group of 10,000 people, each of whom faces a 1/10,000 risk of death, so that there would be one expected death to their group. If each person is willing to pay $900 to eliminate that risk, then collectively members of the group are willing to pay $9 million to prevent one expected death to some member in their group. In this example, the VSL is $9 million. The VSL serves as a measure of the rate of tradeoff between money and risk for small changes in risk. The VSL overstates how much people would be willing to pay to reduce the probability of death from 1.0 to zero, and it understates how much people would need to be compensated for an increase in the probability of death from 0 to 1.0. The underlying thought experiment that is adopted here in valuing COVID mortality is to monetize COVID-19 deaths in the same manner as would be the case if the government could undertake a series of policies that would generate small reductions in risk for any given individual, but collectively would lead to the overall reduction in fatalities. This approach is consistent with that used to value mortality risk reduction provided by policy by government agencies throughout the world and in other analyses of the mortality costs of COVID-19 [4–7].

Following the practices generally adopted by government agencies, the calculations presented here apply the average U.S. VSL to monetize the mortality risk reductions in the U.S. and the average country-specific VSL counterparts to monetize deaths in other countries. As indicated below, the procedure for constructing country-specific VSL levels is to use the U.S. average VSL as the reference point and then to transfer this value to other countries by incorporating different assumptions about the influence of a country’s income level on that country’s average VSL. The focus on average VSL statistics does not imply that there is a country-specific VSL pertinent to all people in a particular country. Even within the U.S., the VSL is not a single figure applicable to everyone. Preferences with respect to risk and willingness to pay to avoid these risks vary. There are many potential sources of heterogeneity in the VSL, which is a value that varies across the population and across countries. Application of average VSL levels across a country’s population is appropriate when the risks are broadly distributed. If the risks are concentrated among those with a lower VSL, as is the case of hazards affecting those with very short remaining life expectancy, then the pertinent VSL and total mortality costs may be lower. For example, to the extent that one is applying an age-adjusted VSL using the counterpart values of a statistical life year, the mortality cost of COVID-19 would be reduced by about half [5]. However, losses from other health impacts may offset such an adjustment. Monetized morbidity costs may equal or exceed the value of the age-related reduction in monetized mortality costs [5, 8] so that the mortality cost calculations do not overstate the overall health impact even with an age adjustment. Other estimates of the morbidity cost associated with COVID based on quality-adjusted life years that are lost also indicate that the monetized value of the losses associated with morbidity impacts are substantial [9]. Another possible refinement is that if the policies that are envisioned would lead to very large changes in risk levels, the applicable VSL would be reduced [5, 7], mostly because substantial risk-reducing expenditures may deplete available funds used for risk reduction. Such influences might have come into play had the government pursued a single, very costly policy that would have prevented all COVID-related deaths. The approach adopted here is to use the average VSL across a country’s population rather than incorporating such considerations. Note that to the extent that these adjustments affect the valuation of mortality risks in all countries similarly, the relative role of the U.S. and the rate of change in the mortality costs associated with the pandemic are unaffected by adjustments for age and non-incremental risk reductions. However, the absolute level of costs is dependent on the absolute levels of the VSL.

3.2 Assigning mortality cost values

The specific values that I am using to monetize the mortality rates use as the starting point my average U.S. VSL number of $11.4 million. This estimate is the value in December 2021 dollars, (calculated using the U.S. Bureau of Labor Statistics 2022 CPI Inflation Calculator) [10], of the VSL estimates that have been derived from a meta-analysis of VSL studies using labor market data to assess the wage increase that workers require to incur fatality risks [11]. U.S. government agencies rely on this meta-analysis in their regulatory impact analyses. The studies included are restricted to those using occupational fatality risk measures from the Census of Fatal Occupational Injuries, which is a comprehensive census of all U.S. occupational fatalities. The $11.4 million estimate incorporates adjustments for any biases due to publication selection effects [12]. This figure is similar to the estimates used by the U.S. Department of Health and Human Services [13] of $9.3 million in 2014 USD (https://aspe.hhs.gov/reports/updating-vsl-estimates) and the U.S. Department of Transportation [14] of $9.4 million in 2014 USD (https://www.transportation.gov/office-policy/transportation-policy/revised-departmental-guidance-on-valuation-of-a-statistical-life-in-economic-analysis). As noted above, a VSL of $11.4 million does not pertain to the amount that people would be willing to pay to avert their death. Rather, it is the aggregate willingness to pay for the small mortality risk reductions across the population of the type one could expect from a publicly funded vaccination effort. The VSL values fatalities by a greater amount than is usually awarded in wrongful death cases, as it generally exceeds the present value of workers’ lifetime earnings by about an order of magnitude.

The use of this average VSL based on labor market decisions involving occupational risks does not imply that all individual decisions regarding vaccinations and protections in the COVID pandemic reflected a sensible balancing of the risks and the costs of risk reduction consistent with the tradeoff implied by the VSL. As Bruin de Bruine et al. [15] document, individual beliefs and protective behaviors were based on different types of information, the filtering of this information through politically oriented outlets, and the political preferences of the individual. Application of the VSL in policy decisions based on accurate assessment of the risks and costs produces decisions that would be consistent with the types of mortality reduction policies for which the VSL is often used.

To transfer the U.S. VSL statistics to construct country-specific estimates of the VSL for other nations, it is helpful to have an established empirical basis for the relationship of the VSL across countries. A substantial literature has shown that the VSL is positively related to a country’s income level. Consequently, to assess the country-specific mortality costs of the pandemic, it is feasible to transfer the U.S. VSL estimate to other countries based on differences across countries in the GNI per capita. The data utilized here are retrieved from Viscusi and Masterman [16], where they report data from “the World Bank’s GNI per capita numbers calculated using the Atlas method which is based on exchange rates and inflation rates.” Differences in attitudes toward risk, religion, health care systems, and social norms may all contribute to national variation in the VSL. However, a major determinant of national VSL differences stems from differences in income, which is a characteristic for which there is available data as well as an economic literature documenting the income-VSL relationship. How the VSL differs from that in the U.S. depends on income levels and the country’s income elasticity of the VSL. The income elasticity of the VSL is defined as the percentage change in the country’s VSL in response to a 1 percentage change in the country’s income. Here the focus is on percentage differences in the country’s GNI per capita compared to the U.S.

The relationship between income and the VSL may vary across countries so that it is useful to explore a variety of such relationships. The primary set of estimates presented here uses an estimated income elasticity of the VSL 1.0, whereby a 1 percentage lower income level in a country will lead it to have a 1 percent lower VSL. This estimate of the average international income elasticity of the VSL is based on both a meta-regression analysis of labor market estimates of the VSL [16] and a meta-regression analysis of stated preference estimates of the VSL [17], each of which yields an average across-country income elasticity estimate of 1.0 using a procedure that accounts for publication selection effects. These calculations use the same income elasticity for all countries but estimate country-specific levels of VSL taking into account the income level in each country as compared to that in the U.S. Drawing on these estimates and other studies, the U.S. Environmental Protection Agency [18] also adopted an income elasticity of 1.0 to transfer U.S. VSL estimates to countries that do not have VSL estimates in their work estimating the global mortality cost component of the social cost of carbon.

Countries with income levels much lower than the U.S. tend to have VSL levels that are more responsive to their lower income levels, making it desirable to explore the sensitivity of the results to different assumptions about the income elasticity of the VSL for low- and middle-income countries. Column 2 of Table A1 in S1 Appendix presents the baseline estimates using an income elasticity of 1.0, listing the VSL for countries throughout the world calculated based on this estimate and the associated COVID-19 mortality costs. Section 3.3 below presents results using other income elasticities that can be used to transfer the U.S. VSL to construct country-specific values, with the overall effect being that higher (lower) income elasticities tend to reduce (raise) the estimates of the global mortality costs for most countries given the relatively high U.S. GNI per capita. To the extent that countries with lower income levels than the U.S. have income elasticities greater than 1.0, the mortality cost estimates for COVID-19 will be reduced for these countries.

By the end of 2022, the total global mortality cost was $29.4 trillion, which is a per capita cost of $3,818. The cost levels by year were $9.0 trillion in the 2019 to 2020 period, $13.1 trillion in 2021, and $7.3 trillion in 2022. Table 1 summarizes these values by year and for the top ten mortality cost countries in each of the time periods indicated in the five panels of Table 1.

The U.S. was the largest contributor to the global mortality costs, with total costs of $12.7 trillion, or per capita costs of $38,370, which are an order of magnitude greater than the worldwide average. The costs by year were $4.1 trillion through 2020, $5.6 trillion in 2021, and $3.1 trillion in 2022. These figures do not imply that the government would have spent these amounts on a single policy that would have eliminated the risks. Since the available policy choices generally pertain to minor reductions in mortality risks rather than complete elimination of COVID risks, concerns regarding whether an expenditure of this magnitude would have been feasible do not arise.

Notwithstanding the considerable temporal variation in the U.S. mortality costs, the relative role of U.S. mortality costs has been remarkably invariant for all five panels of the time periods shown in Table 1. Although the U.S. has 4% of the global population, the U.S. share of global mortality costs was 45% through 2020 (Panel A), 43% in 2021 (Panel B), 44% through 2021 (Panel C), 43% in 2022 (Panel D), and 43% through 2022. The U.S. fluctuations in mortality costs consequently track the worldwide mortality cost pattern quite closely.

Several other affluent countries have consistent prominent roles in the worldwide COVID mortality costs, in part because of their relatively high VSL. The United Kingdom has an average VSL of $8.8 million, which contributed to it having the second highest mortality cost at the end of 2020, the fourth highest mortality cost in 2021, and the third highest mortality costs in 2022, leading to it being the second ranking country in terms of total mortality costs through 2022. Germany has a VSL of $9.4 million and had the third highest in mortality costs through 2022 as a result of having the sixth highest cost through 2020, the third highest cost in 2021, and the second highest cost in 2022.

Despite having lower income levels and a lower VSL, Brazil has had a prominent ranking in terms of total mortality costs. Brazil is also noteworthy in that the large number of deaths in Brazil offsets the impact of a lower VSL of $2.0 million. Brazil ranked fifth in total mortality costs at the end of 2020, was second in 2021, and did not make the top 10 in 2022 despite 74,832 COVID deaths, which is more than the number of deaths for all but two of the countries listed in the top 10 in Panel D of Table 1. The combined influence of the relatively low VSL and large number of deaths resulted in Brazil finishing fourth overall for the cumulative mortality costs at the end of 2022.

3.3 Alternative assumptions regarding the income elasticity of the VSL

The extrapolation of the U.S. VSL to other countries thus far was based on the assumption that the country-specific VSL could be constructed using an income elasticity of 1.0. Other assumptions that can be used to transfer the U.S. VSL to other countries are also possible. For example, if all countries were accorded the same VSL as the U.S., the total global mortality cost of the pandemic would be $75.8 trillion rather than $29.4 trillion. The ordering and relative values of the countries’ mortality costs would be the same as the number of COVID-related deaths. One might suggest that applying a uniform VSL for all countries is the correct approach on the grounds that all lives are equally valuable. However, the mortality cost estimates do not set out to measure the intrinsic moral value of these lives. Rather, the mortality cost calculations measure society’s willingness to pay to reduce mortality risks in these different countries. This valuation of risks varies with the income levels in those countries just as do most other categories of expenditures. Moreover, it is not likely that a country’s VSL that emerges based on an income elasticity of 1.0 generally understates the country-specific VSL since the VSL based on this income elasticity is greater than the VSL that government agencies in these countries use for policy analyses [12].

Use of an elasticity of 1.0 is consistent with U.S. Department of Transportation [19] practices for modifying the U.S. VSL in response to changes in income, but there is more diversity in the VSL income elasticity assumptions used in analyses outside the U.S. A prominent alternative economic approach is to suggest that the VSL for low- and middle-income countries outside the U.S. should be based on an income elasticity greater than 1.0 [20–23]. Adopting a formulation that leads to a more than proportional relationship of the VSL and income will lead to an even lower VSL for these lower income countries than would prevail with an elasticity of 1.0. The rationales that Hammitt and Robinson [20] and Robinson, Hammitt, and O’Keefe [23] offer for using a higher elasticity estimate for lower income countries are based on across-country comparisons of the VSL and the high percentage of income that would be devoted to risk reduction based on the use of a unitary income elasticity. Robinson, Hammitt, and O’Keefe [23] propose that policy analyses include a sensitivity analysis based on an income elasticity of 1.5 for low- and middle-income countries, which will be done below. Hoffman, Krupnick, and Qin [24] have also questioned the common use of a unitary income elasticity for international VSL estimates, but found a different relationship. Their study drew on a set of studies fielding the same stated preference VSL valuation instrument in Canada, China, France, Italy, Japan, Mongolia, the United Kingdom, and the U.S. They found estimated VSL-transfer income elasticities slightly below 1.0 for lower-income countries, implying a smaller diminution of the VSL for low-income countries than would be predicted based on an income elasticity of 1.0. The World Bank study by Narain and Sall [21] adopted an income elasticity of 1.2 for transferring the VSL to low- and middle-income countries. The World Bank and the Institute for Health Metrics and Evaluation [22] analysis of international air pollution costs adopted an elasticity of 1.2 for low- and middle-income countries and 0.8 for high-income countries. OECD [25] recommended an elasticity estimate of 0.8 for OECD countries. The OECD [26] assessment of international mortality costs of air pollution adopted greater income elasticities for low- and middle-income countries but started at a lower elasticity base level for high-income countries. That OECD [26] report used income elasticities of 0.8 for high-income countries, 0.9 for middle-income countries, and 1.0 for low-income countries.

Although the average estimated international income elasticity of the VSL of 1.0 is consistent with the U.S. Environmental Protection Agency’s [18] use of an international income elasticity of 1.0, many empirical estimates of the income elasticity of the VSL are greater for low- and middle-income countries. In the international VSL meta-regression analysis by Viscusi and Masterman [16], the estimated income elasticity at the 10^th percentile of the international VSL distribution was 1.5, which would imply a 1.5% drop in the VSL if a country has a 1% lower income level than the U.S. The income elasticity at the 25^th percentile of the VSL distribution across different countries was 1.2, which would entail a reduction of the VSL by 1.2% for a 1% lower income level. Applying these higher income elasticities has little effect on high-income countries but has a greater impact on the VSL for lower income countries. The VSL for the United Kingdom is $8.8 million based on an elasticity of 1.0, $8.4 million using an income elasticity of 1.2, and $7.8 million using an income elasticity of 1.5. The VSL for Thailand drops more steeply from a VSL of $1.2 million at an income elasticity of 1.0, to $0.7 million with an income elasticity of 1.2, and to $0.4 million with an income elasticity of 1.5. A very low-income country such as Bangladesh has a VSL of $240,000 for an elasticity of 1.0, $110,000 for an elasticity of 1.2, and $40,000 for an elasticity of 1.5.

It is instructive to explore the implications of different income elasticity assumptions of 1.0, 1.2, and 1.5. In addition to presenting estimates transferring the U.S. VSL to other countries based on an income elasticity of 1.0, Table 2 reports results using an upper bound elasticity of 1.5 and an income elasticity of 1.2, both of which have appeared in previous empirical studies of the VSL as well as in policy analyses. The calculation of the VSL_i for any country i using information on the VSL_US in the U.S., the income Y_i in the country i, the income Y_US in the U.S., and the income elasticity, ε, is given by

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Cumulative COVID-19 mortality costs with different income elasticities.

https://doi.org/10.1371/journal.pone.0284273.t002

The mortality cost statistics in Table 2 are shown separately for different groups of countries based on their per capita GNI levels. The bottom income quartile extends from Burundi to Sao Tome and Principe. The 26^th–50^th quartile includes countries from Sudan and the Solomon Islands to Thailand. The 51^st–75^th quartile has a per capita GNI range from Iraq to Oman. Finally, the countries in the top quartile per capita GNI range extend from Slovakia to Bermuda. Panel A pertains to the total costs through 2020, Panel B summarizes the total costs through 2021, and Panel C summarizes the total costs through 2022.

Since the pattern of mortality cost values is similar across all panels, let us focus on the most recent cumulative results in Panel C. The extent to which the mortality costs are affected by use of different income elasticities varies by quartile in the expected fashion. The top quartile is least affected, as there is only a 6% drop in the mortality costs after switching the income elasticity from 1.0 to 1.5. The bottom quartile estimates change much more substantially, exhibiting an 84% decline as a result of the elasticity increase from 1.0 to 1.5. However, because the value of the total global mortality costs is driven principally by the value of fatalities in the upper income countries, the value of the global mortality costs only declines by 18% if one adopted an elasticity of 1.5 throughout the calculations rather than 1.0. In that case, the U.S. mortality cost share of global mortality costs is 53%. The intermediate income elasticity of 1.2 leads to a 9% decline in global mortality costs. The U.S. share of the global mortality costs is 48% in this intermediate case.

Using the different results in Table 2, it is also possible to fine tune the mortality cost estimates if one wishes to assign different income elasticities to different income groups. For example, suppose that the mortality costs through January 1, 2023, have an income elasticity of 1.0 for the top two quartiles, 1.2 for the 26^th–50^th quartile, and 1.5 for the 0–25^th quartile. Then the total global mortality cost is $29.0 trillion which is 1% less than cumulative mortality cost shown in Panel E of Table 1. If instead one had applied an income elasticity of 1.5 for the 0–25^th quartile and 1.2 throughout the 26–75^th quartiles, and 1.0 for the highest quartile, the total mortality cost would have been $27.4 trillion which is a 7% reduction from the income elasticity of 1.0 case.