Mortality Risk Valuation in Policy Assessment

This chapter presents recommended value of statistical life (VSL) base estimates and ranges for use in policy assessments, programmes or projects involving changes in mortality risks. These VSL estimates reflect the existing empirical evidence regarding the trade-offs that individuals make between changes in mortality risks and changes in their income or wealth. The recommended VSL estimates are based on the mean preliminary unweighted VSL (PUVSL) estimates from the meta-analysis results reported in Section 4.5 of Chapter 4. Base estimates are provided for the United States, the OECD, the European Union (EU), as well as for high-, low- and middle-income countries1. This chapter also provides guidelines for adjusting the base VSL estimates across countries and over time and discusses how to handle remaining uncertainty in the estimated values when applying them in specific policy settings. Finally, considerations for future work and applications of the base VSL estimates are provided.

The primary valuation studies and VSL estimates from the countries included in the meta-data for the current analysis are not necessarily representative of all the countries included in the country groups presented below. In fact, some countries are not represented with studies in the meta-data sample, an issue that is particularly pronounced when considering a global VSL estimate. In addition, the income levels (GDP per capita) under which each VSL estimate was derived were not harmonised before estimating average preliminary VSLs. As a result, the PUVSL estimates presented in earlier chapters must be adjusted for by differences between the income levels of the sample for each country group and the actual income levels for each country group.

The recommended VSL estimates may additionally need to be adjusted when evaluating changes in mortality risk of a specific policy in a specific country. There may also be specific rules or regulations in individual countries that necessitate modifications to the values and approaches presented in this report. The results reported in this chapter should therefore be viewed as a starting point for valuations that involve VSL.

To adjust the PUVSL estimates presented in Chapter 4 to reflect differences in income between the meta-data sample and the current income level of the country or country group, an income elasticity of 1 is recommended, based on results from the meta-regression analysis and literature review in Chapter 5. A population-weighted measure of GDP per capita for a country group i can then be calculated from the PUVSL estimates of Chapter 4 as:2

$\mathrm{B}\mathrm{a}\mathrm{s}\mathrm{e}\mathrm{ }{\mathrm{V}\mathrm{S}\mathrm{L}}_i=\mathrm{ }{\mathrm{P}\mathrm{U}\mathrm{V}\mathrm{S}\mathrm{L}}_{Sample\left(i\right)}\mathrm{*}\frac{\mathrm{G}\mathrm{D}\mathrm{P}/{Capita}_{Population\left(i\right)}}{\mathrm{G}\mathrm{D}\mathrm{P}/{Capita}_{Sample\left(i\right)}}$

Table ‎6.1 presents the recommended base VSL estimates and ranges for each group of countries based on the PUSVL estimates in Table 4.2 of Chapter 4 and the income information in Annex G.3 Variation in the ranges reflect the size of the samples, but also the variation of VSL estimates within the respective country group samples. The estimated base VSLs are similar for high-income groups assessed, ranging from USD 7.1 million for OECD member countries to USD 8.5 million for the United States and USD 8.4 million for the EU. Note that Table ‎6.1, in addition to the mean base VSL, also presents a calculated median VSL. This value is calculated from the mean base VSL utilising OECD data on household income for 2022 (OECD, 2025). It is intended to provide an estimate of the VSL for a median income household, a measure that may be desirable for policy purposes. Across OECD countries, the median disposable household income is generally 10-20% below the mean household disposable income.

For each country group shown in Table ‎6.1, the base VSL estimate represents a population-weighted income adjustment that reflects population differences among countries and groups of countries rather than weighting every country equally. Practically, this means that the GDP per capita of more populous countries count more in the calculation of the weighted average base VSL for a group of countries. One area where this has a particularly pronounced effect is for the base VSL estimate for the world (Global) which is reduced from the preliminary unweighted meta-analysis results of USD 5.5 million, to USD 2.7 million, reflecting the lower global population-weighted GDP per capita compared to the sample GDP per capita in the meta-analysis.

In some circumstances it may be appropriate to use base VSL estimates that are adjusted by a country-weighted measure of GDP per capita instead of a population-weighted measure, which provides equal weight to each country in a given country group. Table ‎6.2 provides base VSL estimates that are adjusted using country-weighted GDP per capita. These estimates are relatively similar to those calculated on the basis of population-weighted GDP per capita in Table ‎6.1 for high-income country groups, ranging from USD 7.7 million for OECD Member Countries to USD 8.7 million for the EU. A comparison of the country- and population-weighted VSLs shown in Figure ‎6.1 indicates a larger discrepancy for the global country group than for other groups.

Note that in calculating the recommended base VSL estimates, no adjustment was made for factors other than income, as discussed in detail in Chapter 5. The VSL estimates are also based on SP and RP studies over the 2009-2023 period and are not randomly distributed across countries or regions. No adjustments were made for methods or for other factors that could potentially be “over- or under-represented” compared to an ideal “randomly drawn” evidence base4. This is, however, the nature of working with meta data. For example, it may be the case that certain countries have focused more on certain mortality risks in their VSL studies (e.g. air pollution) than other countries. As shown in this report, the choice of valuation approach (SP or RP) does not significantly influence the results, but the choice of types of individual SP and RP methods (HW, CM, CE and CV) and how these are distributed over time and across countries could potentially play a role in explaining differences in observed VSL estimates between countries and regions (cf Section 3.3 of Chapter 3 and Section 4.6 of Chapter 4) . Other aspects that are not captured in the meta-analysis and data likely also contribute to the observed differences, such as between the EU and the United States, despite the GDP per capita being significantly higher in the United States.

The base VSL estimates provided in Table ‎6.1 and Table ‎6.2 represent a starting point for the valuation of mortality effects. Adjustments may be appropriate to account for the type of policy or programme considered, as well as for the countries/populations involved. One key issue is whether to use a single VSL estimate to assess the mortality effects within a group of countries, or if different VSL estimates should be used across countries or populations, for example due to differences in income levels (GDP per capita).

The most common approach for valuing mortality impacts in international cost benefit analyses (CBA), as recommended in the research and policy guidance literature, is to apply differentiated VSL estimates to value policy-related impacts in different countries (Robinson et al., 2019[2]; USEPA, 2024[3]). This approach is motivated by the fact that GDP per capita is strongly correlated with willingness to pay for risk reductions5. The use of differentiated VSL estimates is the default approach recommended in this report.

The base VSL estimates for a given country j, which is a part of country group $i$(Equation ‎6.1), can be derived by scaling the base VSL estimate of the relevant country group by the ratio of the GDP per capita (2022 USD) of country j and that of the country group i following Equation ‎6.2.6 As discussed in Chapter 5, an income elasticity of 1 is recommended based on the results of the current meta-analysis and a review of relevant literature.

$\mathrm{B}\mathrm{a}\mathrm{s}\mathrm{e}\mathrm{ }{\mathrm{V}\mathrm{S}\mathrm{L}}_j=\mathrm{ }{\mathrm{P}\mathrm{U}\mathrm{V}\mathrm{S}\mathrm{L}}_{Sample\left(i\right)}\mathrm{*}\frac{\mathrm{G}\mathrm{D}\mathrm{P}/{Capita}_{Population\left(j\right)}}{\mathrm{G}\mathrm{D}\mathrm{P}/{Capita}_{Sample\left(i\right)}}$

Scaling the base VSL estimate as described above applies to both national and international studies7. There may be situations in which using differentiated VSL estimates in an international multi-country context may not be desirable, such as for policies that affect a more formally associated group of countries (e.g. the EU) and in the assessment of common policies based on shared analytical frameworks. In such situations, where each country typically has one vote, it could be considered more appropriate to use a single VSL estimate to assess impacts across countries within the country group. In the case of the EU, this could be the country-weighted base VSL estimate for the EU reported in Table ‎6.2. The implication of using a country-weighted base VSL estimate is that estimates from the more populous countries within the group are weighted equally as those with small populations. Another example of where a single VSL estimate could be considered is in the assessment of the benefits of the mortality risk reductions stemming from a policy that affects multiple countries, such as air pollution or climate change-related policies, e.g. Bressler (2021[4]) and Broome (2012[5])8. However, the most common approach to valuing the mortality impacts of climate change in integrated assessment models and in other climate impact studies (e.g. those using estimates of the social cost of carbon) remains the use of VSL estimates that are differentiated by country using the mean GDP per capita of each country (Carleton et al., 2022[6]; Rennert et al., 2022[7]; USEPA, 2024[3]; World Bank, 2024[8]).9 Many of these studies use a VSL estimate from the United States as a basis for differentiation (Rennert et al., 2022[7]; World Bank, 2024[8]). It is recommended that estimates presented in this report be used in future studies as a basis for differentiation instead, as they reflect the best available global evidence. It is further recommended that the VSL estimate for the smallest country group to which a country belongs (e.g. for Portugal, the EU instead of the high-income country group) is used as a basis for calculating country-specific VSL estimates based on income differences following Equation ‎6.2.

In forward-looking CBAs of policy initiatives, benefits and costs accrue into the future. When using VSL estimates to assess such impacts, it is recommended that practitioners adjust the base VSL estimates to account for how they evolve over time. In a first step, this involves adjusting for inflation as well as for real annual growth in GDP per capita (applying an income elasticity of 1) as measured from 2022 to the year(s) for which the VSL is sought. Hence, if real GDP is expected to grow by 2% annually, the VSL estimate should also be expected to increase by 2% annually. In a second step, the monetised mortality benefits (or costs) should be discounted back to the year in which the CBA is being carried out in order to calculate the net present value of future costs and benefits, following best practice or standard guidance on the choice of discount rate from national or international CBA guidelines.

As supported by the meta-regression analysis (Section 5.3 of Chapter 5) and a review of the literature (Section 5.4 of Chapter 5), it is not recommended to adjust the base VSL estimates presented above for factors such as age, gender and the type of mortality risk considered. This applies to VSL estimates for country groups as well as for individual countries. As discussed in Section 5.4.6 of Chapter 5, although cancer was found in to have a significant effect on the VSL estimates in the meta-regression analysis, a “cancer premium” is difficult to operationalise due to the large variation of cancer types and relevant contexts. Other factors, such as placing a higher value on mortality risk impacts for children, or differentiating VSL estimates by age more generally, is also not recommended due to a lack of sufficiently robust evidence to justify such systematic treatment.

Note that there are private and public costs that are additional to the mortality valuation, such as the costs of treatments and hospitalisation that should be added to the VSL when estimating the total social value of preventing a fatality. Analysts should also be aware of the risk of double-counting of morbidity and mortality effects when aggregating health-related impacts in a given CBA.

While this report provides recommended base VSL estimates for use in policy assessment, there may be compelling reasons for making adjustments to the base VSL estimates beyond what is presented here to reflect specific policy and country contexts. Specific rules or regulations in individual countries may also necessitate adjustments that differ from what is presented here. Furthermore, to the extent there are primary studies available in a country, it may be preferrable to use these on their own as basis for a recommended domestic VSL estimate rather than to use the transfer procedures outlined here that draw on a larger evidence base from the group to which the country belongs. The two approaches could also be combined, e.g. for sensitivity analysis. Pros and cons of both approaches will need to be considered in the specific country situation (see also discussion in the next subsection).

The main underlying uncertainty of the base VSL estimates is reflected in the ranges around the mean estimates reported in Table ‎6.1 and Table ‎6.2, based on the meta-analysis confidence intervals. These ranges should be considered when adjusting base VSL estimates to specific policy or country contexts. In addition, it is good practice in CBA to investigate the sensitivity of CBA results to key assumptions, especially if the stakes are high or the estimated costs and benefits of a policy are of similar magnitude.

While it is clear that GDP per capita is strongly correlated with VSL, the measure remains a relatively crude measure of income, and the income elasticity of VSL can have a relatively large effect on VSL estimates across countries following Equation ‎6.2 above. The uncertainty involved in extrapolating VSL estimates across countries based on GDP per capita is particularly large for countries with very low and very high GDP per capita levels.

Furthermore, uncertainty remains regarding most appropriate values to use for the income elasticity of VSL, with studies indicating that the value could be higher for lower-income countries and lower for higher-income countries. It is therefore recommended to perform sensitivity analyses using an income elasticity of VSL value of 0.5 when extrapolating to individual countries within the high-income country group (e.g. from base VSL estimates for OECD Member Countries or the EU) and a value 1.5 when extrapolating to countries within the low- and middle-income country group10. This implies that the adjustment of the VSL estimates will be relatively lower for high-income countries compared to using the recommended default elasticity of 1. For transferring VSL estimates to countries in the low- and middle-income categories, using a higher elasticity would imply a larger adjustment of the VSL estimate, relative to using a default elasticity value of 1. Note that this implies that countries in the high-income country group will use VSL estimates closer to the mean of the group, while the opposite will be the case for countries in lower-income country groups.

Conducting sensitivity analysis using income elasticity estimates of 0.5 and 1.5 can also contribute to determining the extent to which VSL estimates for individual countries can be considered plausible. For example, for a country such as Luxembourg, which is characterised by a high GDP per capita relative to other countries, the calculated VSL estimate would be USD 19.2 million using the default assumption of an income elasticity of VSL of 1. Although this may be considered a relatively high number, there is little guidance in the literature to assess what should be considered high or low bounds on VSL estimates (or WTP more generally). Assessing the mean WTP as indicated by the VSL estimate as a share of GDP per capita or other income measures (e.g. median disposable income or household expenditures), could be a way to “ground truth” VSL estimates for individual countries. If WTP is unrealistically high (or low) compared to a reasonable measure of income or other expenditures, the VSL estimate may be considered implausible11. Considering the uncertainty of the estimates and the lack of firm guidance in the literature, however, it is difficult to formulate more detailed guidance on this issue.

Assuming an income elasticity of VSL of 0.5, for example, would reduce the VSL estimate for Luxembourg to a relatively more modest USD 12.1 million. In contrast, for a country such as Tanzania with a relatively low GDP per capita relative to other countries, the derived VSL estimate would be about USD 240 000 when assuming an income elasticity of VSL of 1, compared to ca. USD 127 000 when assuming an elasticity of 1.5. Again, assessing WTP against GDP per capita and other income measures could provide some perspective regarding the plausibility of the calculated VSL estimates for individual countries.

Less is known about the income elasticity of VSL over time than across countries and populations. It has become relatively standard to use an income elasticity of 1 to adjust VSL with projected growth in real GDP per capita for forward-looking analysis of policy proposals, and this is the approach taken in this report. The current analysis did not adjust VSL estimates taken from primary valuation studies for differences in income when compiling the meta-data for the meta-analysis. Instead, an adjustment for income is made after the meta-analysis to account for the difference between the GDP per capita in 2022 of the relevant country group and the income level reflected in the meta-data sample (see Equation ‎6.1). This approach follows USEPA SAB (2017, p. 5[9]), which argues that adjusting for income on the input-side of the meta-analysis “goes beyond any meta-analytic practices in the literature” and is a type of adjustment more appropriate as part of benefit transfer procedures when adjusting VSL into the future for CBA purposes. However, two examples of meta-analysis studies have used such adjustments recently (Banzhaf, 2022[10]; Ginbo, Adamowicz and Lloyd-Smith, 2023[11]). Hence, this issue appears to have yet to be fully resolved.12

In addition to the assumptions regarding the income elasticity of VSL, a number of other factors could be explored in sensitivity analyses of CBAs. For example, if analysts prefer to use specific quality criteria, which are coded in the dataset (see Chapter 4), they may wish to screen the data for distinct country groups before calculating new mean VSL estimates using the meta-analysis procedure outlined in Chapter 4. Sections 4.6 and 4.7 of Chapter 4, as well as Annex E provide VSL estimates calculated using various screening procedures. Since peer-reviewed journal publications may be of particular interest to some analysts, the mean VSL estimates for the meta-data screened on this basis are provided for illustrative purposes in Table A E.1 in Annex E.

The VSL estimates derived in this report have many potential applications for assessing the benefits (costs) of policies involving reductions (increases) in mortality risks. Such applications can be found in the environment, transport, energy, food safety and health sectors. Geographically, the VSL estimates presented here are applicable at the local, national, as well as international and global levels. Local applications could include the assessment of health benefits from reducing air pollution in cities or reducing water pollution in certain lakes or rivers. National applications could involve valuing mortality effects from changing vehicle fuel standards, tailpipe emissions or water quality. International applications could include the assessment of common EU policies to increase food safety or to reduce cross-border pollution. At the global level, such estimates could be used to update current estimates of the social cost of greenhouse gases.

In addition to the more standard CBA of projects, programmes, or policies, the VSL estimates reported here could also be used in studies that estimate health costs of inaction or the total mortality costs of certain problems (e.g. traffic planning assessments or for certain diseases). Such assessments can, moreover, be forward- or backward-looking. Analyses of the effects of previously implemented policies can be highly informative, such as the ex-post analysis of the Clean Air Act in the United States. The VSL estimates presented here can also be applied to the valuation of environmental and health impacts in corporate green metrics as part of the social costs of emissions from various sources impacting a company’s products and supply chains.

For all such analyses, morbidity impacts should be valued separately and added to the mortality costs, both for morbidity prior to premature death and for non-fatal health outcomes resulting from the policy under evaluation. The updated evidence provided here on how people across countries value mortality risks constitutes an important basis for better-informed and more efficient policy decisions in the future.13

[16] Acland, D. and D. Greenberg (2023), “Distributional weighting and welfare/equity tradeoffs: a new approach”, Journal of Benefit-Cost Analysis, Vol. 14/1, pp. 68-92, https://doi.org/10.1017/BCA.2023.5.

[10] Banzhaf, H. (2022), “The Value of Statistical Life: A Meta-Analysis of Meta-Analyses”, Journal of Benefit-Cost Analysis, Vol. 13/2, pp. 182-197, https://doi.org/10.1017/BCA.2022.9.

[4] Bressler, R. (2021), “The mortality cost of carbon”, Nature Communications 2021 12:1, Vol. 12/1, pp. 1-12, https://doi.org/10.1038/s41467-021-24487-w.

[6] Carleton, T. et al. (2022), “Valuing the Global Mortality Consequences of Climate Change Accounting for Adaptation Costs and Benefits”, The Quarterly Journal of Economics, Vol. 137/4, pp. 2037-2105, https://doi.org/10.1093/QJE/QJAC020.

[11] Ginbo, T., W. Adamowicz and P. Lloyd-Smith (2023), “Valuing Mortality Risk Reductions in Canada: An Updated Meta-Analysis and Policy Guidance”, Canadian Public Policy, Vol. 49/3, pp. 233-251, https://doi.org/10.3138/CPP.2022-052.

[1] OECD (2012), Mortality Risk Valuation in Environment, Health and Transport Policies, OECD Publishing, Paris, https://doi.org/10.1787/9789264130807-en.

[14] Prest, B. et al. (2024), “Equity weighting increases the social cost of carbon”, Science, Vol. 385/6710, pp. 715-717, https://doi.org/10.1126/SCIENCE.ADN1488.

[7] Rennert, K. et al. (2022), “Comprehensive evidence implies a higher social cost of CO2”, Nature 2022 610:7933, Vol. 610/7933, pp. 687-692, https://doi.org/10.1038/s41586-022-05224-9.

[2] Robinson, L. et al. (2019), “Reference Case Guidelines for Benefit-Cost Analysis in Global Health and Development”, SSRN Electronic Journal, https://doi.org/10.2139/SSRN.4015886.

[17] Robinson, L., J. Hammitt and L. O’Keeffe (2019), “Valuing Mortality Risk Reductions in Global Benefit-Cost Analysis”, Journal of Benefit-Cost Analysis, Vol. 10, pp. 15-50, https://doi.org/10.1017/BCA.2018.26.

[15] Sunstein, C. (2023), “Inequality and the Value of a Statistical Life”, Journal of Benefit-Cost Analysis, Vol. 14/1, pp. 1-7, https://doi.org/10.1017/BCA.2023.7.

[3] USEPA (2024), Mortality Risk Valuation, https://www.epa.gov/environmental-economics/mortality-risk-valuation#process (accessed on 25 October 2024).

[9] USEPA SAB (2017), Review of EPA’s Proposed Methodology for Updating Mortality Risk Valuation Estimates for Policy Analysis.

[13] Viscusi, W. (2024), “Why Office of Management and Budget’s (OMB) Social Welfare Function Is Not Society’s Social Welfare Function”, Journal of Benefit-Cost Analysis, pp. 1-24, https://doi.org/10.1017/BCA.2024.25.

[12] World Bank (2025), GNI per capita vs. GDP per capita, 2023, https://ourworldindata.org/grapher/gni-per-capita-vs-gdp-per-capita.

[8] World Bank (2024), The Cost of Inaction: Quantifying the Impact of Climate Change on Health in Low- and Middle-Income Countries, World Bank Group, Washington D.C., http://documents.worldbank.org/curated/en/099111324172540265/P5005831a1804a05f19aae18bc0f1396763.

[5] WW Norton & Company (ed.) (2012), Climate Matters: Ethics in a Warming World.