ENVIRONMENT AND NATURAL RESOURCES GLOBAL PRACTICE Methodology for Valuing the Health Impacts of Air Pollution Discussion of Challenges and Proposed Solutions Urvashi Narain and Chris Sall Methodology for Valuing the Health Impacts of Air Pollution Discussion of Challenges and Proposed Solutions Urvashi Narain and Chris Sall © 2016 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington, DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org Some rights reserved 1 2 3 4 19 18 17 16 This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Nothing herein shall constitute or be considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions This work is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) http://creativecommons.org/licenses/by/3.0/igo. Under the Creative Commons Attribution license, you are free to copy, distribute, transmit, and adapt this work, including for commercial purposes, under the ­following conditions: Attribution—Please cite the work as follows: Narain, Urvashi, and Chris Sall. 2016. “Methodology for Valuing the Health Impacts of Air Pollution: Discussion of Challenges and Proposed Solutions.” World Bank, Washington DC. License: Creative Commons Attribution CC BY 3.0 IGO Translations—If you create a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be considered content or error in this an official World Bank translation. The World Bank shall not be liable for any ­ translation. Adaptations—If you create an adaptation of this work, please add the following disclaimer along with the attribution: This is an adaptation of an original work by The World Bank. Views and opinions expressed in the adaptation are the sole responsibility of the author or authors of the adaptation and are not endorsed by The World Bank. Third-party content—The World Bank does not necessarily own each component of the content contained within the work. The World Bank therefore does not warrant that the use of any third- party-owned individual component or part contained in the work will not infringe on the rights of those third parties. The risk of claims resulting from such infringement rests solely with you. If you wish to re-use a component of the work, it is your responsibility to determine whether permission is needed for that re-use and to obtain permission from the copyright owner. Examples of components can include, but are not limited to, tables, figures, or images. All queries on rights and licenses should be addressed to the Publishing and Knowledge Division, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@ worldbank.org. Cover photo: Chris Stowers. Used with permission. Further permission required for reuse. Cover design: Bill Pragluski, Critical Stages, LLC. Contents Acknowledgments............................................................................................................................. vi 1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 2 Decisions on What Is to Be Valued: Metrics of Health Impacts . . . . . . . . . . . . . . . . . . . 3 Metrics of Health Impacts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Valuing Mortality and Morbidity Together. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Measuring Premature Mortality. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Notes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 3 Approaches to Valuing the Economic Costs of Premature Mortality. . . . . . . . . . . . . . 10 Willingness-to-Pay Approach. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Other Approaches to Valuing Health Impacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Use of the Willing-to-Pay–Based Approach versus the Forgone Output Approach. . . . . . 13 Factors Influencing the Value of Statistical Life. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Age. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Health Status. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Income. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 Risk Characteristics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 The Value per Statistical Life Year. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Base Value of Statistical Life Estimates for Policy Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . 24 Transferring Base Estimates of the Value of Statistical Life to Value Health Impacts in Other Countries. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 Unit Value Transfer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 Function-Based Transfer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 Credibility Testing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Choice of Exchange Rate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 Implementing the Forgone Output Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Defining the Workforce. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 Working Life Expectancy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 Child Mortality. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Accounting for Income Growth. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 Discount Rate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 Notes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 4 Valuing Morbidity Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 Boxes 2.1 Question: Should Mortality and Morbidity Impacts Be Valued Together in Terms of Disability-Adjusted Life Years?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.2 Question: Should Mortality Impacts Be Measured as Number of Deaths or Number of Life Years Lost?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 3.1 Question: When Should the Willingness-to-Pay Approach Be Used Instead of the Forgone Output Approach?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Methodology for Valuing the Health Impacts of Air Pollution iii 3.2 Question: When Using the Value of Statistical Life, Should the Value of Statistical Life Be Adjusted for Age Differences in the Adult Population Affected by Air Pollution?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 3.3 Question: How Should Deaths among Children Be Monetized?. . . . . . . . . . . . . . . . . . 17 3.4 Question: Should Mortality Risks Be Valued Differently for People in Different Health States?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 3.5 Question: Should the Value of Statistical Life Be Adjusted for Differences in Income?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 3.6 Question: Should the Value of Statistical Life Be Adjusted to Reflect the Nature of Different Health Risks from Pollution?. . . . . . . . . . . . . . . . . . . . . . . . . . . 21 3.7 Question: Should the Value of Statistical Life or Value per Statistical Life Year Be Applied in Monetizing Premature Mortality Costs? How Should the Value Per Statistical Life Be Estimated?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 3.8 Question: What Base Value of Statistical Life and Value per Statistical Life Year Values Should Be Used?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 3.9 Question: How Should the Base Value of Statistical Life Be Transferred between Countries?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 3.10 Question: How Should the Workforce and Working Life Be Defined?. . . . . . . . . . . . . 44 3.11 Question: How Should Deaths among Children Be Valued under the Forgone Output Approach?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 3.12 Question: What Assumptions Should Be Made about Future Income Growth? . . . . . 45 3.13 Question: What Discount Rate Should Be Used?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 4.1 Question: How Should Morbidity Costs Be Valued? . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 Figures 2.1 Worldwide Deaths Due to Household Air Pollution from Cooking with Solid Fuels, 2013. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.2 Worldwide Years of Life Lost Due to Household Air Pollution from Cooking with Solid Fuels, 2013. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 3.1 The Tradeoff between Wealth and the Probability of Survival (or Risk of Death). . . . 11 3.2 Benefit-Transfer Function from Biausque (2012) Used in World Bank Cost of Environmental Degradation Studies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 3.3 Benefit-Transfer Functions and Benchmark Value of Statistical Life Estimates. . . . . . 32 3.4 Ratio of Value of Statistical Life to Gross Domestic Product per Capita, Assuming Different Income Elasticities (e), All Years . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 3.5 Transferred Values of Statistical Life Compared with Stated-Preference Study Results in Middle-Income Countries. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 3.6 Labor Force Participation, by Men and Women in South Asia, 2013 . . . . . . . . . . . . . . 42 3.7 Labor Force Participation Rates, by Income Group, 2013. . . . . . . . . . . . . . . . . . . . . . . . 43 Tables 2.1 Disability-Adjusted Life Years for Fatal and Nonfatal Health Endpoints, Example from World Bank Study of India (2013). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.2 Global Health Effects of Exposure to Ambient and Household Pollution (millions), 2013. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 iv Methodology for Valuing the Health Impacts of Air Pollution 3.1 Examples of Value of Statistical Life Estimates Used by Public Agencies in Policy Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 3.2 Examples of Value of Statistical Life Estimates from Stated Preference Studies in Developing Countries. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 3.3 Examples of Value of Statistical Life Estimates from Meta-Analyses of Stated-Preference Studies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 3.4 Average Values of Statistical Life from Database of Stated-Preference Studies. . . . . . . 27 3.5 Examples of Guideline Value per Statistical Life Year Estimates Used by Public Agencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.6 Median Ratio of Transferred Value of Statistical Life Estimates to Gross Domestic Product per Capita, by Income Group, All Years (1990–2013). . . . . . . . . . . 34 3.7 Median Ratio of Estimated Value of Statistical Life to Expected Future Earnings for the Average Worker, All Years (1990–2013). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 3.8 Median Ratio of Estimated Value of Statistical Life to Expected Lifetime Consumption for Average Adult Ages 40–44, All Years (1990–2013). . . . . . . . . . . . . . 35 3.9 Median Ratio of Estimated Value of Statistical Life to Expected Lifetime Consumption for Average Adult Ages 45–49, All Years (1990–2013). . . . . . . . . . . . . . 36 3.10 Value of Statistical Life Estimates and Remaining Life Expectancy from Stated-Preference Studies in Organisation for Economic Co-operation and Development Database for which Age Is Reported. . . . . . . . . . . . . . . . . . . . . . . . . . 37 4.1 Share of Morbidity Costs in Total Health Costs from Ambient Air Pollution, World Bank Studies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 Methodology for Valuing the Health Impacts of Air Pollution v Acknowledgments This report represents part of a joint study by the World Bank and the Institute of Health Metrics and Evaluation at the University of Washington, Seattle (IHME) of the global eco- nomic costs of air pollution. The study, supported by the World Bank’s Pollution Management and Environmental Health (PMEH) program, aims to raise awareness about the severity of air pollution around the world and to strengthen the business case for countries to take action in reducing pollution. The report was written together by Urvashi Narain (Senior Environmental Economist, World Bank) and Chris Sall (Consultant, World Bank). The authors contributed equally to the research and writing of the report and are listed alphabetically. The report benefited greatly from discussions with Maureen Cropper (University of Maryland), Alan Krupnick (Resources for the Future), Anna Malinovskaya (Resources for the Future), Michael Brauer (University of  British Colombia), Aaron Cohen (Health Effects Institute), Mohammad Forouzanfar (IHME), Kara Estep (IHME), Kevin O’Rourke (IHME), Bjorn Larsen (Consultant, World Bank), Elena Strukova (Consultant, World Bank), and others. The report emerged from a series of working-group meetings with the community of environmental economists at the World Bank, chaired by Carter Brandon (Lead Economist, World Bank) and Kseniya Lvovsky (Practice Manager, World Bank), which focused on developing a consistent and updated framework across World Bank operations for estimating the health impacts and economic costs of air pollution. The authors are grateful for all those who contributed their valuable insights in these meetings. The authors also thank Flore Martinant de Preneuf (Senior Communications Officer, World Bank), Cindy Fisher (Publishing Officer, World Bank), Rumit Pancholi (Consultant, World Bank), and Erin Ann Barrett (Consultant, World Bank) for production management. The manuscript was edited by Marcy Gessel (Publications Professionals). Any remaining errors or omissions are the authors’ own. vi Methodology for Valuing the Health Impacts of Air Pollution 1. Introduction This discussion note is meant to inform a joint publication by the World Bank and Institute for Health Metrics and Evaluation (IHME) on the economic costs of air pollution. Air pol- lution is a global challenge and one that is acutely felt in developing countries. Illnesses caused by ambient and household air pollution claim the lives of nearly 6 million people each year, with a disproportionate majority of deaths occurring in low- and middle-income countries. The goal of the joint World Bank–IHME report is to raise awareness about the severity of this challenge and to strengthen the business case for countries to take action on reducing air pollution. To that end, the report will present new evidence on the health impacts of air pollution from the Global Burden of Disease 2013 study (GBD 2013), provide estimates of the economic costs that societies bear as a result of these impacts, and highlight the potential for economic benefits resulting from air quality management and emissions reductions. A secondary goal of the joint World Bank–IHME report is to further the development of a consistent framework for valuing the costs of air pollution in World Bank operations. Whether in assessing potential investments or in advising client governments on policy, the World Bank undertakes a variety of activities that require it to estimate the economic costs of pollution. In recent decades, tremendous progress has been made in understand- ing both the epidemiology and the economics of health risks from pollution. The joint World Bank–IHME report provides an opportunity to assess what has been done and to bring more uniformity and consistency to the way pollution costs are valued across World Bank operations. The need for greater consistency was noted, for example, in a 2014 review by Cropper and Khanna of the World Bank’s methodology for estimating damages from particulate emissions for the adjusted net savings (ANS) indicator (Cropper and Khanna 2014). The joint report with the IHME will explore several alternative approaches to valuing health risks from air pollution and will provide guidance on key issues in implementing these approaches, including when certain approaches may be more appropriate than others. This discussion note serves as a background paper for the joint report and provides a detailed discussion of the key methodological choices that must be made in valuing the health impacts of pollution. This note also makes recommendations on how these challenges can be addressed. Although past efforts to value the health impacts of pollution have greatly contributed to the discussion of challenges and potential solutions, analysts have tended to make some methodological choices on an ad hoc basis. The hope in developing this note is to bring greater clarity to what the issues are and to provide guidance on how the issues can be addressed consistently. The note provides clear recommendations where possible and a framing of issues where the literature or the context does not yet provide clarity on potential solutions. Lastly, although the primary purpose of this note is to serve as a background paper for the joint report and therefore to make recommendations for the methodological Methodology for Valuing the Health Impacts of Air Pollution 1 choices that are relevant in that context, this note also, to the extent possible, discusses how these choices may differ in different contexts (project analysis, for example). The note should nevertheless be treated as a living document—one that will need to be updated regularly as applications and consultations highlight other challenges and provide greater clarity on potential solutions. 2 Methodology for Valuing the Health Impacts of Air Pollution 2. Decisions on What Is to Be Valued: Metrics of Health Impacts The choice of what metric to use to quantify health outcomes attributable to pollution has important implications for how the economic costs of these outcomes are valued and for their magnitude. This section discusses two key questions in the choice of metrics: 1. Should both mortality and morbidity impacts be valued together in terms of disability-­ adjusted life years (DALYs) lost? 2. Should mortality impacts be measured as the number of deaths or number of years of life lost? Metrics of Health Impacts Metrics of the health burden of pollution estimated for GBD 2013 include deaths, years of life lost (YLLs), years lived with disability (YLDs), and DALYs. Deaths and YLLs are mortality metrics, YLDs measure morbidity, and DALYs are a measure of total mortality and morbidity. DALYs are commonly used in the epidemiology literature as a summary measure of health loss, taking into account both fatal and nonfatal conditions. Under certain assumptions, DALYs are equivalent to the inverse measure of quality-adjusted life years (QALYs)—a con- cept that was developed a couple of decades before the concept of DALYs and that is widely used in cost-effectiveness studies of alternative disease prevention and control strategies. These metrics (or a subset of them) are found in most studies that estimate the health impacts of pollution. In GBD 2013, for each country and cause of death, YLLs are calculated against a standard life table, which was updated by GBD researchers to account for recent demographic changes and population projections (GBD 2013 Collaborators 2015a). In calculating YLLs for GBD 2013, the GBD researchers do not discount for years of life lost at later ages, a procedure that contrasts with that used by the World Health Organization (WHO). (WHO originally calculated only discounted YLLs but now reports both discounted and undiscounted versions.) Also, the life expectancy for men and women is assumed to be the same (Murray et al. 2012).1 YLDs per health outcome are calculated as the number of prev- alent cases multiplied by a disability weight, which is “the general public’s assessment of the severity of health loss” associated with that condition (GBD 2013 Collaborators 2015b: 743). Disability weights are a relative measure of different health states, ranging from 0 (no health loss) to 1 (equivalent to death). Weights for the GBD 2013 were assigned on the basis of face-to-face surveys in Bangladesh, Hungary, Indonesia, Italy, the Netherlands, Peru, Sweden, Tanzania, and the United States as well as an open Internet survey (GBD 2013 Collaborators 2015b, 743; Salomon et al. 2012). Finally, DALYs are calculated as the sum of YLLs and YLDs. Methodology for Valuing the Health Impacts of Air Pollution 3 Valuing Mortality and Morbidity Together Although DALYs are a useful summary measure of health impacts, they may be problematic for the purposes of economic valuation. The disability weights used to calculate YLDs (and thus DALYs) are meant to be a universal, value-neutral, and context-independent measure of health loss. Although the GBD survey results indicate consistency in the rankings of dis- ability weights across countries and levels of development, disability weights are not intended to be a measure of welfare (that is, lost utility or value) or a reflection of the social conse- quences of various health states. This makes DALYs problematic for use in economic ­valuation,2 particularly when aggregated across different health outcomes. As an example,3 consider two patients suffering from vision loss. One is in a society in which corrective lenses are readily available and people with impaired vision have good job prospects. The other is in a society in which corrective lenses are not available or are prohibitively expensive and job opportunities are limited. When measured in terms of YLDs and DALYs, both peo- ple are considered to have suffered an equal reduction in health; however, the consequences of impaired vision are vastly different and would likely be valued differently by individuals in the two societies.4 From this perspective, Hammitt (2002) argues that DALYs are incon- sistent with the Willingness-to-Pay (WTP) approach (discussed in the next section) under which characteristics such as wealth and age may influence how individuals value health risks. By the same logic, DALYs are also inconsistent with the forgone output approach (also discussed in the next section) because the effect of health loss on an individual’s capac- ity to perform work may depend on her profession, the industry in which she works, and other context-specific factors, including how households adjust labor inputs in response to death or illness. Using DALYs as a metric of health impacts for economic valuation requires some additional restrictive assumptions about individual preferences (Pliskin et al. 1980, cited in Hubbell 2002; Hammitt 2002, 2013). First, longevity and the quality of life must be independent in the indi- vidual’s utility function, meaning that someone’s preference to live longer does not depend on how healthy she is and that preferences for health are not influenced by age or remaining life- span. Second, there must be constant proportionality in tradeoffs between the quantity and quality of life, such that an individual would trade living 20 years in “fair” health over 10 years in “good” health the same way she would trade 10 years in “fair” health over 5 years in “good” health. Third, the individual must be risk neutral with respect to years of life such that she will prefer living longer given equal odds and that she will not take a gamble to extend her life expectancy if the odds are not as good. Fourth, there must be independence in utility func- tions over time, meaning that an individual’s health state in previous periods will not influence her preferences in subsequent periods. Yet, empirical WTP studies have shown that (a) people disproportionately prefer reduced fatality risks over improvements in quality of health, (b) WTP for reducing mortality risks is disproportionately affected by the duration and sever- ity of illness, and (c) WTP for reducing mortality risks varies little with age or may decline only later in life (Hammitt and Haninger 2011; Krupnick 2007). Similar assumptions apply in monetizing the cost of YLLs as with DALYs. Assigning a ­ single monetary value per YLL implies that individual preferences to extend life are inde- pendent of age (for example, a 30-year-old would value a year of life the same way a 4 Methodology for Valuing the Health Impacts of Air Pollution 75-year-old would) and that WTP is proportional to the gain in life expectancy (in other words, that WTP for 10 years of additional life should be 10 times that for 1 year of life gained). As noted, economic studies of how people value changes in the probability of death suggest that WTP is not proportional to age. However, as will be discussed more later, relatively few studies have explicitly looked at how people value changes in life expec- tancy. Also, although changes in mortality risks and life expectancy are mathematically related, people may not think about them in the same way and may not attach the same monetary value to one versus the other. Despite criticisms on theoretical grounds, it is nonetheless common international prac- tice  to measure both fatal and nonfatal health outcomes in terms of DALYs in assessing their economic cost (Cropper and Khanna 2014). Recent cost of environmental degrada- tion (COED) studies by the World Bank that evaluate health outcomes in terms of DALYs include those done in Colombia (Golub et al. 2014), India (World Bank 2013), Jordan (Croitoru et al. 2010), Kosovo (World Bank 2013c), Macedonia (Meisner et al. 2015), and Nicaragua (Klytchnikova et al. 2013). Table 2.1, drawn from the India study, shows how DALYs are used to quantify premature mortality risks as well as other health endpoints. In the India study, each premature death from air pollution is assumed to result in a loss of 7.5  DALYs for adults and 70.0 DALYs for children, while each nonfatal case of chronic bronchitis leads to a loss of 4.0 DALYs (0.2 x 20 = 4). Note in table 2.1 that other nonfatal ­ health endpoints such as hospital admissions or outpatient visits to the emergency room are also translated into terms of DALYs (but these are very small, with the time lost measured in days). Table 2.1  Disability-Adjusted Life Years for Fatal and Nonfatal Health Endpoints, Example from World Bank Study of India (2013) Fatal health outcomes Average YLLs DALYs per 1,000 cases Mortality (adults) 7.5 years 7,500.0 Mortality (children) 70.0 years 70,000.0 Nonfatal health outcomes Disability weight Average duration DALYs per 1,000 cases Lower respiratory illness (children) 0.28 10.0 days 7.7 Respiratory symptoms (adults) 0.05 0.5 days 0.1 Restricted activity days (adults) 0.10 1.0 day 0.3 Emergency room outpatient visits 0.30 5.0 days 4.1 Hospital admissions 0.40 14.0 days 15.3 Chronic bronchitis 0.20 20.0 years 4,000.0 Source: World Bank 2013a. Note: DALYs represent the sum of YLLs for fatal outcomes plus years lived with disability (YLDs) for nonfatal outcomes. In the tables above, the calculation of DALYs follows the method originally used for the 1990 Global Burden of Disease (GBD) study by the World Health Organization and subsequent updates for 2000 and 2004. Under this method, YLDs = incident cases x disability weight x average duration. GBD 2010 and GBD 2013, on the other hand, calculate YLDs as the number of prevalent cases x disability weight. Disability weights for lower respiratory illness and chronic bronchitis are those used by the National Institutes of Health in the United States. Weights for other endpoints are drawn from Larsen (2004). Average years of life lost per death from air pollution are calculated by applying age weights and have been discounted at 3 percent per year. DALY = disability-adjusted life-year; YLL = years of life lost. Methodology for Valuing the Health Impacts of Air Pollution 5 Table 2.2  Global Health Effects of Exposure to Ambient and Household Pollution (millions), 2013 Risk category DALYs YLDs YLLs Ambient PM2.5 69.6 1.8 67.8 Household PM2.5 81.0 5.4 75.6 Source: GDB 2013 Collaborators 2015c. Note: Household PM2.5 is from cooking with solid fuels. PM2.5 = particulate matter less than 2.5 micrometers in Aerodynamic diameter; DALYs = disability-adjusted life years; YLDs = years lived with disability; YLLs = years of life lost. Box 2.1  Question: Should Mortality and Morbidity Impacts Be Valued Together in Terms of Disability-Adjusted Life Years? Proposal: Given the problems associated with disability-adjusted life years (DALYs) when used in economic valuation, the fact that mortality is easier to explain, and the fact that deaths (or years of life lost—YLLs) represent the bulk of the overall health burden of pollution in the Global Burden of Disease study, it is recommended that only mortality impacts be valued while measur- ing the economic costs of air pollution for the joint report. Outside the context of this report, both the morbidity and mortality outcomes will be relevant for the economic analysis in other situations. For example, and as recommended by the World Health Organization, both fatal and nonfatal health outcomes should be used in assessments of cost-­effectiveness of particular projects or policies. Or when morbidity impacts are a non- proportion of the total health impacts of air pollution, researchers may want to value trivial ­ DALYs and not just YLLs, problems associated with DALYs when used in economic valuation notwithstanding. It is also important to note that morbidity represents a very small share of the total health impacts of air pollution in the context of the GBD analysis. The shares of YLLs and YLDs in total DALYs from exposure to ambient and household PM2.5 (particulate matter less than 2.5 micrometers in aerodynamic diameter) are shown in table 2.2, as estimated for GBD 2013. Because YLLs contribute 97 percent of DALYs from ambient PM2.5 exposure and 93 percent of DALYs from household air pollution exposure, valuing health losses in the form of DALYs yields practically the same results as valuing premature mortality risks in the form of reduced life expectancy (YLLs).5 (See box 2.1.) Measuring Premature Mortality Theoretical issues aside, whether premature mortality is measured in terms of excess deaths or lost life expectancy raises important ethical questions about how health impacts suffered by the very old and very young should be weighted. According to the GBD 2013 estimates, 62 percent of deaths attributed to ambient PM2.5 and 59 percent of deaths from household air pollution in 2013 were suffered by the over age 65 population. Children under the age 6 Methodology for Valuing the Health Impacts of Air Pollution of 15, meanwhile, accounted for only 4 percent of deaths from ambient PM2.5 and 8 percent of deaths from household air pollution in 2013. When expressed in terms of YLLs, the pro- file of health impacts looks much different. YLLs among the over age 65 population repre- sented 34 percent of total YLLs due to ambient PM2.5 and 30 percent of YLLs due to household air pollution. About 16 percent of YLLs from ambient PM2.5 and 26 percent of YLLs from household air pollution were among children under the age of 15. The differ- ences in the age profile of premature deaths and YLLs due to household air pollution are illustrated in figures 2.1 and 2.2. Although YLLs per death are greater for young people, they are not the same as life expec- tancy, and their usefulness may be limited for the purposes of economic valuation. Because YLLs are calculated against a standard life table, YLLs per death are assumed to be the same for similar-age people around the world, regardless of actual life expectancy. For example, life expectancy in Swaziland was 21 years at age 50. The calculated number of YLLs per death for a person of that age is 35 years for all countries. Thus, valuing the economic benefit of an avoided death from air pollution due to ischemic heart disease for a 50-year-old in Swaziland using YLLs may lead to overestimates because YLLs ignore competing risks and use a lower overall risk profile for all causes of death in that country. (See box. 2.2.) Figure 2.1  Worldwide Deaths Due to Household Air Pollution from Cooking with Solid Fuels, 2013 700 600 500 Deaths (thousands) 400 300 200 100 0 s 19 29 39 49 59 69 79 4 9 14 24 34 44 54 64 74 r nt de to to fa to to to to to to to to to to to to to to ol 1 5 In 15 25 35 45 55 65 75 10 20 30 40 50 60 70 d an 80 Age at time of death Lung cancer Lower respiratory infections Ischemic stroke Ischemic heart disease Hemorrhagic stroke COPD Source: GBD 2013 Collaborators 2015c. Note: COPD = chronic obstructive pulmonary disease. Methodology for Valuing the Health Impacts of Air Pollution 7 Figure 2.2  Worldwide Years of Life Lost Due to Household Air Pollution from Cooking with Solid Fuels, 2013 16 14 12 Years of life lost (millions) 10 8 6 4 2 0 ts 4 9 14 19 24 29 34 39 44 49 54 59 64 69 74 79 r de n to to fa to to to to to to to to to to to to to to ol 1 5 In 10 15 20 25 30 35 40 45 50 55 60 65 70 75 d an 80 Age at time of death Lung cancer Lower respiratory infections Ischemic stroke Ischemic heart disease Hemorrhagic stroke COPD Source: GBD 2013 Collaborators 2015c. Note: COPD = chronic obstructive pulmonary disease. Box 2.2  Question: Should Mortality Impacts Be Measured as Number of Deaths or Number of Life Years Lost? Proposal: The literature offers almost no guidance as to whether it is better to measure number of deaths or reduced life expectancy when evaluating the economic costs of premature mortality from air pollution. Both deaths and years of life lost (YLLs) are common in the medical litera- ture, though the usefulness of YLLs versus actual life expectancy increases may be limited for the purposes of economic valuation. The choice between valuing deaths or reduced life expectancy in assessing the burden of disease from air pollution hinges on whether impacts among children or older people should be weighted more heavily in the calculation. This is as much an ethical ques- tion as it is an economic one. For the purposes of assessing the global burden of disease, separate calculations will be made in valuing both deaths and reduced life expectancy. In other contexts, the choice of metric may be guided by the context of the project or policy at hand. 8 Methodology for Valuing the Health Impacts of Air Pollution Notes 1. The choice not to apply age weighting or to differentiate life expectancy for men and women in calculating YLLs represents a departure from common practice before the GBD 2010. See Haagsma ­ et al. 2014. 2. The exception is with cost-effectiveness analysis. For example, WHO has recommended the use of  DALYs as a standard parameter for comparing the cost-effectiveness of health interventions and policies (such as by comparing implementation costs per avoided DALY) (Tan-Torres Edejer et al. 2008; WHO 2008). U.S. government agencies have also used DALYs in assessing the cost-­ effectiveness of policies that reduce health risks (see Robinson 2007). 3. The following hypothetical is posed by Voigt and King (2014). 4. The example by Voigt and King (2014) points to additional concerns about equity. For a relevant discussion of equity and the choice of health metrics in economic analysis, see Whitehead and Ali (2010). 5. This point is raised by Cropper and Khanna (2014). Methodology for Valuing the Health Impacts of Air Pollution 9 3. Approaches to Valuing the Economic Costs of Premature Mortality This section considers the different approaches to valuing the economic costs of premature mortality and the methodological challenges in implementing them. In particular, it tries to provide answers to the following questions: 1. When should the willingness-to-pay (WTP) approach be used versus the forgone-output approach? 2. Should value of statistical life (VSL) be adjusted for age differences, health status, income, or risk characteristics? 3. How should deaths among children be valued? 4. Should VSL or value per statistical life year (VSLY) be used? How should VSLY be estimated? 5. What base VSL should be used, and how should the VSL be transferred between countries? 6. How should working age, future growth rates, and discount rates be defined when imple- menting the forgone-output approach? Willingness-to-Pay Approach By causing illness, pollution is a detriment to the things that individuals fundamentally value, including consumption, leisure, good health, and life itself. Individuals make tradeoffs between things of value every day. For example, they sacrifice leisure by going to work so that they can increase consumption. In valuing the increased risk of death due to pollution, the task of the economist then is to weigh the tradeoffs that individuals would be willing to make to reduce their chances of dying. What the economist seeks to measure is the individual’s willingness to pay (WTP) in exchange for a reduced chance of dying—or willingness to accept (WTA) com- pensation in exchange for an increased fatality risk—to maintain the same level of utility.1 Graphically, the marginal rate of substitution between wealth and the probability of survival (or risk of death) is depicted in figure 3.1 as the slope of an individual’s indifference curve (see Cropper et al 2011, 315–316). Supposing that pollution reduces the probability of survival from P1 to P2, the cost to the individual would be equal to the vertical distance from W1 to W2. The economist would monetize the loss of value suffered by the individual due to pollution by assigning a dollar amount to ΔW, which is what the individual would be willing to pay to avoid such an increase in mortality risk and thereby sustain the same level of utility. The value of statistical life (VSL) represents an aggregate of individuals’ willingness to pay (WTP) for marginal reductions in their mortality risks (R). It is not the value of any single person’s life or death, nor does it represent a society’s judgment as to what that value should be. In figure 3.1, for small changes in survival probability, the VSL can be approximated as VSL = ΔW/ΔP. As a stylized example of how the VSL is derived and how it relates to the cost of pollution, suppose that pollution increases the odds that any given individual would die in a ­ year on average from 2 in 10,000 to 3 in 10,000.2 This is ΔP. Probability implies that among 10,000 people, 1 person dies each year due to the increased risk from pollution. Suppose fur- ther that individuals in this society are willing to pay US$30 on average to reduce their risk of 10 Methodology for Valuing the Health Impacts of Air Pollution Figure 3.1  The Tradeoff between Wealth and the Probability of Survival (or Risk of Death) Indifference curve Wealth W1 ∆W W2 ∆P 0 P1 P2 1 Probability of survival (P = 1 – risk of death) Source: World Bank, adapted from Cropper et al. 2011. Note: Δ = change. death by a margin of 1 in 10,000. This is ΔW. Summing this average WTP value over 10,000 individuals, the value of one statistical life would be equal to US$300,000. Empirically, the VSL is derived from studies of people’s stated preferences or revealed prefer- ences (actual behavior). In stated-preference studies, individuals are presented with hypothet- ical scenarios and asked to make choices. For example, they might be asked about how much they would be willing to pay for a new medical treatment that would reduce their risk of dying from cancer, or they might be asked to choose between paying for different programs to reduce the risk of traffic fatalities in their city. In any case, researchers must: (a) present respondents with information about their baseline risk of death and the nature of the risk; (b) describe the context; (c) define how much their risk will be reduced (or increased) and who exactly will benefit from the risk reduction (for example, adult individuals or their children); (d) clarify the time period over which risks will be reduced (for example, now or 10 years later); (e) explain who will be expected to pay (for example, the individual or general public); and (f) provide other relevant information (Kling et al. 2012). For the purposes of evaluating the costs of pollution, the appeal of stated-preference studies is that they enable researchers to investigate the preferences of a broad array of individuals, including those most affected by pollution, and to ask about health risks similar to those actu- ally experienced by people exposed to pollution. The main disadvantage of such studies is that Methodology for Valuing the Health Impacts of Air Pollution 11 they are based on hypothetical situations in which respondents may disregard their actual budget constraints or make spurious choices without real consequences (Cropper et al. 2011; OECD 2012). Also, respondents may be asked about small changes in survival probability that are difficult to comprehend, and their choices may be influenced by how those choices are embedded (for example, whether a hypothetical program is presented by itself, in comparison with another program, or bundled in a package) or the order in which choices of different magnitude are posed. Whether such differences lead to systematic bias or unreliability in VSL estimates derived from stated-preference studies is open to debate (see Carson 2012; Hausman 2012; Kling et al. 2012). For some goods, individuals tend to overstate their WTP when asked in surveys; for reducing illness-related risks, however, previous reviews have found that indi- viduals often understate their WTP when survey results are compared with revealed market behavior (Cropper et al. 2011; Kochi et al. 2006).3 In revealed-preference studies, researchers investigate individuals’ WTP by observing their behavior in actual markets. The most common approach in estimating the VSL has been to examine differences in wages for more or less risky occupations (see Viscusi 2004). This is the approach that the U.S. Environmental Protection Agency (EPA) has primarily taken in valuing mortality risk reductions (Robinson 2007). Other approaches may value pollution costs by  looking at differences in real estate prices between more and less polluted areas (see Zheng et al. 2014), or gauge individuals’ willingness to spend on protective measures such as face masks that limit their exposure to air pollution (see Zheng et al. 2015). Although revealed-preference studies of wage-risk differentials are often seen as more credi- ble, they face drawbacks that may limit their usefulness for valuing pollution-related health risks. Wage-risk studies used to estimate the VSL assume that workers are well informed about their actual risk of on-the-job injury and that they are in a competitive labor market in which they are perfectly mobile and can negotiate higher wages with employers for taking riskier jobs. These conditions do not hold for workers in many developing countries. Furthermore, because wage-risk studies focus on working adults, they do not capture the preferences of people outside the labor force “whether because of illness, retirement, child raising, or other reasons” (Cropper et al. 2011, 317). Studies of wage differentials for morality risks also face a host of econometric issues, such as determining how to account for unobserved characteris- tics of workers that incline some to take more or less risky jobs for reasons other than pay (Cropper et al. 2011, 317; Kniesner and Viscusi 2005). Other Approaches to Valuing Health Impacts Alternative approaches to valuing health impacts include the cost of illness (COI) approach. The COI approach seeks to account for the direct and indirect financial costs of death and i ­llness. Direct financial costs include ambulance costs, medical expenses, wages lost by sick workers, and time spent caring for ill family members, among others. Indirect financial costs include such costs as productivity losses to firms from declining customer service quality and missed business opportunities because of sick employees. The COI approach could be used to value the financial costs of health impacts arising from either acute or chronic exposure. As an example of the COI approach in practice, the U.K. Department for Transport includes direct and indirect financial costs in valuing the costs of road traffic fatalities. It counts these financial costs apart 12 Methodology for Valuing the Health Impacts of Air Pollution from what the agency calls “human costs,” which “are based on willingness to pay to avoid pain, grief and suffering to the casualty, relatives and friends, as well as intrinsic loss of enjoyment of life” (U.K. DFT 2012, 2). The COI approach is also frequently used in measuring the costs of morbidity (nonfatal health outcomes), as will be discussed in greater detail in a later section. The forgone output (FO) approach, sometimes referred to as the human capital approach, is related to the COI approach. It is an income-based approach and equates the financial cost of premature mortality with the present value of forgone lifetime earnings. The Ministry of Social Development (MDS) in Chile, for example, has adopted the FO approach in its guidelines for valuing premature mortality. MDS estimates the average cost of an early death at UF 12,868 (Chile MDS 2014). A common criticism of this approach is that it excludes losses suffered by individuals outside the labor force, including retirees. The income-based approach also raises questions about how unpaid work that contributes to the economy, such as subsistence farm- ing and domestic activities, should be valued. An adjusted approach seen in countries such as China is to value mortality among all individuals regardless of participation in the labor force using gross domestic product (GDP) per capita discounted over remaining life expectancy (see World Bank–SEPA 2007). Although the FO approach accounts specifically for reduced labor output due to the loss of individuals’ time endowment, computable general equilibrium (CGE) modeling may be applied to measure the wider productivity losses to the economy. These models simulate how the effects of premature deaths and nonfatal health outcomes are propagated across sectors and over time, leading to changes in overall economic activity and consumption. For example, Matus and coauthors (2012) use an expanded version of the Emissions Prediction and Policy Analysis (EPPA) model developed by the Massachusetts Institute of Technology to simulate the broader economic impacts caused by air pollution exposure in China as a result of divert- ing labor from productive activities and reducing the time for leisure. Tracking the effects of exposure to PM10 and ozone in major cities from 1970 to 2005, they estimate that air pollution reduced annual consumption in China in 2005 by 7 percent, or US$69.0 billion (1997 prices). Including welfare losses—estimated in terms of lost leisure time and valued at the wage rate, not the WTP to avoid mortality risk—they figure total economic losses due to air pollution in 2005 to be US$184.5 billion. Similar analyses using the health-adjusted EPPA model have been done for the United States (Matus et al. 2008) and Europe (Nam et al. 2009). Analysis by the World Bank in India has used CGE modeling to estimate the GDP impacts of reducing air pollution emissions through the use of a tax on coal or emissions (World Bank 2013b). Use of the Willing-to-Pay–Based Approach versus the Forgone Output Approach The WTP-based approach is best suited for analyses of economic welfare and has become the standard approach in high-income countries for valuing mortality risks associated with pollution (see Cropper 2000; OECD 2012; Viscusi 1993). The forgone output approach ­ continues to be used for policy analysis by public agencies in some countries such as Chile and China. Both the WTP-based approach and forgone output approach may be applicable for cross-country comparisons of health impacts and costs or other applications depending Methodology for Valuing the Health Impacts of Air Pollution 13 on the scope of the application. In theory, the present value of lifetime earnings should be less than the VSL, which captures the utility derived from intangibles such as being alive and spending time with loved ones in addition to consumption (Hammitt and Robinson 2011). The forgone output provides an income-based measure of health impacts and may be more appropriate for financial analysis and accounting, for example, in calculating pollution costs as a component of adjusted net savings (ANS) within the System of National Accounts frame- work. ANS, or “genuine savings,” is a measure of the change in the value of a nation’s assets, including manufactured capital as well as natural and human capital (see Hamilton and Clemens 1999; World Bank 2005, 2011). Positive savings represent an investment in future well-being as a nation accumulates the assets needed to drive economic growth and at least sustain current levels of consumption. Within the ANS framework, premature mortality due to pollution represents a disinvestment in a nation’s human capital stock. As with the degrada- tion of other forms of capital, the value of this disinvestment would be valued according to an expected loss of income over the lifetime of the asset.4 Losses valued under a FO approach are typically much smaller than those valued under a WTP approach. Because of this, in COED studies by the World Bank, the FO approach is still frequently used to estimate the lower bound of losses in economic welfare due to pollution.5 FO may represent to the lower bounds of WTP so long as average utility of consumption is greater than marginal utility of consumption, which is true for strictly concave utility func- tions (Cropper 2000). However, although welfare-based and income-based measures typically provide a range of health costs, they should not be averaged and presented as a single measure. Also, in COED studies, welfare-based estimates of damages should not be added to other pro- ductivity-based measures of environmental degradation that are presented as a percentage loss of GDP, as GDP is an income-based measure. If presenting the welfare-based estimate as a percentage of GDP, appropriate language should be added to explain to readers how they should interpret this measure. Finally, both types of approaches can be used in project analysis: income-based measures in proj- ect financial analysis, and welfare-based measures in project economic analysis. (See box 3.1) Box 3.1  Question: When Should the Willingness-to-Pay Approach Be Used Instead of the Forgone Output Approach? Proposal: Both the willingness to pay (WTP) approach and forgone output (FO) approach may be applicable for cross-country comparisons of health impacts or other applications, depending on the scope of the applications. Estimates based on both approaches will be provided for the joint study. FO is more appropriate for financial analysis and accounting, as in calculating pollution costs as a component of adjusted net savings or in project financial analysis. The WTP approach, on the other hand, should be used in project economic analysis. Both approaches can be used in cost of environmental degradation studies. Although welfare-based and income-based measures typically provide a range of health costs, they should not be averaged and presented as a single measure. Welfare-based estimates of health damages should also not be presented as a percentage of gross domestic product without appropriate language to clarify that the comparison is only being made for convenience as a benchmark of magnitude. 14 Methodology for Valuing the Health Impacts of Air Pollution Factors Influencing the Value of Statistical Life A variety of factors influence the VSL, including the income, wealth, age, life expectancy, cul- tural beliefs, and health status of the affected individuals who make up a society. Moreover, as Cropper and coauthors write, an individual has no single VSL: “Different people have different preferences, and for an individual, the value of a 1-in-10,000 risk reduction may depend on whether death results immediately from injury or from a lingering illness, whether it is caused by a hazard viewed as particularly fearsome, whether one is rich or poor or old or young, and other factors” (Cropper et al. 2011, 316). The following section describes salient population and risk characteristics that affect the VSL. Population characteristics discussed include age, health status, and income. Age The relationship between age and the value of mortality risk reductions is a complex and con- tentious issue. As noted above, the age question is important because of how the health impacts of air pollution are distributed. Experts and public agencies around the world are split as to whether the VSL should be adjusted in assessing health benefits and costs for people of different ages. The US EPA, OMB, and DOT have discontinued the practice of discounting the VSL for the elderly or making other adjustments for age in valuing mortality risks, with the expert panels advising these agencies citing a lack of consistent empirical support (Morgan et al. 2007; NRC 2008; U.S. EPA 2014; U.S. OMB 2003). The Organisation for Co-operation and Development (OECD) also disfavors making age adjustments to the VSL (OECD 2012). Meanwhile, the Canadian ­ government has regularly applied a 25 percent discount to monetize reduced mortality risks for the over age 65 population (Chestnut and Civita 2009). The European Commission Directorate-General for Environment (EC DG Environment) has recommended that the VSL be discounted by 30 percent for the elderly (Hurley et al. 2005). Previous studies by the World Bank of the health costs of pollution have also applied age discounting to value prema- ture mortality, either directly or implicitly through the use of a constant VSLY to monetize DALYs (Hammitt and Robinson 2011). If the VSLY is used to monetize health impacts rather than the VSL, the U.S. Office of Management and Budget (OMB) has recommended a higher VSLY be used for older adults (U.S. OMB 2003). The effect of age on the VSL is theoretically indeterminate (Morgan et al. 2007; Hammitt 2007; Hammitt and Robinson 2014). As people age, their life expectancy shortens, but they may also be wealthier. As their preferences evolve, they may place a greater value on longevity or pre- serving their health. Stated-preference studies have found some evidence of a lower VSL only among the very old.6 For example, in a survey by Alberini and coauthors (2004), respondents in Canada who were over the age of 70 were willing to pay about 33 percent less than younger respondents to reduce their risk of dying over the next 10 years. Respondents in the United States over the age of 70 were willing to pay about 20 percent less than younger respondents, although the difference in WTP for the U.S. sample was statistically insignificant. Similarly, in another stated-preference study, Krupnick and coauthors (2010) observed that WTP among respondents over the age of 70 in the Chinese cities of Shanghai, Nanning, and Jiujiang declined Methodology for Valuing the Health Impacts of Air Pollution 15 by 20 percent for contemporaneous risk reductions, after controlling for differences in health and other factors. Reviewing the results from more than 20 other stated-preference studies, Krupnick concludes that “the evidence for a senior discount is too thin to be relied on for pol- icy purposes” (Krupnick 2007, 262). Blomquist and coauthors (2011) posit that the relationship between age and the VSL may be nonmonotonic. The authors conducted two separate surveys in the United States in which adults weighed the tradeoffs between drugs that better control asthma symptoms but also increase fatality risks. In the first survey, parents with asthmatic children between the ages of 4 to 17 were asked to make choices to treat their children. In the second survey, adults in the general population ages 23 and older were asked about their own preferences. The survey results show that the VSL for children is about 1.7 times higher than the average adult VSL. The VSL gradually declines for young people between the ages of 4 and 30, rises for adults over 30, and then declines again for people in retirement age. Another study by  Hammitt and Haninger (2010) estimated that the WTP of parents to reduce risks to their children is 1.5 to 2.0 times larger than the WTP to reduce their own mortality risks or  the risks faced by other adults. Not only does the higher VSL for young children reflect  the  altruistic concerns of parents, but also it is influenced by the fact that parents tend  to be more affluent than young people in their teens and twenties with lower VSLs (Viscusi 2010). The practice of assigning higher values to mortality risk reductions among children has tended to be less controversial than discounting the VSL for older adults (Hammitt and Robinson 2014; Viscusi 2010). Normally, the VSL should reflect the aggregate of affected individuals’ own WTP for reducing their risk of death, but children represent a special case. Children have little concept of risk and are usually not held responsible for making decisions affecting their well-being or chance of death. The U.S. EPA has endorsed the use of parent surveys to value costs and benefits to children’s health, finding that “the parental perspective provides value estimates from individuals likely to have the child’s best interest at heart, incorporates the household context into the estimation approaches, and can be implemented using a variety of valuation techniques” (U.S. EPA 2003, 2–11). Although the EPA does not find sufficient evi- dence to derive age-specific VSLs for children (U.S. EPA 2014), it does suggest that these val- ues are systematically higher (U.S. EPA 2003). The U.S. OMB recommends the use of monetary values that are at least as large as those used for adults (U.S. OMB 2003). The OECD has rec- ommended that the VSL for children should be 1.5 to 2.0 times higher than the mean adult VSL when assessing public health risks that mainly affect children or evaluating programs that specifically target children (OECD 2012). Virtually all of the COED studies by the World Bank in recent years have forgone the use of a VSL in monetizing impacts on children’s health from pollution and instead have taken a for- gone output approach.7 The commonly cited reason for this approach is that no VSL studies have been done for children, and children have poorly defined risk preferences. An exception is the 2007 study of pollution damages in China by the World Bank and the Chinese environ- mental protection authority, which applies the adult VSL to value child mortality following U.S. EPA guidance (World Bank—SEPA 2007). Under a forgone output approach, children’s deaths are monetized as the discounted value of their expected lifetime earnings (see box 3.3 and box 3.11.) 16 Methodology for Valuing the Health Impacts of Air Pollution Box 3.2  Question: When Using the Value of Statistical Life, Should the Value of Statistical Life Be Adjusted for Age Differences in the Adult Population Affected by Air Pollution? Proposal: Given that the effect of age on VSL is theoretically indeterminate and that the empirical evidence is mixed, experts and public agencies around the world are split as to whether the VSL should be adjusted in assessing health benefits and costs for people of different ages. Consequently, conditional on using VSL, an age-independent VSL will be used for adults, though sensitivity analysis may be done to compare the effect of discounting the VSL by 20–30 percent for persons aged 70 and older, corresponding roughly to the differences in WTP observed by Alberini and coauthors (2004), Krupnick and coauthors (2010), and the rate used by the European Commission Directorate-General for the Environment. Note that use of a VSLY, as opposed to the VSL, implic- itly discounts health impacts among older people with shorter remaining lifetimes. The choice to use an age-invariant VSL in other contexts will depend on the nature of the health risk being evaluated and whether a particular policy or project targets people in a specific age group. Box 3.3  Question: How Should Deaths among Children Be Monetized? Proposal: The VSL for children should be set according to what adults are willing to pay to reduce risks to children in the country where the health impacts are taking place. If such an estimate is not available, then it is recommended that the default be to use the adult VSL as the existing empirical evidence on what parents are willing to pay is mixed. Furthermore, use of the forgone income approach to value deaths among children while using a WTP-based approach to value deaths among adults is also not recommended as this implicitly discounts the value of children’s deaths, and also mixes welfare-based measures and income-based measures in a single indicator. Health Status As noted by the U.S. EPA, individual health status “is an especially relevant factor for valuation of environmental risks because individuals with impaired health are often the most vulnerable to mortality risks from environmental causes. For example, particulate air pollution appears to disproportionately affect individuals in an already impaired state of health” (U.S. EPA 2014, B-5). However, the empirical evidence for how individual health status influences the VSL is mixed at best. A study by Tsevat and coauthors (1998) of elderly hospitalized patients found that patients placed a higher premium on extending life than did medical care providers and family members. Stated-preference studies in the United States and Canada by Krupnick and coauthors (2002) and Alberini and coauthors (2004) showed that individuals in poor health did not have a significantly different WTP for mortality risk reductions than others. A more recent survey of people’s WTP to avoid air pollution in Beijing and Shanghai similarly did not reveal any difference in WTP among respondents suffering from respiratory disease versus those who reported good health (Tan and Zhao 2014). As with the influence of age on the VSL, the EPA finds insufficient theoretical and empiri- cal support to warrant the use of a health-adjusted VSL for policy purposes (U.S. EPA 2014). Methodology for Valuing the Health Impacts of Air Pollution 17 This advice mirrors that of earlier expert panels (Morgan et al. 2007; NRC 2008). The OECD also has avoided making adjustments for varying health status between countries in its esti- mates of the costs of air pollution (OECD 2014; WHO-OECD 2015). None of the COED studies by the World Bank have adjusted the VSL for an affected population’s general state of health. (See box 3.4.) Income Both theory and the available body of empirical evidence support the notion that the value attached to mortality risk reductions increases for individuals with higher incomes. For equity concerns, governments have typically not adjusted VSLs to reflect cross-sectional differences between poorer and richer citizens. In country-level policy analysis, a single VSL is typically used for the entire country despite pronounced differences in income between people living in different regions (see World Bank-SEPA 2007). The reasoning behind this choice is that the costs of pollution should be aggregated at the societal level at which the collective economic burden is typically shared and the policies to control pollution are typically made: the nation state (WHO-OECD 2015). Although the notion that the VSL should rise with income is commonly accepted, the extent to which the VSL varies with income is the subject of debate. Lindhjem and coauthors (2011) conducted a meta-regression analysis for the OECD of VSL estimates from more than 80 stated-preference studies done in 38 countries between 1970 and 2008. They estimate an ­ income elasticity of between 1.17 and 1.31 for the entire sample set.8 This finding suggests that WTP for mortality risk reductions would increase more quickly than would income and that the ratio of the VSL to per capita income would be greater in high-income countries than in low-income or middle-income countries.9 After screening for studies that meet certain quality standards,10 however, the elasticity estimated by Lindhjem and coauthors falls to between 0.77 and 0.88, implying that the VSL is income inelastic. Imposing further restric- tions to include only studies that satisfy validity tests reduces the elasticity to between 0.33 and 0.75.11 Although this restricted sample is composed mainly of VSL estimates from stud- ies in high-income OECD countries, it does include estimates from a number of middle-​ income countries. Attempting to eliminate the effect of methodological differences, Lindhjem and coauthors proceed to use an even more restricted sample of estimates drawn from stud- ies that employ the best-practice survey instrument developed by Krupnick and coauthors (2002). This set yields income elasticities in the range of 0.30 to 0.44. Using the same ques- tionnaire, a more recent WTP study by Hoffmann and coauthors (2012) in Ulaanbaatar, Mongolia, estimates an average within-sample income elasticity of 0.38 to 0.60 for WTP for Box 3.4  Question: Should Mortality Risks Be Valued Differently for People in Different Health States? Proposal: Although individuals in poor health may be more susceptible to the health effects of pollution, theoretical and empirical evidence are insufficient to suggest that their WTP for reduced mortality risk is any different. The VSL, therefore, need not be adjusted to reflect differ- ences in health status. 18 Methodology for Valuing the Health Impacts of Air Pollution contemporaneous risk reductions and 0.38 to 0.52 for longer-term risk reductions. Also using the same survey instrument, Krupnick and coauthors (2010) estimate an even lower average income elasticity in the range of 0.15 to 0.25 for respondents in the Chinese cities of Shanghai, Jiujiang, and Nanning. Others argue that income elasticity of the VSL should be greater than 1, as suggested by the literature on risk aversion (see Kaplow 2005). Kniesner and coauthors (2010), for example, use data on wages, occupational risks, and demographics in the United States. Rather than looking at the average effect of wage difference, they employ a quantile regression model to capture how people at different income levels value mortality risk reductions. They find that income elasticity ranges from 1.23 for people in the top 10th percentile of wages to 2.24 for people in the bottom 10th percentile. In a meta-analysis of VSL estimates from stated-preference stud- ies, Biausque (2012) takes a similar methodological approach to explore differences in income elasticities estimated from studies in poorer versus richer countries. He estimates income elas- ticities above 1 for countries in the bottom 40th percentile of income and below 1 for countries at higher income levels. Viscusi (2010) argues that it is reasonable to assume an elasticity of at least 1 in assessing costs and benefits to the general public—well above the values of 0.5–0.6 estimated from wage-risk studies of people of working age and in decent health (Aldy and Viscusi 2007). The paucity of contingent valuation studies in low-income countries makes it difficult to say whether the observed relationship between the VSL and income in middle-income and high-income countries also holds for poorer countries. Robinson and Hammitt (2009), Hammitt and Robinson (2011), and Cropper and Sahin (2009) argue that basic economic reasons explain why the income elasticity of WTP for risk reductions is likely higher in low-income countries than in richer countries and may be greater than one. In low-income countries, survival probabilities are lower, private discount rates are higher, and a larger share of consumption expenditure is devoted to satisfying basic needs. In such an environ- ment where baseline health risks are often higher and basic social services and safety nets lacking, individuals’ attitudes toward risk are also likely different. Given that, it seems rea- sonable that WTP for relatively small reductions in mortality risk, such as those commonly posed in stated-preference studies, would represent a smaller percentage of income for people in lower-income countries than in more affluent countries such as the United States. ­ Comparing the ratio of the VSL to per capita income as estimated from previous meta-­ analyses of empirical WTP studies in developing countries, Cropper and Sahin (2009) find this to indeed be the case. Further support for the claim that income elasticity of the VSL is higher in developing coun- tries is provided by longitudinal studies. These studies gauge how WTP for mortality risk reductions has varied at different stages of economic development. For example, using wage- risk data from 1982 to 1997 in Taiwan, Hammitt, Liu, and Liu (2000) estimate the income elasticity of the VSL to be in the range of 2.0–3.0. Using wage-risk data from 1940 to 1980, Costa and Kahn (2004) estimate the elasticity for the United States at 1.5–1.7. There are no longitudinal estimates of the VSL from stated-preference studies. Moreover, little consensus exists on how the VSL should be adjusted to account for the effect of income growth over time. The U.S. EPA has recommended making an adjustment for Methodology for Valuing the Health Impacts of Air Pollution 19 Box 3.5  Question: Should the Value of Statistical Life Be Adjusted for Differences in Income? Proposal: For the purposes of the global assessment, for which most assessments are at the national level, a single country-level VSL will be used. For country-level assessments, subnational VSL may be applied. This decision should be made in consultation with local stakeholders. The VSL should, however, be adjusted for cross-sectional differences in income between coun- tries as well as for income growth over time. Although the empirical evidence from many stated-­ preference studies suggests an income elasticity value of less than one, a low elasticity produces estimates of VSL for low-income countries that are often not credible. Other empirical studies have provided support for elasticity values above one. Consequently, it is suggested that a range of elasticity values be considered, especially for situations in which large differences in income exist among the affected population and in which mortality risks are being valued over a long time period. Especially when the VSL is transferred to a different context from the one in which it was originally estimated, the choice of elasticity value could have a material influence on the resulting cost estimates. This choice should be communicated clearly and discussed with local stakeholders involved in the analysis. For the World Bank–IHME report, differentiated elasticity values will be assumed for countries of different income levels. The value of 0.8 recommended by the OECD (2012, 2014) will be assumed for high-income countries, with a range of 0.6–1.0 for sensitivity analysis. For low-­ income and middle-income countries, a higher central value of 1.2 will be assumed, with a range of 1.0–1.4 for sensitivity analysis. The choice of elasticity values for the global assessment is tied closely to the choice of the base VSL and will be discussed in greater detail in a later section of this report on benefit-transfer methods (see “Transferring base estimates of the VSL to value health impacts in other countries”). income growth when assessing the potential benefits of reducing mortality risks in future years (U.S. EPA 2014). A prospective study done by the EPA in 2000, for example, examined the benefits of the Clean Air Act from 1990 to 2010. In the study, the agency assumed a wide range of income elasticities (0.08, 0.40, and 1.00). The U.S. Department of Transportation, meanwhile, assumes an income elasticity of 1 to project the VSL in assessing costs and benefits in future years (U.S. DOT 2015) (see box 3.5). Risk Characteristics Whether different estimates of the VSL should be applied for different risks or causes of death is another contentious issue. Dimensions of risk that affect the VSL include whether a risk is voluntary or involuntary, ordinary or catastrophic, immediate or delayed, natural or constructed, old or new, controllable or uncontrollable, necessary or unnecessary, and occasional or continuous (U.S. EPA 2010, B-3). For example, in the European Union and United Kingdom, economic assessments by public agencies may use higher values to mon- etize cancer risks because of the dread that people feel toward cancer (cited in Cropper et al. 2011, 330). Empirical evidence of a “cancer premium” is mixed. Hammitt and Haninger (2011) find little difference in WTP for reduced risk of death from cancers, other diseases, and traffic accidents. Alberini and Ščańy (2013), on the other hand, observe that the VSL 20 Methodology for Valuing the Health Impacts of Air Pollution Box 3.6  Question: Should the Value of Statistical Life Be Adjusted to Reflect the Nature of Different Health Risks from Pollution? Proposal: A single VSL will be used to value mortality from air pollution regardless of the cause of death. There is not enough evidence to suggest that values other than the default should be used. The same holds for the size of the risk and for baseline risks. increases with dread, exposure to risk, and respondents’ perceptions of their baseline level of risk. They estimate that the WTP for avoided cancer deaths and respiratory illnesses is significantly higher than that for motor crashes. In a nationally representative stated-­ preference study in the United States, Viscusi, Huber, and Bell (2014) found evidence of a  21 percent premium in WTP for avoiding the latent risk of cancer compared with acute fatality risks. As standard practice, the United Kingdom typically doubles the VSL in  estimating the economic benefits of asbestos removal compared with traffic safety (U.K. Treasury 2011, 62). WTP should increase with the magnitude of risk, and Cropper and coauthors (2011) rec- ommend that this assumption be used as a minimum criterion for acceptability of VSL studies. In theory, WTP should increase proportionally to the size of the risk reduction, but this has not been borne out by many empirical studies. For example, Alberini and coau- thors (2004) find that the mean WTP among U.S. respondents of reducing the annual risk of mortality by a factor of 10-in-10,000 is about 1.6 times that for a 5-in-10,000 risk reduction. Hoffmann and coauthors (2012) estimate that the mean VSL in Ulaanbaatar, ­ Mongolia, for reducing annual mortality risks by 10 in 10,000 is only 1.15 times the WTP for a risk reduction of 5 in 10,000. Individuals living in riskier environments, with lower baseline survival rates and other kinds of competing risks, may value health risks differently from those in less risky environments, all else being equal. Eeckhoudt and Hammitt (2001) have shown that the higher level of background mortality or competing financial risks faced by an individual, the lower the WTP to reduce any one risk. If the competing risk is much larger than the target risk, the result may be a “why bother?” effect (Eeckhoudt and Hammitt 2001, 274). Because the indi- vidual is unlikely to survive the competing risk anyway, WTP to reduce a target risk may be much lower. However, for smaller changes in baseline mortality rates, such as those com- monly posed by competing environmental risk factors, the overall effect is usually negligible. (See box 3.6.) The Value per Statistical Life Year In the United States and other OECD countries, the WTP-based approach using the VSL is the standard method for monetizing the economic costs of premature mortality (WHO-OECD 2015). An alternative approach, which is found in international practice, is to monetize pre- mature mortality in the form of reduced life expectancy using a value per statistical life year Methodology for Valuing the Health Impacts of Air Pollution 21 (VSLY) (Cropper and Khanna 2014).12 The VSLY is commonly used to value the cost of DALYs or YLLs, while the VSL is used to monetize the number of statistical deaths. But the issue of how the VSLY should be appropriately estimated is contentious. In practice, the VSLY is usually derived by dividing the VSL by the discounted number of years remaining in the average person’s expected lifespan (Cropper and Khanna 2014). In  the World Bank’s 2013 study of pollution costs in India, for example, the discounted remaining life expectancy for the average adult is assumed to be 7.5 years. Previously, in the World Bank’s calculation of air pollution damages as part of adjusted net savings, it was assumed to be 22 years. This method for deriving a constant VSLY implies that that individuals’ WTP for mortality risk reductions, and hence the VSL, should decline propor- ­ tionally with age. As previously discussed in the section on age and the VSL, however, this assumption is not borne out by empirical evidence, and there is little support for using a single VSLY to value life years for people of all age groups (Aldy and Viscusi 2007; Cropper and Khanna 2014; Hammitt 2013; Krupnick 2007; Morgan et al. 2007; NRC 2008; Ryen and Svensson 2015; U.S. EPA 2014; Viscusi 2014). Valuing health impacts in the form  of DALYs or YLL using a constant VSLY further implies that that the VSL should decline proportionally with the state of an individual’s health. However, in comparing the VSLY for smokers and nonsmokers, Viscusi and Hersch (2008) have found this not to be the case. Smokers have similar VSLY values as nonsmokers despite generally being in poorer health. Relatively few studies have attempted to determine the VSLY empirically (OECD 2012).13 Desaigues and coauthors (2007, 2011) have produced two recent examples. The authors con- ducted a contingent valuation study in nine European countries to estimate a policy-relevant VSLY for the European Union as part of an European Commission–funded effort. In the sur- vey, adult respondents were asked about hypothetical air pollution control measures that would lead to small gains in life expectancy of three to six months but would also increase the cost of living. As explained to respondents, this gain in life expectancy does not mean “addi- tional months of misery at the end of one’s life” (Desaigues et al. 2011, 905); rather it is more like decelerated aging. The authors argue that these marginal changes in life expectancy are easier to understand and provide better results than asking respondents about small reduc- tions in the probability of dying (see also Morris and Hammitt 2001). The results indicate an appropriate VSLY for cost-benefit analysis of air pollution policies in the EU-25 of around €40,000, with a confidence interval of €25,000 to €100,000 (PPP-adjusted, year 2005 prices).14 The authors find that respondents’ WTP for marginal gains in life expectancy did not vary significantly by age. The VSLY estimated by the Desaigues et al. study is below the range of values that has been used by the EC DG Environment elsewhere for cost-benefit analysis of environmental poli- cies. Analyses of the Clean Air for Europe Programme (CAFE) and the EU Clean Air Package have employed a base VSLY of €59,000 to €133,000 (PPP-adjusted, year 2005 prices) (Hurley et al. 2005; Holland 2014).15 The lower and upper values of this range correspond to the median and mean VSLY from a three-country study in France, Italy, and the United Kingdom by Alberini, Hunt, and Markandya (2004, 2006), adjusted for price inflation. This study employed a survey instrument similar to that originally developed by Alberini, Krupnick, and others in the United States and Canada (see Alberini et al. 2004 and Krupnick et al. 2002). 22 Methodology for Valuing the Health Impacts of Air Pollution In the survey, respondents were asked about their WTP for marginal reductions in fatality risks over the next 10 years. Alberini, Hunt, and Markandya then computed the VSLY by dividing respondents’ WTP by the gain in life expectancy implied by an annual risk reduction of 5 in 10,000. In other words, respondents were not asked directly about how they valued life expectancy. Gains in life expectancy were calculated using the life tables for the three coun- tries. To gauge whether there was a difference between the life tables and subjective percep- tions of remaining life expectancy, the authors also asked respondents to self-assess their remaining life expectancy. Using this subjective measure of life expectancy, the median and mean values for the VSLY are just slightly higher at €62,000 and €151,000 (PPP-adjusted, year 2005 prices).16 The authors also find the VSLY increases with age, peaking among 60- to 69-year-olds before declining thereafter. The median and mean VSL calculated from the sample for an annual risk reduction of 5 in 10,000 are €1.122 million and €2.414 million ­ (PPP-adjusted, year 2005 prices).17 Use of the VSLY to monetize YLLs introduces additional uncertainties into the analysis.18 One issue is whether individuals who are more susceptible to pollution-related illnesses have the same life expectancy as the general population. The presence of co-morbid conditions and multiple hazards may further complicate matters. An individual’s life expectancy could be influenced by any number of shocks to the individual’s survival curve, both now and into the future (Hammitt 2007; Jones-Lee et al 2015). Another issue with applying the VSLY to monetize YLLs is the implicit rate of discount with respect to remaining life expectancy. Moore and Viscusi (1990) estimate discount rates for deferred health losses and health benefits in the range of 1–14 percent using labor market data, and Dreyfus and Viscusi (1995) estimate a higher range of 11–17 percent by looking at prices paid for cars with higher fuel efficiency and better safety features. The rate of time preference may vary with individual characteristics, such as education level and state of health (Scharff and Viscusi 2011). Jones-Lee and coauthors (2015) point out the importance of ensuring con- sistency between the discount rate used to calculate health impacts (YLLs or DALYs) and the rate used to discount life expectancy in deriving the VSLY from the VSL. Finally, there is the question of when the undiscounted VSLY should be allowed to vary for YLLs suffered by people of different ages and at different periods of time. Whereas the study by Desaigues and coauthors (2011) did not find significant variation of WTP for extended life expectancy among adults of different ages, the study by Alberini, Hunt, and Markandya (2004, 2006) estimated an age-varying VSLY for adults over age 40. As noted, the literature on the relationship between WTP for mortality risk reductions is inconclusive, though there is some evidence that WTP may decline only late in life. This situation would also suggest that the adult VSLY should increase with age. To that effect, the U.S. EPA had previously used a two-tiered VSLY of US$163,000 and US$258,000 (year 1999 prices) for people ages 65 and below and for people over age 65, respectively, to evaluate the benefits of the Clear Skies Initiative (U.S. EPA 2002, 36). The U.S. OMB has also recommended that the VSLY be allowed to increase with age (U.S. OMB 2003), though it has not said how this should be done. The European Commission’s cost-benefit analyses of Europe’s air pollution control strategies assume an age-invariant VSLY and VSL (Hurley et al. 2005; Holland 2014). They also make no adjustment to the undiscounted base VSLY for YLLs lost in future periods. (See box 3.7.) Methodology for Valuing the Health Impacts of Air Pollution 23 Box 3.7  Question: Should the Value of Statistical Life or Value per Statistical Life Year Be Applied in Monetizing Premature Mortality Costs? How Should the Value Per Statistical Life Be Estimated? Proposal: The choice between using VSL versus VSLY depends in part on the metric of health impacts being used. The VSL is used for valuing the number of deaths and the VSLY is used for valuing YLLs or reduced life expectancy. As noted, this study will value both deaths and effects on life expectancy, so both the VSL and VSLY will be used. The empirical basis for estimating the VSLY from WTP for gains in life expectancy is less devel- oped, and there are fewer studies from which to derive a base value for the VSLY than there are for the VSL. Although it is common practice to derive the VSLY from the VSL by dividing the VSL by average remaining life expectancy, this practice is not really consistent with the empirical lit- erature on mortality risks and on the levels at which adults of different ages value risk reductions. Because no basis of evidence is robust enough to determine how WTP for gains in life expectancy varies with age, for the purposes of this study a constant VSLY will be used. This is common prac- tice in the literature and in policy analysis using the VSLY. Also, given the limited guidance from the literature, a range of VSLY values will be used (1) from the Desaigues et al. study, (2) from the Alberini et al. study, and (3) by deriving the VSLY indirectly from the VSL (see the next section for choice of VSL and box 3.8 for details). For country-level analysis, if the VSLY is used, choices about such issues as a relevant base value, discounting, and adjusting the VSLY for age should be made in consultation with local stakeholders. Base Value of Statistical Life Estimates for Policy Analysis Table 3.1 illustrates guideline values for the VSL that have been used by public agencies in different countries and regions for policy analysis. In the table, all amounts have been con- verted into PPP-adjusted U.S. dollars at year 2011 prices. The VSL estimates in the table are ­ onsumer first adjusted for inflation to 2011 prices in local currency units according to the c price index (CPI) in the respective country or region and then converted to U.S. dollars at purchasing power parity (PPP) using exchange rates for 2011 based on actual individual con- sumption (AIC) (Lindhjem et al. 2011; OECD 2012). Most of the agencies in the developing countries listed in table 3.1 have derived VSL estimates for policy analysis by transferring values from studies performed in other countries, mainly the United States. Table 3.2 shows estimates of the VSL from stated-preference studies in middle-income coun- tries, drawing primarily from the database assembled by the OECD (see Lindhjem et al. 2011). Table 3.3 shows highlighted VSL estimates from meta-analyses of stated-preference studies. The review by Kochi and coauthors (2006) only included studies from high-income countries, whereas the analyses by Dekker and coauthors (2011) and Lindhjem and coauthors (2011) include studies from a number of middle-income countries such as China, Malaysia, and Thailand. The WTP studies in the meta-analyses are for environmental, traffic-related, and general health risks. Table 3.4 provides a more detailed breakdown of descriptive statistics for 24 Methodology for Valuing the Health Impacts of Air Pollution Table 3.1  Examples of Value of Statistical Life Estimates Used by Public Agencies in Policy Analysis Methodology for Valuing the Health Impacts of Air Pollution VSL (Million U.S. VSL/GDP per Agency Application Source dollars, PPP) capita United States Office of Management and Budget (OMB) Guideline value 1.3–12.7 28–276 U.S. OMB (2003) Department of Transportation (DOT) Guideline value 8.9 (5.1–12.6) 176 (99–246) U.S. DOT (2015) Department of Homeland Security Report by US Customs and Border Protection 6.6 131 U.S. OMB (2015) Environmental Protection Agency NAAQS final rule review for PM standards 8.2 179 U.S. OMB (2015) Department of Labor OSHA hazard communication final rule 9.0 182 U.S. OMB (2015) Food and Drug Administration Examples of values used in rule-making 6.3, 7.0 135, 151 U.S. OMB (2015) United Kingdom Department for Transport (DFT) Accident and casualty costs table 2.3 62 U.K. DFT (2015) Department for Environment and Rural Affairs (Defra) Guidelines for valuing risk to life from flooding 2.0 62 U.K. Defra (2008) Canada Policy Horizons Canada Guideline value 5.5 (3.0–8.0) 133 (71–194) Chestnut and De Civita (2009) Australia Office of Best Practice Regulation Guideline value 2.6 60 Australia OBPR (2014) Mexico Ministry of Environment and Natural Resources Evaluation of fuel quality standards, from BT 0.9 19 Rojas-Bracho et al (2013) (SEMARNAT) Colombia Commission for Regulating Potable Water and Assessing benefits of improved water and 1.0 108 Colombia CRA (2008) Basic Sanitation sanitation, from BT Bogota, Local Secretary of Environment Value in city air pollution control strategy, from BT 0.9 88 Golub et al (2014) Peru Supervisory Organization for Energy Investment Value for assessing accidental deaths in energy 1.5 199 Vásquez Cordano (2006) (OSINERG) industry, from BT Malaysia Road Transport Department Guidelines for calculating losses in road accidents 1.0 60 Cited in Mohd Faudzi et al (2013) Note: VSL converted from base year and currency to 2011 U.S. dollars at PPP; the ratio of the VSL to GNI per capita in the year for which the VSL is estimated (in 2011 U.S. dollars, PPP); from BT is the value derived 25 from studies performed in other countries using benefit-transfer function. VSL = value of statistical life; PPP = purchasing power parity; NAAQS = National Ambient Air Quality Standards; PM = particulate matter. the VSL estimates contained in the OECD database and additional studies that have been pub- lished in recent years. All dollar amounts in Tables 3.2, 3.3, and 3.4 are expressed in PPP- adjusted U.S. dollars at 2011 prices. Only a few governments have set guidelines for baseline estimates of the VSLY (Table 3.5). As noted, the EC DG Environment uses a range of €59,000 to €133,000 (PPP-adjusted, year 2005 prices) (Hurley et al. 2005; Holland 2014). The United Kingdom recommends a VSLY of £35,000 for chronic mortality and £18,000 for acute mortality in evaluating the costs and ben- efits of air pollution regulations (2012 prices) (U.K. Defra 2007, 2013). The U.K. values are drawn from a domestic survey by Chilton and coauthors (2004),19 which asked respondents about their WTP for gains in life expectancy of one, three, and six months in “poor” versus “good” health. The Norwegian Ministry of Finance recommends a VSLY of NKr 425,000 (year 2005 prices) (cited in OECD 2012, 29). Best-practice guidelines by the Australian government Table 3.2  Examples of Value of Statistical Life Estimates from Stated Preference Studies in Developing Countries Year of Type of VSL (2011 U.S. Country VSL/Y Source survey risk dollars, PPP) Thailand 2003 Traffic 159,000 16 Vassanadumrongdee and Matsuoka (2005) India 2005 Traffic 74,000 23 Bhattacharya, Alberini, and Cropper (2006) Bangladesh 2003 Health 9,000 5 Mahmud (2009) Malaysia 2004 Traffic 1,794,000 103 ADB-ASEAN (2005) Malaysia 1999 Traffic 1,883,000 128 Nor Ghani and Mohd Faudzi (2003) Turkey 2012 Health 641,000 36 TekeSin and Ara (2014) Mongolia 2010 Health 638,000 83 Hoffman et al (2012) Iran 2013 Traffic 2,636,000 169 Ainy et al (2014) Sudan 2013 Traffic 243,000 63 Mofadal, Kanitpong, and Jiwattanakulpaisarn (2015) China 2006 Health 398,000 63 Krupnick, Hoffmann, and Qin (2010) Note: VSL is the mean VSL estimate for contemporaneous reductions in noncancer risks. VSL estimates in local currency units have been adjusted to year 2011 prices using the national consumer price index and converted to U.S. dollars at PPP rates; VSL/Y is ratio of the VSL to GDP per capita in the year of the survey. VSL = value of statistical life; PPP = purchasing power parity. Table 3.3  Examples of Value of Statistical Life Estimates from Meta-Analyses of Stated-Preference Studies Publication dates of underlying Highlighted VSL estimates Meta-analysis studies (number of studies) (Million 2011 US dollars, PPP) Kochi, Hubbell, and Kramer (2006) 1988–2002 (14) 3.5 Dekker et al. (2011) 1983–2008 (26) 2.8–9.1 Lindhjem et al. (2011); OECD (2012) 1982–2007 (26) 3.5 (OECD countries) 4.3 (EU27 countries) Source: Adapted from Cropper and Krupnick (2015). Note: VSL estimates have been adjusted to year 2011 prices using the national consumer price index and at PPP rates. VSL = value of statistical life; PPP = purchasing power parity; OECD = Organisation for Economic Co-operation and Development; EU27 = Austria, Belgium, Bulgaria, Cyprus, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Poland, Portugal, Romania, the Slovak Republic, Slovenia, Spain, Sweden, and the United Kingdom. 26 Methodology for Valuing the Health Impacts of Air Pollution Table 3.4  Average Values of Statistical Life from Database of Stated-Preference Studies VSL (2011 GDP per capita VSL / Sample Surveys Obs Years U.S. dollars, (2011 U.S. GDP per PPP) dollars, PPP) capita All countries Mean 40 449 1990–2013 1,767,397 19,510 91 Median 40 449 1990–2013 2,422,302 30,671 79 Mean, passes scope test 26 257 1994–2013 1,482,549 20,459 72 Median, passes scope test 26 257 1994–2013 2,082,871 32,248 65 Middle income countries Mean 15 201 1999–2013 382,440 7,007 55 Median 15 201 1999–2013 481,347 6,360 76 Mean, passes scope test 9 86 2003–2013 176,801 5,730 31 Median, passes scope test 9 86 2003–2013 302,092 6,360 47 OECD countries Mean 24 240 1990–2010 4,055,891 34,960 116 Median 24 240 1990–2010 3,624,531 35,548 102 Mean, passes scope test 16 167 1994–2010 3,832,843 36,839 104 Median, passes scope test 16 167 1994–2010 3,182,057 35,810 89 Sources: OECD, “Meta-Analysis of Value of Statistical Life Estimates” database, http://www.oecd.org/env/tools-evaluation/env-value​ -statistical-life.htm; additional observations from Ainy et al. 2014; Alolayan, Evans, and Hammitt 2013; Hoffmann et al. 2012; Krupnick, Hoffmann, and Qin 2010; Mofadal, Kanitpong, and Jiwattanakulpaisarn 2015; Tekes ¸in and Ara 2014; and Veisten et al. 2013. Note: Surveys indicates the number of willingness to pay studies in the sample. Obs are the total number of value of statistical life estimates from these studies. GDP per capita is for the year and country in which the survey was conducted, expressed in constant year 2011 prices. VSL/GDP per capita is the ratio of the average VSL to gross domestic product per capita. Only surveys satisfying the following quality-screening criteria have been included: (1) value for risk change reported, (2) main sample larger than 200 observations or subsample larger than 100 observations, and (3) sample representative of the general population. Passes scope test indicates that the subsample has been further restricted to studies that pass a test for internal or external consistency. Multiple observations are allowed per survey. Sample averages are weighted by the inverse number of observations from each survey. Underlying VSL estimates in national currency have been adjusted to year 2011 prices using the national consumer price index and at PPP rates. Mean VSL estimates are estimated from the log-transformed distribution. VSL = value of statistical life; PPP = purchasing power parity; OECD = Organisation for Economic Co-operation and Development; GDP = gross domestic product. Table 3.5  Examples of Guideline Value per Statistical Life Year Estimates Used by Public Agencies VSLY Agency (2011 U.S. dollars, PPP) United Kingdom Defra 24,000 (acute) 46,000 (chronic) NORWAY Ministry of Finance 49,000 AUSTRALIA OBPR 113,000 EC DG Environment 82,000–184,000 Sources: Australia OBPR (2014); Holland (2014); OECD (2012); and U.K. Defra (2007). Note: VSL estimates in local currency units have been adjusted to year 2011 prices using the national or regional (Euro area) consumer price index and converted to U.S. dollars at PPP rates. VSLY = value per statistical life year; PPP = purchasing power parity; Defra = Department for Environment and Rural Affairs; OBR = Office of Best Practice Regulation; EC DG Environment = European Commission Directorate-General for the Environment. Methodology for Valuing the Health Impacts of Air Pollution 27 Box 3.8  Question: What Base Value of Statistical Life and Value per Statistical Life Year Values Should Be Used? Proposal: The choice of base value of statistical life (VSL) depends on the risk and context being evaluated. Ideally, a global assessment of premature mortality costs in different countries would use empirically estimated VSLs from those countries. However, studies on willingness to pay for mortality risk reductions are still lacking in many countries, so this approach is not feasible. Instead, for the World Bank–IHME report, the mean VSL estimate from the quality-screened sample of studies in Organisation for Economic Co-operation and Development (OECD) coun- tries that pass some form of a scope test for internal or external consistency will be used as a base value: US$3.83 million (2011 U.S. dollars, purchasing power parity) (see table 3.4). For sensitivity analysis, a range of base VSL estimates from the distribution of stated-preference study estimates in different countries will be tested. For deriving the base value per statistical life year from the base VSL in the World Bank–IHME report, the VSL estimates in the database of stated-preference studies are divided by remaining life expectancy for the study population. Table 3.10 provides descriptive statistics for the subsam- ple of stated-preference studies from OECD countries for which the average age of respondents is reported. The mean VSL for the subsample is US$3.5 million (2011 U.S. dollars, purchasing power parity), which is slightly lower than the mean VSL for the larger sample. Associated gross domestic product per capita for the surveys and countries included in the subsample of OECD country studies is US$37,350. Respondents in this subsample are 50 years old on average, with a remaining life expectancy of 33 years. Discounting remaining life expectancy by an annual rate of 4 percent would imply that the VSLY is US$189,706. As noted, empirical studies provide little support for deriving the VSLY in this way, but it is nevertheless common practice and so is repeated in the World Bank–IHME report solely for the purposes of sensitivity analysis. highlight the usefulness of the VSLY in evaluating policies that are designed to prevent injuries and disease that are not immediately fatal and suggest adjusting the VSLY to monetize nonfa- tal conditions using disability weights (Australia OBPR 2014; Australia ASCC 2008). Australia’s guideline value for the VSLY for use in policy analysis is $A 182,000 (2014 prices). The U.S. OMB has also advised government agencies to present both VSL and VSLY-based estimates when valuing mortality risks (U.S. OMB 2003); however, no U.S. agency currently stipulates a guideline value for the VSLY. (See box 3.8.) Transferring Base Estimates Of The Value Of Statistical Life To Value Health Impacts In Other Countries VSL values are often transferred from the context in which they were originally estimated to a new context for policy analysis where it may be either impractical or cost-prohibitive to con- duct a WTP study. This approach is called benefit transfer. This section discusses two approaches to benefit transfer: unit value transfer and function-based transfer. Unit Value Transfer A common method for benefit transfer (BT) is unit value transfer, which involves taking an estimate from one study or an average value from several studies and applying it directly 28 Methodology for Valuing the Health Impacts of Air Pollution to the policy context. The OECD recommends this method “as the simplest and most trans- parent way of transfer between countries” (OECD 2012, 118). Unit value transfer assumes that individuals in the study context and policy setting have similar tastes and characteris- tics and that the type of risk being valued is the same, yielding a similar utility for mar- ginal  risk reductions (OECD 2012; U.S. EPA 2014). To capture population differences between the study and policy setting, the transferred VSL may be adjusted for salient char- acteristics such as per capita income. Use of the method thus requires discretion on the part of the analyst in screening high-quality studies for transfer and making appropri- ate adjustments for the policy case. Under these conditions, unit value transfer can be as accurate as more complex meta-regression analyses and other methods (Lindhjem and Navrud 2015). Recent analyses by the OECD have relied on unit value transfer in estimating country-specific VSLs to assess the public health costs of air pollution in Europe, China, and India (OECD 2014; WHO-OECD 2015). The authors begin by selecting a base VSL of about US$3 million (year 2005, PPP), the median VSL from a quality-screened sample of VSL estimates from studies in OECD countries analyzed by Lindhjem and coauthors (2011).20 They then make ­ the  following adjustment for cross-sectional differences in income between countries and post-2005 growth: b  Y  VSLc ,2010 = VSLOECD ,2005 ×  c ,2005  × (1 + %ΔP + %ΔY )b  YOECD ,2005  where VSLc,2010 is the VSL for country C in 2010; VSLOECD,2005 is the base value (VSL for OECD countries in 2005); Y is GDP per capita (in PPP terms); b is the income elasticity of the VSL, assumed by the authors to be 0.8, the midpoint of the best estimate of 0.7–0.9 (OECD 2012); %DP represents price inflation, as estimated by the consumer price index; and %DY is post- 2005 income growth. Several studies by the World Bank have also used VSL estimates derived from such a method, including analyses of Ecuador (Sander, Mira-Salama, and Feuerbacher 2015), Georgia (World Bank 2015b), and Nicaragua (Klytchnikova et al. 2013). In the Ecuador and Georgia examples, the base VSL estimated for OECD countries is transferred to the policy context assuming the same income elasticity of 0.8. The Nicaragua example uses a lower and upper base VSL from two other meta-analyses, which is transferred assuming an income elas- ticity of unity (1.0). Function-Based Transfer Another method commonly used for BT is function-based transfer, which proposes that the WTP may be described as a function of quantifiable characteristics in the study case and trans- ferred to the policy case using the estimated function. This method is demonstrated in a recent COED study of Kosovo by the World Bank (2013c), which applies a BT function derived from the meta-regression analysis of the OECD’s VSL database by Biausque (2012).21 The VSL is formulated as VSL = 1.044 · Y1.022 · r−0.445 Methodology for Valuing the Health Impacts of Air Pollution 29 where Y is per capita GDP (in PPP terms, as above), and r is the change in risk mortality. For the Kosovo example, an annual change in risk mortality of 1 in 10,000 is assumed. A similar analysis by the World Bank of the costs of air pollution in Argentina uses the same func- tional form to specify the VSL but presents a range of VSLs assuming different rates of annual change in mortality risks, from 1 in 10,000 to 5 in 10,000. Note that the BT function used by the World Bank in its assessments of Kosovo and Argentina assumes that the VSL is income-elastic, unlike studies done by the World Bank in other middle-income countries that have used a lower elasticity. This difference implies that the WTP for mortality risk reductions when compared as a percentage of income is lower for people in Kosovo and Argentina than it is for those in Georgia, for example, where an elasticity coefficient of 0.8 is assumed. The BT function proposed in the Biausque (2012) meta-analysis is illustrated graphically in figure 3.2. In their meta-analysis, Milligan and coauthors (2014) tailor a BT function specifically for low-income and middle-income countries. The function’s parameters are estimated using the OECD database of quality-screened VSL estimates from stated-preference studies,22 with some additional observations from middle-income and high-income countries (India, Republic of Korea, Mongolia, Sweden, and the United States). Including observations from all income levels (n = 410), they estimate an income elasticity of 0.6. Restricting the sample Figure 3.2  Benefit-Transfer Function from Biausque (2012) Used in World Bank Cost of Environmental Degradation Studies 4.5 VSL = 1.044 . Y 1.022 . r –0.445 4.0 3.5 3.0 VSL (million US$) 2.5 2.0 1.5 1.0 0.5 0 10,000 20,000 30,000 40,000 50,000 Y (GDP per capita, US$) r = 1/10,000 r = 5/10,000 r = 10/10,000 Source: World Bank. Note: in the transfer function, Y is GDP per capita, r is the annual reduction in mortality risk. VSL = value of statistical life; GDP = gross domestic product. 30 Methodology for Valuing the Health Impacts of Air Pollution to only lower-middle and middle income countries (n = 152), they calculate a much higher elasticity of 2.5. Although Milligan and coauthors (2014) specify the BT function in the context of evaluating road safety, the VSL estimates they use from the OECD database are also taken from environmental and health-related studies. The BT functions they estimate for road safety are: VSLT,HIC= 67.088 · Y 0.693 · r−0.481 VSLT,LMC = 2.226 · 10−7 · Y 2.478 · r −0.698 Where VSLT,HIC is the transferred VSL for traffic safety in high-income countries, VSLT,LMC is the transferred VSL for low-income and middle-income countries, Y is GDP per capita (at PPP exchange rates), and r is the change in risk mortality. Milligan and coauthors assume a  threshold of US$14,000 (PPP-adjusted, 2005 prices) for GDP per capita in high-income countries. In tailoring the BT function for road safety, they set r equal to one- half the crude death rate for traffic accidents in high-income countries and developing countries. Using data for 2011 from the WHO, they estimate the rate of traffic fatalities at 9.0 and 19.9 per 100,000 people in high-income countries and developing countries, respectively (cited in Milligan et al. 2014, 245). Inserting the assumed r values, the authors simplify the BT functions as VSLT,HIC = 8.2474 · 103 · Y 0.693 VSLT,LMC = 1.3732 · 10−4 · Y 2.478 Milligan and coauthors suggest that the BT function can be adjusted to value environmental and health-related risks in the same way (Milligan et al. 2014, 246). Re-estimating the BT functions for contemporaneous reductions in noncancer fatality risks related to general health using the VSL estimates from the OECD database and additional studies summarized in table 3.4 yields VSLHIC = 933.555 · Y 0.729 · r−0.088 VSLLMC = 0.058 · Y 1.792 · r −0.022 According to the GBD 2013 data, fatality rates in 2013 attributed to ambient PM2.5 exposure were 47 and 56 deaths per 100,000 people in high-income countries and developing countries, respectively. Assuming a baseline risk reduction equal to half the crude mortality rate, as done by Milligan and coauthors, simplifies the BT functions this way: VSLHIC = 1,947.645 · Y 0.729 VSLLMC = 0.069 · Y 1.792 Figure 3.3 illustrates the estimated VSLs transferred using the piece-wise BT functions for countries of different income levels, along with the BT function estimated by Biausque. For the Biausque function, annual risk changes of 1.0 in 10,000, 2.5 in 10,000, and 5.0 in 10,000 are assumed. Methodology for Valuing the Health Impacts of Air Pollution 31 Figure 3.3  Benefit-Transfer Functions and Benchmark Value of Statistical Life Estimates 12 11 10 VSL (million US$, PPP, year 2005) 9 8 7 6 5 4 3 2 1 0 10 20 30 40 50 GDP per capita (thousand US$, PPP, year 2005) Low- and middle-income High-income countries Biausque (2012), assuming countries 1.0-in-10,000 annual risk change VSL estimated as Biausque (2012), assuming Biausque (2012), assuming 70 x GDP per capita 2.5-in-10,000 annual risk change 5.0-in-10,000 annual risk change U.S. EPA upper-bound VSL U.S. EPA central (mean) VSL OECD central (mean) VSL OECD central (median) VSL U.S. EPA lower-bound VSL Source: World Bank. Note: VSL = value of statistical life; GDP = gross domestic product; PPP = purchasing power parity; EPA = Environmental Protection Agency; OECD = Organisation for Economic Co-operation and Development. Credibility Testing The fact that fewer empirical studies of the VSL have been done in low-income and middle-­ income countries raises questions about the credibility of the VSL when it is transferred between countries with large differences in average income and vastly disparate social ­contexts. Although there is little empirical basis for judging the reasonableness of the transferred VSL where in-country studies are lacking, several rule-of-thumb methods have been proposed. One method is to benchmark the VSL against GDP per capita. The International Road Assessment Programme (IRAP) and World Bank have recommended this method to estimate a rough VSL for the purpose of assessing traffic safety benefits. The logic behind comparing the VSL to GDP per capita stems from the relationship commonly assumed in BT functions such that differences in the VSL are proportional to income differences: VSLtransfer = VSLstudy * (Ytransfer /Ystudy) 32 Methodology for Valuing the Health Impacts of Air Pollution where transfer denotes the country for which the VSL is being transferred, study refers to the study context from which the original base value is estimated, and Y is GDP per capita. The function implies VSLstudy /Ystudy = VSLtransfer /Ytransfer. Cropper and Sahin (2009) suggest that a reasonable VSL/Y ratio for middle-income coun- tries is 80:1. IRAP and the World Bank employ a value of 70 times GDP per capita (Dahdah and McMahon 2008). Larsen (2015) suggests a lower ratio of 50:1 for low-income and middle-income countries. By comparison, Miller (2000) has reported a ratio of 120:1 from a meta-analysis of WTP studies in 13 high-income countries. The VSL/Y ratio reported by Miller is more or less in line with the average ratio for the stated-preference studies from OECD countries shown in table 3.4, which is between 102:1 and 116:1. The average ratio for VSL estimates from stated preference countries in middle-income countries in table 3.4 is between 55:1 and 76:1.23 The difference in means between the VSL/Y ratios of middle- and high-income countries is statistically significant,24 further suggesting that the income elasticity of the VSL is higher in the poorer countries than in the richer countries included in the database of studies. Figure 3.4 illustrates the spread of VSL/Y ratios for countries in different income groups for different assumed values of e using the chosen base VSL for the World Bank–IHME report: US$3.8 million (2011 US$, PPP). Average GDP per capita associated with this base value is about US$37,000 (2011 US$, PPP). Each observation in figure 3.4 represents a country and year for which the VSL is transferred. The distribution for e = 1 is not shown in the figure because when this elasticity is assumed, the ratio of VSL to Y is the same for all countries regardless of income (104:1). As figure 3.4 shows, values of e below 1 imply that the ratio of the VSL to Y is larger for poorer countries than for richer ones. The opposite is true for elasticities that are higher than 1. The ratio of the VSL to Y is lower for poorer countries, as seen with the VSL/Y ratios in the stated preference studies in table 3.4. The resulting VSL/Y ratios for differ- ent values of e are summarized in table 3.6. As an alternative to benchmarking the VSL against GDP per capita, Hammitt and Robinson (2011) have suggested comparing the VSL to expected lifetime income and consumption. The authors take this approach to test the assumption that the income elasticity is above 1 in trans- ferring VSL estimates from studies in high-income countries to low-income countries in Africa. Their approach is guided by the economic theory behind the VSL, which tells us that the amount individuals are willing to pay to reduce their risk of dying should reflect the utility they will experience over their remaining lifetimes. The level of utility, in turn, depends on such factors as age, life expectancy, survival probability, and their level of consumption.25 From a simple point of view, how much individuals can consume is typically limited by their earnings.26 This implies that the VSL should be at least the discounted present value of future income and consumption. In table 3.7, transferred VSL estimates using a range of values for e are compared to the value of future income. The present value of labor income is discounted over the remaining working life for the average adult between the ages of 15 and 69. Remaining working life takes into account survival probabilities and labor participation rates for adults of different ages. As with Methodology for Valuing the Health Impacts of Air Pollution 33 Figure 3.4  Ratio of Value of Statistical Life to Gross Domestic Product per Capita, Assuming Different Income Elasticities (e), All Years a. Income elasticity = 0.8 b. Income elasticity = 1.2 200 200 180 180 160 160 140 140 Observations Observations 120 120 100 100 80 80 60 60 40 40 20 20 0 0 0 20 40 60 80 0 0 0 0 0 0 0 0 0 10 12 14 16 18 20 22 24 20 40 60 80 0 0 0 0 0 0 0 0 10 12 14 16 18 20 22 24 Ratio of VSL to GDP per captia Ratio of VSL to GDP per captia c. Income elasticity = 1.4 200 180 160 140 Observations 120 100 80 60 40 20 0 0 20 40 60 80 0 0 0 0 0 0 0 0 10 12 14 16 18 20 22 24 Ratio of VSL to GDP per captia High-income countries Middle-income countries Low-income countries Source: World Bank. Note: Base VSL of US$3.8 million and GDP per capita of US$37,000 are assumed (2011 US$, PPP); VSL estimates are transferred at PPP rates. VSL = value of statistical life; GDP = gross domestic product. Table 3.6  Median Ratio of Transferred Value of Statistical Life Estimates to Gross Domestic Product per Capita, by Income Group, All Years (1990–2013) Median VSL to GDP per capita ratio, assuming different income elasticities Income group 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0 Low income 409 291 206 147 104 74 52 37 26 19 13 9 7 5 3 Lower middle 272 214 168 132 104 82 64 51 40 31 25 19 15 12 9 income Upper middle 178 156 136 119 104 91 80 70 61 53 46 41 36 31 27 income High income 109 108 106 105 104 103 102 100 99 98 97 96 95 94 93 Source: World Bank. Note: Base VSL of US$3.8 million and GDP per capita of US$37,000 are assumed (2011 US$, PPP); VSL estimates are transferred at PPP rates; table made of balanced sample of countries for which data are available for 1990, 1995, 2000, 2005, 2010, and 2013; each observation is a country and year. VSL = value of statistical life; GDP = gross domestic product. 34 Methodology for Valuing the Health Impacts of Air Pollution Table 3.7  Median Ratio of Estimated Value of Statistical Life to Expected Future Earnings for the Average Worker, All Years (1990–2013) Ratio of VSL to expected future labor income, assuming different Income income elasticities of VSL group 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0 Low income 17.9 12.4 8.8 6.3 4.5 3.2 2.2 1.6 1.1 0.8 0.6 0.4 0.3 0.2 0.2 Lower middle 12.4 10.0 8.1 6.5 5.2 4.1 3.2 2.5 1.9 1.5 1.2 1.0 0.8 0.6 0.5 income Upper middle 9.8 8.6 7.6 6.7 5.9 5.1 4.5 4.0 3.5 3.1 2.7 2.4 2.1 1.8 1.6 income High income 5.4 5.4 5.4 5.5 5.4 5.3 5.2 5.1 5.0 5.0 5.0 4.9 4.8 4.7 4.7 Source: World Bank. Note: Base VSL of US$3.8 million and GDP per capita of US$37,000 are assumed (2011 US$, PPP); Expected future labor income is estimated assuming a constant average annual wage per employee and the expected remaining working life for the average worker in the labor force (ages 15 to 69), discounted at a rate of 4 percent per year for high-income countries and 6 percent per year for low-income and middle-income countries; table made of balanced sample of countries for which data are available for 1990, 1995, 2000, 2005, 2010, and 2013; each observation is a country and year. VSL = value of statistical life. Table 3.8  Median Ratio of Estimated Value of Statistical Life to Expected Lifetime Consumption for Average Adult Ages 40–44, All Years (1990–2013) Ratio of VSL to remaining lifetime consumption, Income Life assuming different income elasticities of VSL group expectancy 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0 Low income 32 32.5 22.9 16.3 11.5 7.9 5.5 3.9 2.8 2.0 1.4 1.0 0.7 0.5 0.4 0.3 Lower middle 34 23.2 18.0 14.3 11.1 8.7 6.8 5.4 4.4 3.5 2.8 2.2 1.8 1.4 1.2 0.9 income Upper middle 37 15.4 13.5 11.8 10.3 9.0 8.0 7.1 6.3 5.5 4.8 4.2 3.7 3.3 2.9 2.6 income High income 40 7.7 7.7 7.5 7.3 7.1 7.1 7.0 6.9 6.9 7.0 6.8 6.7 6.6 6.6 6.5 Source: World Bank. Note: Base VSL of US$3.8 million and GDP per capita of US$37,000 are assumed (2011 US$, PPP); Remaining lifetime consumption is estimated assuming constant per capita consumption in real terms, discounted into the future at a rate of 4 percent per year for high-income countries and 6 percent per year for low-income and middle-income countries; table made of balanced sample of countries for which data are available for 1990, 1995, 2000, 2005, 2010, and 2013; each observation is a country and year. VSL = value of statistical life. consumption, real wages are held constant and future income is discounted at an annual rate of 4 percent for high-income countries and 6 percent for low-income and middle-income countries. Tables 3.8 and 3.9 illustrate how the estimates of VSLs for different countries and years ­ compare with the present value of future consumption for the average adult between the ages of 40 and 44 and 45 and 49. These age groups correspond with the median ages of survey respondents in ­ middle-income and high-income countries, respectively, in the OECD database of VSL studies (see table 3.10). Life expectancy for respondents in the year of survey is estimated using data from the GBD 2013. On average, survey respondents had a remaining life expectancy of 33–36 years. Methodology for Valuing the Health Impacts of Air Pollution 35 Table 3.9  Median Ratio of Estimated Value of Statistical Life to Expected Lifetime Consumption for Average Adult Ages 45–49, All Years (1990–2013) Ratio of VSL to remaining lifetime consumption, Income Life assuming different income elasticities of VSL group expectancy 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0 Low income 28 34.2 24.0 16.9 11.9 8.3 5.8 4.1 2.9 2.1 1.5 1.1 0.8 0.5 0.4 0.3 Lower middle 29 24.3 18.9 14.9 11.6 9.1 7.2 5.6 4.5 3.6 2.9 2.3 1.9 1.5 1.2 1.0 income Upper middle 33 15.9 14.0 12.2 10.6 9.3 8.3 7.3 6.5 5.7 4.9 4.4 3.9 3.4 3.0 2.7 income High income 35 8.2 8.0 7.9 7.7 7.5 7.4 7.4 7.2 7.3 7.3 7.1 7.1 6.9 6.9 6.8 Source: World Bank. Note: Base VSL of US$3.8 million and GDP per capita of US$37,000 are assumed (2011 US$, PPP); Remaining lifetime consumption is estimated assuming constant per capita consumption in real terms, discounted into the future at a rate of 4 percent per year for high-income countries and 6 percent per year for low-income and middle-income countries; table made of a balanced sample of countries for which data are available for 1990, 1995, 2000, 2005, 2010, and 2013; each observation is a country and year. VSL = value of statistical life. By comparison, life expectancy for the average adult ages 40–44 in the low-income countries was 32 years in 2013. Conservatively, in comparing the ratio of the VSL with future consumption in the tables, real consumption per capita is assumed to be constant and is discounted into the future at a rate of 4 percent for high-income countries and 6 percent for low-income and ­middle-income countries. For the World Bank–IHME report, in which a base VSL of US$3.8 million and GDP per capita of around US$37,000 (2011 US$, PPP) is assumed, the tables demonstrate that apply- ing income elasticity values of 1 or lower may lead to VSL/Y ratios in low-income and middle-income countries beyond those observed in stated preference studies to date. At the ­ same time, elasticity values of 1.4 or greater would result in VSL estimates that are lower than the expected future earnings for the average working adult in the majority of low-income countries and years. For elasticity values of 1.6 or greater, the present value of future consumption for adults between the ages of 40 and 49 would exceed the VSL in about ­ ­one-third of low-income countries in 2013. Elasticity values of 1.6 or greater would also cause expected future earnings to be greater than the estimated VSL for the majority of lower-­ middle income countries.27 In figure 3.5, transferred VSLs using the base VSL for the World Bank–IHME report and the range of elasticity values for middle-income countries are compared with the empirical esti- mates of the VSL in the database of stated preference studies (see table 3.4). As figure 3.5 shows, the empirically estimated VSL is within the range of transferred VSLs for half of the studies in the database. GDP per capita in these countries ranges from US$3,419 (China in 1999) to US$17,864 (Malaysia in 2004). However, the range of transferred VSLs is still above the empirical estimates for the two poorest countries, Bangladesh in 2003 and India in 2005, with GDP per capita of US$1,781 and US$3,213, respectively. The average transfer error is smallest when the central elasticity value of 1.2 is assumed. 36 Methodology for Valuing the Health Impacts of Air Pollution Methodology for Valuing the Health Impacts of Air Pollution Table 3.10  Value of Statistical Life Estimates and Remaining Life Expectancy from Stated-Preference Studies in Organisation for Economic Co-operation and Development Database for which Age Is Reported Per capita Lifetime GDP per Remaining VSL / VSL (2011 consumption consumption VSL / lifetime capita Sample Surveys Obs Earliest Latest Age life GDP per USD, PPP) (2011 USD, (2011 USD, consumption (2011 USD, expectancy capita PPP) PPP) PPP) Middle-income country studies Mean 11 154 1999 2012 45 34 249,143 6,430 94,674 2.6 5,882 42 Median 11 154 1999 2012 43 36 337,163 6,168 85,701 3.9 5,675 59 Mean, passes scope test 7 78 2003 2012 46 32 171,947 6,284 92,161 1.9 6,077 28 Median, passes scope test 7 78 2003 2012 44 35 296,942 5,176 64,159 4.6 6,360 47 OECD country studies Mean 19 198 1995 2010 51 33 3,470,963 27,196 495,252 7.0 35,841 97 Median 19 198 1995 2010 50 33 3,257,430 27,153 480,561 6.8 35,810 91 Mean, passes scope test 14 142 1995 2010 51 33 3,504,928 27,905 509,853 6.9 37,350 94 Median, passes scope test 14 142 1995 2010 50 33 3,182,057 27,197 501,965 6.3 36,962 86 Source: World Bank using data from Ainy et al. 2014; Alolayan, Evans, and Hammitt 2013; Hoffmann et al. 2012; Institute for Health Metrics and Evaluation Global Burden of Disease study 2013 database; Krupnick, Hoffmann, and Qin 2010; Mofadal, Kanitpong, and Jiwattanakulpaisarn 2015; Organisation for Economic Co-operation and Development, “Meta-Analysis of Value of Statistical Life Estimates” database; Tekes¸in and Ara 2014; and Veisten et al. 2013. 37 Figure 3.5  Transferred Values of Statistical Life Compared with Stated-Preference Study Results in Middle-Income Countries 3.0 2.5 VSL (Mn 2011 US$, PPP) 2.0 1.5 1.0 0.5 0 ] ] ] ] ] ] ] ] ] 0] 1] 2] 3] [1 [2 [3 [4 [5 [6 [7 [8 [9 [1 [1 [1 [1 2 3 05 6 3 0 9 13 5 01 00 00 00 01 99 00 04 3 99 13 20 20 00 ,2 ,2 ,2 ,2 ,2 ,1 ,2 20 19 20 ,2 a, n, ey nd na sh lia na na a, a, ., di da nd ep go rk de la hi hi hi si si In Su la ay ay Tu ai C C C R on la ai Th al al ng ic M Th M M m Ba la Is n, Ira Mean empirical estimate of VSL in study VSL transferred from OECD base value, income elasticity 1.4 VSL transferred from OECD base value, VSL transferred from OECD base value, income elasticity of 1.2 income elasticity of 1.0 Sources: [1] Tekes¸in and Ara 2014; [2] Vassanadumrondgee and Matsuoka 2005; [3] Bhattacharya, Alberini, and Cropper 2007; [4] Krupnick, Hoffmann, and Qin 2010; [5] Mahmud 2009; [6] Hoffmann et al. 2012; [7] Hammitt and Zhou 2006; [8] Mofadal, Kanitpong, and Jiwattanakulpaisarn 2015; [9] Krupnick et al. 2006; [10] ADB-ASEAN 2005; [11] Gibson et al. 2007; [12] Ghani and Faudzi 2003; [13] Ainy et al. 2014. Note: VSL = value of statistical life; OECD = Organisation for Economic Co-operation and Development. Choice of Exchange Rate Transferring the VSL by comparing GDP per capita in the study country with the policy ­country, the choice of exchange rate is particularly important. In general, VSL estimates in local currency units are converted into US$ at PPP rates rather than market exchange rates in making comparisons between countries because PPP rates are a better reflection of opportu- ­ nity costs in the economy (Krupnick, Hoffmann, and Qin 2010, 23). However, as Cropper and Khanna (2014) note in their review of the World Bank’s methodology for calculating air pollu- tion damages, if health costs are reported as a percentage of GNI—as they are in the World Bank’s adjusted net savings database—and GNI is measured at market exchange rates, then the estimated VSL should be transferred using market rates, too. Also, for economic analysis of pollution reduction projects and policies, benefits and costs should be estimated on the same basis using either the PPP or market exchange rate. Although the benefits of pollution reduc- tion policies in terms of avoided health impacts are typically monetized on a PPP basis, it is more common to report costs at market exchange rates, leading to potential inconsistencies when calculating net benefits over costs. (See box 3.9.) 38 Methodology for Valuing the Health Impacts of Air Pollution Box 3.9  Question: How Should the Base Value of Statistical Life Be Transferred between Countries? Proposal: Country-specific values of statistical life (VSLs) may be derived from a base VSL following two methods: (1) a simple unit-value transfer adjusting for cross-sectional dif- ferences in income, price inflation, and growth over time; and (2) a function-based trans- fer, with  parameters estimated from a meta-regression analysis using the Organisation for Economic Co-operation and Development database, as in figure 3.2. For the unit value trans- fer, and as previously discussed, a range of income elasticities should be used for the sensitivity analysis. For each country, the ratio of the estimated VSL to gross domestic product (GDP) or expected lifetime income should also be used as a benchmark to judge the reasonableness of the VSL. Unit-value transfer has the advantage of being the most straightforward transfer method, though it requires stricter assumptions about the comparability of individual tastes and characteristics in the study context versus the policy setting and about the nature of the risk being valued. It is the method that will be applied for the global assessment, assuming the range of income elasticity val- ues previously noted (0.6–1.0 for high-income countries, with a central value of 0.8, and 1.0–1.4 for low-income and middle-income countries, with a central value of 1.2). Function-based trans- fers, on the other hand, may be highly sensitive to how the function is specified and to what vari- ables are included. Re-estimating a base VSL or transfer function may be too time intensive and burdensome for individual project analysis. If functions from other analyses or contexts are used, they should be carefully examined and compared to rule-of-thumb measures such as the ratio of the estimated VSL to GDP to judge their reliability. For the purposes of valuing the global burden of disease associated with air pollution, transferring the base VSL at PPP rates is appropriate for cross-country comparisons. For cost-benefit analy- sis of specific projects or policies to reduce pollution, however, using market rates may be more appropriate for the calculation of net benefits. This is because costs are usually taken at market rates, so the monetary basis for comparing costs and benefits should be consistent. Implementing the Forgone Output Approach The following section discusses some of the practical issues in implementing a forgone labor output approach for the financial accounting of pollution health costs, including defin- ing the workforce, valuing child mortality, and accounting for future income growth. Under this approach, the expected loss of income for the average person in each age group is typically valued as T PV (I ) = ∑ I (1 + g )i /(1+ r )i i =0 where I is average per capita labor income in the present year; T is the expected number of working years for the average person in a particular age group (conditional on survival prob- abilities and labor force participation rates); g is the annual rate of income growth; and r is the social discount rate. Ideally, I would be differentiated by age groups, as people of different ages Methodology for Valuing the Health Impacts of Air Pollution 39 may have varying levels of education, experience, and other characteristics that affect their average earning potential. However, a lack of detailed wage data in many countries does not allow for this level of disaggregation as part of a global assessment. Defining the Workforce For the sake of convenience, in estimating forgone labor output, one typically assumes working age to be 15–64 for cross-country analyses.28 Defining the scope of the workforce in ­ this way excludes the contribution of persons aged 65 and over to the economy. According to data from the International Labor Organization (ILO), about 25 percent of the population ages 65 and over in middle-income countries were economically active in 2013. For low-​ income countries, the figure was even higher, around 56 percent.29 As noted, the 65-and-over population suffers a disproportionate share of health impacts from pollution. Out of con- cerns for equity, analysts may use an adjusted form of the FO approach that values YLLs among the 65-and-over population using the economywide average for GDP per capita, as is often done in countries such as China. However, this practice does not really capture the value of lost income or reduced labor productivity in the economy and would be contrary to the purposes of financial accounting for the ANS indicator, for which the FO approach is proposed. Also, although retired people contribute to the economy as consumers of goods and services, these sorts of losses beyond the individual’s time endowment are outside the scope of the FO approach. Working Life Expectancy Working life expectancy, T, may be calculated by adjusting life expectancy during working age by the probability that an individual will be active in the labor force: M Tj = ∑ s j ,t ⋅ t − 1 t = j +1 where sj,t is the probability that a person of age j will survive to the beginning of age t; is the labor force participation rate;30 and M is the end of working age. Survival probabilities, s, may be calculated from mean death rates as s j ,t = (1 − d j )(1 − d j +1 )...(1 − dt −1 ) sj,t = sj · sj+1 · … · st-1 where dj is the portion of individuals alive at the beginning of year j who die during that year, and the probability of survival is 1 – d. The calculation of T is done recursively, starting with the oldest age group. For example, if maximum working age is assumed to be 79, (such that all individuals stop working once they turn 80) T79 = s79,80 · ℓ 79 40 Methodology for Valuing the Health Impacts of Air Pollution 80 T78 = ∑ s78,t ⋅ t −1 = s78,79 ⋅ 78 + s78,80 ⋅ 79 = s78,79 ( 78 + T79 ) t = 79 T15 = s15,16 (ℓ15 + T16) Because working age is typically assumed to begin at 15, children represent a special case. Using ages 13 and 14 as an example, T would be 80 T14 = ∑ s14,t ⋅ t −1 = s14,15 ⋅ s15,16 ( 15 + T16 ) = (s14,15 ⋅T15 ) t = 16 80 T13 = ∑ s13,t ⋅ t −1 = s13,14 ⋅T14 t =16 Of course, not all individuals begin working as soon as they are 15, and not all individuals will work until they are 80. Weighting T by labor force participation is one way of accounting for variation in the demographic makeup of the labor force, including delayed entry until later age and retirement. Workforce participation rates vary widely by region, and there is longstanding evidence that participation by women is frequently underestimated (see Boserup 1970; Benería 1982, 1992; Benería, Berik, and Floro 2015). Low participation rates among women are the main reason that overall participation rates for both sexes in some regions are much lower than in others. These estimates of labor force participation by the ILO are based on reported rates in nation- ally representative labor force surveys and population censuses. Over the past decades, the statistical definition of employment used in these surveys has expanded to account for greater diversity in the way in which people work, and data collection has greatly improved. However, labor force statistics are still prone to measurement error, particularly in capturing work by people in the informal sector. In countries with high levels of informality in the labor market, this measurement bias particularly affects women, whose work may be intermittent, tempo- rary, short term, or home based and who may have multiple work arrangements, such as car- ing for children as well as engaging in market activities. Also, although international standards for employment and labor force statistics have evolved to be more inclusive, definitions and reporting by agencies in different countries vary. Hirway and Jose (2011) provide evidence of the potential magnitude of bias in the labor force participation statistics for India. According to data from India’s national labor force survey of 1999–2000, only 25 percent of rural women and 13 percent of urban women were active in the workforce in the states of Haryana, Madhya Pradesh, Gujarat, Orissa, Tamil Nadu, and Meghalaya. By comparison, 51 percent of men in rural and urban areas were estimated to be active. However, using data from a large-scale time-use survey from 1998 to 1999 in these states, Hirway and Jose take advantage of more detailed information about how people spend their time and to accurately distinguish between work and nonwork. With these data, they find that 58 percent of rural women and 31 percent of urban women were actually active in the Methodology for Valuing the Health Impacts of Air Pollution 41 Figure 3.6  Labor Force Participation, by Men and Women in South Asia, 2013 100 90 80 % population active in workforce 70 60 50 40 30 20 10 0 9 9 9 9 9 4 4 4 4 4 –1 –2 –3 –4 –5 –2 –3 –4 –5 –6 15 25 35 45 55 20 30 40 50 60 Age group Women Men Source: World Bank, using data from International Labour Organization, ILOSTAT Database, http://www.ilo.org/ilostat. workforce, compared with 63 percent of rural men and 59 percent of urban men. Thus, although there is still a gap between workforce participation by men and women, this gap may be much smaller than reflected in labor force survey and population census data. To illustrate the size of this gap in the ILO estimates of participation using labor force survey and popula- tion census data, average rates of participation for men and women in South Asia in 2013 are shown in figure 3.6. By lowering working life expectancy, gender bias in the labor force participation data would bias the overall estimates of forgone labor output due to air pollution exposure. This oversight would suggest that women are much less productive and would, in effect, assume zero loss of income for a large share of the working-age population that no doubt contributes to the econ- omy. On the flip side, failing to account for labor force participation rates as part of working life expectancy would lead to overestimates of forgone labor output by assuming that all indi- viduals will work from the moment they turn 15 until they are 79, as long as they are still alive. Faced with this dilemma, a reasonable compromise is to assume that estimated rates of labor force participation for men apply to both men and women. This option addresses concerns about gender bias in the labor force participation data while also accounting for varying work life patterns between countries. Indeed, there is precedence for this approach in the real world, and courts have sometimes assumed the average working life of men in calculating the lost earnings capacity of women. For example, in establishing a special fund for the victims of the 9/11 attacks, the U.S. Justice Department used the average working life of men to determine 42 Methodology for Valuing the Health Impacts of Air Pollution Figure 3.7  Labor Force Participation Rates, by Income Group, 2013 100 80 Labor force participation rate (% population) 60 40 20 0 + 9 9 9 9 9 4 4 4 4 4 65 –1 –2 –3 –4 –5 –2 –3 –4 –5 –6 15 25 35 45 55 20 30 40 50 60 Age group High income: non-OECD High income: OECD Upper middle income Lower middle income Low income Source: World Bank, using data from International Labour Organization, ILOSTAT database, http://www.ilo.org/ilostat/. Note: Data are for labor force participation among males. compensation amounts for both men and women “to avoid gender bias … and to ensure consistency” (Feinberg 2004). Adopting this approach for a global assessment of forgone ­ labor output would imply labor force participation rates for the general population such as those illustrated in figure 3.7. (See box 3.10.) Child Mortality In previous COED studies by the World Bank, the forgone output approach has been used to value of the cost of child mortality by taking the present value of wages beginning 10 or 15 years in the future, when those children would be expected to enter the workforce.31 As Larsen (2015) has argued, this approach assumes zero replacement of lost children, an approach which may be inconsistent with actual household decision making and family planning, par- ticularly in countries with high baseline rates of child mortality. Families in these settings may take into account higher levels of health risk when planning the desired number of children in their households, especially if death occurs at a very young age. If this is the case, the actual loss in terms of labor productivity would be less than predicted. Furthermore, because the majority of YLLs among children are suffered in countries with higher rates of solid fuel use Methodology for Valuing the Health Impacts of Air Pollution 43 Box 3.10  Question: How Should the Workforce and Working Life Be Defined? Proposal: Defining working age as 15–64 is arbitrary. Excluding the population ages 65 and over can be avoided by accounting for country-specific labor force participation rates among the elderly conditional on survival probability when specifying T. This approach is seen in the eco- nomic literature on human capital accounting. (See the United Nations University International Human Dimensions Programme on Environmental Change and United Nations Environment Programme estimates of human capital for the Inclusive Wealth Report 2014, for example.) However, although the International Labor Organization provides data on labor force participa- tion among the population ages 65 and over in 185 countries, there are no comparable data on average retirement age for the subset of the older population that is economically active. Although this makes it difficult to avoid imposing an arbitrary cap on maximum working age altogether, given the high rates of labor participation among older persons in low-income and medium-­ income countries, a compromise would be to raise the maximum age to 79 and to discount work- ing life expectancy for labor participation. On grounds of equity, an analysis could alternatively value years of life lost at gross domestic product per capita for people of all ages. Though simpler to implement, strictly speaking this alternative approach would represent a departure from an income-based measure of forgone output for financial accounting purposes. ­ Adjusting life expectancy by rates of labor force participation introduces possible bias in the esti- mates of forgone labor output for regions where workforce participation by women is underre- ported. The possibility of bias in the labor statistics is especially a problem in contexts in which women tend to work outside the formal labor market. For the global assessment in the World Bank–IHME report, a compromise approach is taken by assuming the same rates of labor force participation for both men and women. Box 3.11  Question: How Should Deaths among Children Be Valued under the Forgone Output Approach? Proposal: As done in previous cost of environmental degradation studies, for this study the pres- ent value of forgone income among children will be calculated by discounting average wages into the future, assuming that working age begins at 15. Again, although this definition of the start of working age is somewhat arbitrary, it is conventional practice. The term T may be conditioned on survival probability and average working-life expectancy for persons in each age group, to avoid overstating damages in countries with high baseline rates of child mortality (and lower life expectancies). This has the same effect as discounting expected lifetime income for children more heavily into the future. At present, there is insufficient evidence to make any adjustments for household family planning decisions and child replacement rates, which might partially offset expected losses in labor output. 44 Methodology for Valuing the Health Impacts of Air Pollution for cooking and a greater burden from household air pollution—and the average damage cost for the loss of a child under the forgone output approach is significantly higher than that for adults in their late 50s and 60s—this has the effect of amplifying damage estimates in low-­ income countries. (See box 3.11.) Accounting for Income Growth Currently, in the World Bank’s global estimates of air pollution damages for the ANS indicator, a constant annual income growth rate of 2.5 percent is assumed for all countries and years. Country-specific studies in fast-growing developing countries have used higher rates, for exam- ple, 6–8 percent in the World Bank–SEPA study of pollution costs in China (2007). For the United States, the U.S. EPA typically assumes an annual income growth rate of 1 percent. (box 3.12.) Discount Rate Related to the assumed rate of future income growth is the discount rate. In The Changing Wealth of Nations, the World Bank estimates human capital for China using a lifetime income approach and sets the discount rate of 8.26 percent (World Bank 2011). This rate is assumed on the basis of the Ramsey formula: C r = r +h C where the discount rate, r, is equal to the pure rate of time preference, ρ, plus the elasticity of utility with respect to consumption. For the World Bank’s human capital accounts in China, the pure rate of time preference is assumed to be 1.5 and utility is assumed to be iso-elastic with respect to consumption (h = 1). With average annual growth in real annual per capita consump- tion of 6.76 percent between 1970 and 2008, the social discount rate equals 8.26 percent (8.26 = 1.50 + 6.76). This is close to the discount rate of 8.5 percent assumed in the United Nations University International Human Dimensions Programme on Environmental Change and United Nations Environment Programme’s human capital accounts for the Inclusive Wealth Report (see UNU-IHDP and UNEP 2012). However, for slower-growing economies, setting r at 8.5 percent would be too high. For most assets included in the World Bank’s comprehensive wealth accounts and ANS calculation, a social discount rate of 4 percent is chosen. (See box 3.13.) Box 3.12  Question: What Assumptions Should Be Made about Future Income Growth? Proposal: Under a forgone output approach, assuming a constant income growth rate of 2.5 percent may understate damages in fast-growing economies such as China and India. It may also overstate damages in high-income countries with lower rates of growth. Yet, any assumption about long-term income growth, especially in valuing health impacts among young people with long working-life expectancies, involves significant uncertainty. For the global assessment, and to the extent possible, a constant rate of 3 percent will be assumed for low-income and middle-­ income countries in keep- ing with recent guidance by the World Bank for estimating long-term rates of growth in project economic analysis. A growth rate of 2 percent will be assumed for high-income countries. Methodology for Valuing the Health Impacts of Air Pollution 45 Box 3.13  Question: What Discount Rate Should Be Used? Proposal: For the global assessment, a constant social discount rate of 6 percent will be assumed for low-income and middle-income countries in keeping with recent guidance by the World Bank on setting social discount rates for project economic analysis. A rate of 4 percent will be assumed for high-income countries. Notes 1. Neoclassical economic theory suggests that people’s WTP and WTA for nonmarket goods should be equal, but this is often not the case. The contrast suggests that people’s WTP and WTA are influ- enced by behavioral factors and by differences in the properties of welfare measures that depend on how public goods are provided (see Carson 2012 and Kling et al. 2012). 2. This example borrows from the example in OECD (2012, 14). 3. This claim is based on reviews of wage-risk studies and stated-preference studies in North America. In their review of the more recent literature, Robinson and Hammitt (2015) find that “high-quality” stated-preference studies tend to yield VSL estimates close to those from wage-risk studies. 4. Similarly, in the literature on human capital accounting, the accumulation of human capital is often valued in terms of the expected gain in lifetime earnings potential due to higher levels of schooling among workers, following the approach first developed in the United States by Dale Jorgenson and Barbara Fraumeni (1992). 5. Recent examples include Croitoru et al. 2010 (Jordan); Klytchnikova et al. 2013 (Nicaragua); World Bank 2013a (India); World Bank 2013c (Kosovo); Meisner et al. 2015 (Macedonia); Sander et al. 2015 (Ecuador). 6. In contrast with the stated-preference studies, wage-risk studies have more consistently demon- strated an inverted U-shaped relationship between the VSL and age such that the VSL rises with age for younger workers, peaks for workers around the age of 50, and then declines (Aldy and Viscusi 2007, 2008; Kniesner, Viscusi, and Ziliak 2006; Viscusi 2014). However, as noted, these wage-risk studies deal mainly with the working-age population in decent health considering occupational injuries rather than illness-related health risks, so they do not reflect the preferences of individuals across the broad swath of age groups affected by pollution, particularly the elderly. 7. Recent examples include Golub et al. 2014 (Colombia); World Bank 2015a (Georgia); Croitoru et al. 2010 (Jordan); World Bank 2013a (India); World Bank 2013c (Kosovo); Meisner et al. 2015 (Macedonia). 8. For the meta-regression analysis, income is approximated by country-level GDP per capita constant purchasing power parities [PPPs]). (in ­ 9. In other words, people of higher incomes assign a larger value to marginal reductions in mortality risk than do people of lower incomes, and this difference in value exceeds their proportional dif- ferences in income. Note, however, that the income elasticity of WTP is different from the income elasticity of demand, and usually smaller, so the usual definition of a luxury good does not apply in the same way. 10. These quality standards are (1) the study reports a value for the risk change (for example, 5 in 10,000); (2) the survey sample includes a sample of more than 200 observations and subsamples of 100 observations or more; and (3) the sample is representative of the broader population. 11. The tests considered are internal and external scope tests. 12. The VSLY is also commonly referred to as the value of a life year, or VOLY. 13. Examples include Johannesson and Johansson (1996), Chilton et al. (2004), and Desaigues et al. (2007, 2011). 14. US$56,000, with an uncertainty interval of US$35,000 to US$141,000 (PPP-adjusted, year 2011 prices). The EU-25 are Austria, Belgium, Cyprus, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, the 46 Methodology for Valuing the Health Impacts of Air Pollution Netherlands, Poland, Portugal, Slovenia, the Slovak Republic, Spain, Sweden, and the United Kingdom. 15. US$82,000 to US$184,000 (PPP-adjusted, year 2011 prices) 16. US$85,000 to US$209,000 (PPP-adjusted, year 2011 prices) 17. US$1.545 million to US$3.316 million (PPP-adjusted, year 2011 prices) 18. Some of these uncertainties are elaborated by the U.S. EPA in its sensitivity analysis of the health benefits of the Clean Air Act using the VSLY (U.S. EPA 1997, appendix I). 19. The U.K. VSLY figures are PPP-adjusted and expressed in 2012 prices, as in the Defra guidance note. 20. The average associated GDP per capita for the OECD countries studied is about US$30,000 (year 2005, PPP). 21. See also Lindhjem and coauthors (2011), OECD (2012), and Lindhjem and Navrud (2015). 22. The quality-screened VSL estimates included in the sample meet the following conditions: (1) the study reports a value for the risk change (e.g., 5 in 10,000); (2) the estimates are drawn from a study sample of more than 200 observations or a sub-sample of 100 observations or more; (3) the study sample is representative of the broader population; (4) the estimates pass an internal and/or external scope test for validity. 23. Restricting the sample of studies to those that satisfy internal or external scope tests yields slightly lower VSL/Y ratios of 103:1 to 120:1 for high-income countries and 30:1 to 37:1 for middle-income countries. 24. The log-transformed VSL/Y ratios in the database of WTP studies follow a lognormal distribution. In comparing the distributions of logged VSL/Y ratios for middle-income and high-income coun- tries, the null hypotheses that the average VSL/Y ratio is equal or higher for middle-income coun- tries can be rejected at the 99 percent level. 25. The life-cycle model of consumption and savings that underpins the VSL is described by Yaari (1965); Shepard and Zeckhauer (1982, 1984); Cropper and Sussman (1990); Cropper and Sahin (2009); and others. 26. Wealth, savings, insurance, and other financial instruments or assets may bend this constraint. 27. Note that these conclusions are for the base VSL in the World Bank–IHME report and would not necessarily hold for other base VSL estimates and GDP per capita values. The results should not be generalized to other cases without doing the analysis. 28. This is the age limit assumed by the World Bank in estimating labor productivity losses from pollution as a component of adjusted net savings for the World Development Indicators database ­ (World Bank 2015). See Liu and Fraumeni (2014) for an example from the human capital account- ing literature. 29. That said, the age 65-and-over population represents only 4 and 3 percent of the workforce in mid- dle-income and low-income countries, respectively. Data are not available on the share of the age 65-and-over population in total wages. 30. This assumes that life expectancy increases monotonically with age, which is not the case for coun- tries with high rates of infant mortality where life expectancy at birth is lower than it is for children who survive to age 1. 31. See World Bank 2013a, for example. Methodology for Valuing the Health Impacts of Air Pollution 47 4. Valuing Morbidity Costs No standard method exists for monetizing the costs of nonfatal health outcomes (OECD 2014). Hunt and Ferguson (2010) propose that the economic cost of morbidity equals the sum of the following: •• Resource costs: direct financial costs for avoiding or protecting against exposure or for the care and treatment of pollution-associated illness; •• Opportunity costs: indirect costs from the loss of time for work and leisure; and •• Disutility costs: the cost of pain, suffering, discomfort, or inconvenience. The difficulty in aggregating morbidity costs owes to the complex nature of nonfatal out- comes, as noted by the WHO and OECD in their study of air pollution costs in Europe (WHO-OECD 2015). Morbidity involves a plurality of health endpoints (lost work days, hospital admissions, expenditures for long-term medical treatment, and so on) suffered by a plurality of agents (the patient as well as the patient’s friends, family, coworkers, and so on). This raises the possibility of double counting or inconsistency in cost estimation meth- ods for different outcomes. Data may be missing on health expenditures and medical treat- ment costs for a wide array of countries. Differences in health care systems between countries—and how health care costs are allocated to patients, care providers, and the pub- lic treasury—is also a problem. Finally, there are fewer studies of WTP to avoid different kinds of illnesses and other medical conditions associated with pollution exposure than there are of WTP to avoid mortality risk, so measuring disutility costs is more difficult. Expressed WTP to avoid pain and suffering from illness may also be entangled with other costs (such as expensive or lengthy medical treatment) that individuals associate with conditions such as chronic bronchitis. Hunt and coauthors (2016) have further proposed a core set of health endpoints that could potentially be used to value the disutility, resource, and opportunity costs associated with air pollution: •• Respiratory and cardiovascular hospital admissions related to PM or ozone exposure •• Restricted activity days and days of lost work due to PM and ozone exposure •• Chronic bronchitis in adults due to PM exposure •• Acute bronchitis in children ages 6 to 18 due to PM exposure •• Acute lower respiratory illness in children ages 5 and under due to PM exposure The common set of endpoints is the result of an extensive review of regulatory analysis in Europe, the United States, and non-OECD countries such as China, as well as the GBD. Despite this, Hunt and coauthors note that “the empirical evidence for these pollutant-health out- comes remains limited” (Hunt et al. 2016, 5). Furthermore, some outcomes such as lost work- days are “strongly socio-culturally determined” and “point to major uncertainties about transferability” (Hunt et al. 2016, 39). The question of the transferability is especially thorny. Because they are strongly influenced by social context and are often measured in different ways, unit cost values for different morbidity 48 Methodology for Valuing the Health Impacts of Air Pollution endpoints vary widely, even among countries of similar income levels. For example, in their review of the literature, Hunt and coauthors (2016) find that estimates of WTP to avoid chronic bronchitis range from around US$40,000 to US$90,000 in European countries to US$880,000 to US$2.03 million in the United States. Hospital treatment costs for respiratory and cardio- vascular conditions associated with air pollution range from US$1,500 to US$2,600 per case in European countries and to US$20,000 to US$40,000 per case in the United States. Though disutility costs represent the largest and arguably the most easily transferable component of morbidity costs, it is unclear whether they can be transferred from one country to another in the same way as the VSL. Confronting these problems of measurement, the OECD has taken the pragmatic stance that morbidity costs are at least greater than nothing and so can be estimated provisionally as a small percentage of mortality costs (see OECD 2012 and 2014 and Hunt et al. 2016). As a benchmark value, the OECD has proposed a 10 percent markup of mortality costs to account for morbidity on the basis of the relative share of avoided morbidity costs in the total health benefits of air pollution control as estimated in regulatory analyses by the U.S. EPA and EC DG Environment. Whether this benchmark value holds for countries outside the OECD is ques- tionable. As shown in table 3.6, previous studies by the World Bank in middle-income coun- tries have estimated a wide range of morbidity costs. Studies in Nicaragua and China, for instance, have found that morbidity costs may be as great as 44 percent of total health costs (World Ban–SEPA 2007; Golub et al. 2014). However, because the studies included in table 4.1 take differing approaches to estimating morbidity costs, these figures are not really compara- ble. (See box 4.1.) Table 4.1  Share of Morbidity Costs in Total Health Costs from Ambient Air Pollution, World Bank Studies Morbidity Year share Country Morbidity cost components Source studied of total costs (%) Argentina 2012 7–16 Opportunity costs: Days of illness x 50% average daily World Bank income (2015a) China 2003 18–44 Disutility costs: WTP to avoid chronic bronchitis (using World Bank– VSL and QALY weights) SEPA (2007) Resource costs: Hospital admission costs (direct and indirect) Colombia 2012 20 Disutility costs: WTP calculated as 2.27 x (resource + Golub et al. opportunity costs), as done by Bogota government (2014) Resource costs: Upper bound of publicly listed prices that public insurers pay health care providers Opportunity costs: For urban PM, time lost valued at 75% urban wage, applied to both working and nonworking people; for HAP, time lost valued at 75% rural wage; for chronic bronchitis, time loss valued over 20-year period table continues next page Methodology for Valuing the Health Impacts of Air Pollution 49 Table 4.1  Share of Morbidity Costs in Total Health Costs from Ambient Air Pollution, World Bank Studies (continued) Morbidity Year share Country Morbidity cost components Source studied of total costs (%) India 2009 7 Disutility costs: YLDs valued using VSLY (as part of World Bank DALYs) (2013a) Resource costs: Medical treatment costs from previous studies and consultations with medical providers, costs of chronic bronchitis and COPD taken over 20-year period Opportunity costs: 75% average wage x days of illness Jordan 2006 28 Disutility costs: VSL used to value YLDs Croituru, Resource costs: Medical treatment costs Cervigni, and Jabarin (2010) Kosovo 2010 10 Disutility costs: WTP is 2 x (resource + opportunity World Bank costs) (2013c) Resource costs: Total public medical costs (not just cost to patient) for hospitalization, doctor visits, and ER visits Opportunity costs: Time of illness valued at 50% average wage for patients and caregivers Nicaragua 2007 44 Disutility costs: YLDs valued using VSLY Klytchnikova Resource costs: Hospital admissions, ER visits, doctor et al. (2013) visits Opportunity costs: Time of illness valued at 75% urban wage for both wage earners and nonwage earners; for chronic bronchitis, costs are discounted over 20-year period Box 4.1  Question: How Should Morbidity Costs Be Valued? Proposal: Standard approaches to measuring the costs of morbidity across countries are still in an early stage of development, and consistency is still lacking. The most common approach is to value resource and opportunity costs using country-specific data on medical treatment costs and wages. Disutility costs of nonfatal health outcomes are also sometimes valued in terms of years lived with disability (YLDs) or disability-adjusted life years (DALYs) using a value per statistical life year (VSLY), though issues with the use of YLDs and DALYs for economic valuation and the derivation of the VSLY remain. For the World Bank–IHME study, the goal is to provide estimates of the overall economic costs of the health burden from air pollution in countries around the world. As has been shown, nonfatal health outcomes represent a relatively small fraction of the total health burden. Because there is still no standard approach for valuing morbidity costs and gathering country-specific data on resource costs would be prohibitive, it may be reasonable for the World Bank–IHME study to exclude morbidity costs. Mortality costs may be presented with the caveat that total health costs including morbidity may be at least another 10 percent, on the basis of expert judgment and pre- vious studies by the World Bank and OECD. 50 Methodology for Valuing the Health Impacts of Air Pollution References ADB-ASEAN (Asian Development Bank and Association of Southeast Asian Nations). 2005. “The Cost of Road Traffic Accidents in Malaysia.” Accident Costing Report AC 5: Malaysia, ADB-ASEAN Regional Road Safety Program, ADB and ASEAN, Manila, the Philippines. Ainy, Elaheh, Hamid Soori, Mojtaba Ganjali, Henry Le, Taban Baghfalaki. 2014. “Estimating Cost of Road Traffic Injuries in Iran Using Willingness to Pay (WTP) Method.” PLoS ONE 9 (12): e112721. Alberini, Anna, Maureen Cropper, Alan Krupnick, Nathalie Simon. 2004. “Does the Value of a Statistical Life Vary with Age and Health Status? Evidence from the US and Canada.” Journal of Environmental Economics and Management 48 (1): 769–92. Alberini, Anna, Anil Markandya, and Alistair Hunt. 2004. “Willingness to Pay to Reduce Mortality Risks: Evidence from a Three-Country Contingent Valuation Study.” FEEM Working Paper 111.04, Fondazione Eni Enrico Mattei Note di Lavoro series, September. ———. 2006. “Willingness to Pay to Reduce Mortality Risks: Evidence from a Three-Country Contingent Valuation Study.” Environmental and Resource Economics 33: 251–64. Alberini, Anna, and Milan Ščasný. 2013. “Exploring Heterogeneity in the Value of a Statistical Life: Cause of Death v. Risk Perceptions.” Ecological Economics 94: 143–55. Aldy, Joseph and W. Kip Viscusi. 2007. “Age Differences in the Value of Statistical Life: Revealed Preference Evidence.” Review of Environmental Economics and Policy 1 (2): 241–60. Alolayan, Mohammad, John Evans, James Hammitt. 2013. “Valuing Mortality Risk in Kuwait: Stated- Preference with a New Consistency Test.” Harvard University working paper, http://isites.­ harvard​ .edu/fs/docs/icb.topic1446005.files/February%2011%20-%20James%20Hammitt/Kuwait%20 VSL%20December%202013.pdf. Australia ASCC (Australian Safety and Compensation Council). 2008. “The Health of Nations: The Value of a Statistical Life,” http://www.safeworkaustralia.gov.au/sites/swa/about/publications/pages​ /rr200807valueofastatisticallife. Australia OBPR (Office of Best Practice Regulation, Australia). 2014. “Best Practice Regulation Guidance Note: Value of Statistical Life.” Report by OBPR, Department of the Prime Minister and Cabinet, Government of Australia, December. Avnery, Shiri, Denise Mauzerall, Junfeng Liu, and Larry Horowitz. 2011. “Global Crop Yield Reductions Due to Surface Ozone Exposure: 1. Year 2000 Crop Production Losses and Economic Damage.” Atmospheric Environment 45 (13): 2284–96. Benería, Lourdes. 1982. “Accounting for Women’s Work.” In Women and Development: The Sexual Division of Labor in Rural Societies, edited by Lourdes Benería. New York: Praeger. ———. 1992. “Accounting for Women’s Work: The Progress of Two Decades.” World Development 20 (11): 1547–60. Benería, Lourdes, Günseli Berik, and Maria Floro. 2015. Gender, Development, and Globalization: Economics as if All People Mattered, 2nd ed. Oxon, UK: Routledge. Bhattacharya, Soma, Anna Alberini, and Maureen Cropper. 2006. “Value of Mortality Risk Reductions in Delhi, India.” Policy Research Working Paper 3995, World Bank, Washington, DC. ———. 2007. “The Value of Mortality Risk Reductions in Delhi, India.” Journal of Risk and Uncertainty 34 (1): 21–47. Biausque, Vincent. 2012. “The Value of Statistical Life: A Meta-Analysis.” Report ENV/EPOC/ WPNEP(2010)9/FINAL, Working Party on National Environmental Policies, Organisation for Economic Co-operation and Development, Paris. Blomquist, Glenn, Mark Dickie, and Richard O’Conor. 2011. “Willingness to Pay for Improving Fatality Risks and Asthma Symptoms: Values for Children and Adults of All Ages.” Resource and Energy Economics 33 (2): 410–25. Boserup, Ester. 1970. Women’s Role in Economic Development. London: George Allen and Unwin. Methodology for Valuing the Health Impacts of Air Pollution 51 Carson, Robert. 2012. “Contingent Valuation: A Practical Alternative When Prices Aren’t Available.” Journal of Economic Perspectives 26 (4): 27–42. Chen, Stephen. 2014. “Agriculture Feels the Choke as China Smog Starts to Foster Disastrous Conditions.” South China Morning Post, February 25, http://www.scmp.com/news/china/­article/1434700​/­china​ -smog-threatens-agriculture-nuclear-fallout-conditions-warn. Chestnut, Lauraine, and Paul De Civita. 2009. “Economic Valuation of Mortality Risk Reduction: Review and Recommendations for Policy and Regulatory Analysis.” Research Paper, Policy Research Initiative, Government of Canada, March. Chile MDS (Ministerio de Desarrollo Social, Chile). 2014. “Estimación del valor de los costos sociales por fallecimiento prematuro en chile (Estimation of the Social Costs of Premature Mortality in Chile).” Report by the Division of Social Evaluation of Investments, Sub-Secretary of Social Evaluation, Ministry of Social Development, Government of Chile, March. Chilton, S., J. Covey, M. Jones-Lee, G. Loomes, H. Metcalf. 2004. “Valuation of Health Benefits Associated with Reductions in Air Pollution.” Report for Department for Environment and Rural Affairs (DEFRA), Government of United Kingdom, May. Colombia CRA (Comisión de Regulación de Agua Potable y Saneamiento Básico, Colombia). 2009. “Regulación de agua potable y saneamiento básico, revista no. 15 (Regulation of Potable Water and Basic Sanitation, Review No. 15).” http://www.cra.gov.co/es/novedades​ boletines-y-publicaciones​ /­ /15491-boletines-publicaciones. Costa, Dora, and Matthew Kahn. 2004. “Changes in the Value of Life, 1940–1980.” Journal of Risk and Uncertainty 29 (2): 159–80. Croitoru, Lelia, Raffaello Cervigni, and Amer Jabarin. 2010. “Air Pollution: The Case of Jordan” in The Cost of Environmental Degradation: Case Studies from the Middle East and North Africa, edited by Lelia Croitoru and Maria Sarraf, 37–52. Washington, DC: World Bank. Cropper, Maureen. 2000. “Has Economic Research Answered the Needs of Environmental Policy?” Journal of Environmental Economics and Management 39 (3): 328–50. Cropper, Maureen, James Hammitt, and Lisa Robinson. 2011. “Valuing Mortality Risk Reductions: Progress and Challenges.” Annual Review of Resource Economics 3 (1): 313–36. Cropper, Maureen, and Shefali Khanna. 2014. “How Should the World Bank Estimate Air Pollution Damages?” Discussion Paper RFF DP 14–30, Resources for the Future, Washington, DC, September. Cropper, Maureen, and Alan Krupnick. 2015. “Estimating the Economic Costs of Pollution.” Presentation at Global Alliance on Health and Pollution meeting, New York, NY, November 2015. Cropper, Maureen, and Sebnem Sahin. 2009. “Valuing Mortality and Morbidity in the Context of Disaster Risks.” Policy Research Working Paper 4832, Policy Research Working Paper 4832, World Bank, Washington, DC. Cropper, M.L., and F.G. Sussman. 1990. “Valuing Future Risks to Life.” Journal of Environmental Economics and Management 19: 160–174. Dekker, Thijs, Roy Brouwer, Marjan Hofkes, and Klaus Moeltner. 2011. “The Effect of Risk Context on the Value of a Statistical Life: A Bayesian Meta-model.” Environ Resource Econ 49: 597–624. Desaigues, Brigitte, Dominique Ami, Marie Hutchison, Ari Rabl, Sue Chilton, Hugh Metcalf, Alistair Hunt, Ramon Ortiz, Ståle Navrud, Peter Kaderjak, Richárd Szántó, Jytte Seested Nielsen, Claude Jeanrenaud, Sonia Pellegrini, Markéta Braun Kohlová, Milan Scasny, Vojtěch Máca, Jan Urban, Marie-Eve Stoeckel, Ania Bartczak, Olimpia Markiewicz, Pere Riera, and Veronica Farreras. 2007. “Final Report on the Monetary Valuation of Mortality and Morbidity Risks from Air Pollution.” New Energy Externalities Developments for Sustainability (NEEDS) project, 6th Framework Programme, Paris. Desaigues, Brigitte, Dominique Ami, Ania Bartczak, Markéta Braun Kohlová, Sue Chilton, Mikołaj Czajkowski, Veronica Farreras, Alistair Hunt, Marie Hutchinson, Claude Jeanrenaud, Peter Kaderjack, Vojtěch Máca, Olimpia Markiewicz, Agnieszka Markowska, Hugh Metcalf, Ståle Navrud, Jytte Seested Nielsen, Ramon Ortiz, Sonia Pellegrini, Ari Rabl, Pere Riera, Milan Scasny, 52 Methodology for Valuing the Health Impacts of Air Pollution Marie-Eve Stoeckl, Richárd Szánto, Jan Urban. 2011. “Economic Valuation of Air Pollution Mortality: A 9-Country Contingent Valuation Survey of a Value of a Life Year (VOLY).” Ecological Indicators 11 (3): 902–10. Dreyfus, Mark, and W. Kip Viscusi. 1995. “Rates of Time Preference and Consumer Valuations of Automobile Safety and Fuel Efficiency.” Journal of Law and Economics 38 (1): 79–105. Eeckhoudt, Louis, and James Hammitt. 2001. “Background Risks and the Value of a Statistical Life.” Journal of Risk and Uncertainty 23 (3): 261–79. EIU (Economist Intelligence Unit). 2014. “A Summary of the Liveability Ranking and Overview,” avail- able at http://pages.eiu.com/rs/eiu2/images/Liveability_rankings_2014.pdf. Feinberg, Kenneth R. 2004. “Final Report of the Special Master for the September 11th Victim Compensation Fund of 2001.” U.S. Department of Justice, Washington, D.C. https://pca-cpa.org​ /­wp-content/uploads/sites/175/2016/01/September-11-Victim-Compensation-Fund-Final-Report​ -Volume-I-1.pdf. Fraumeni, Barbara and Gang Liu. 2014. “Human Capital: Country Estimates Using Alternative Approaches” in Inclusive Wealth Report 2014: Measuring Progress Toward Sustainability, pro- duced by United Nations University International Human Dimensions Programme on Global Environmental Change and United Nations Environment Programme, 109–22. Cambridge, UK: Cambridge University Press. GBD 2013 Collaborators (Global Burden of Disease Study 2013 Collaborators). 2015a. “Global, Regional, and National Age-Sex Specific All-Cause and Cause-Specific Mortality for 240 Causes of Death, 1990–2013: A Systematic Analysis for the Global Burden of Disease Study 2013.” The Lancet 385 (9963): 117–71. ———. 2015b. “Global, Regional, and National Incidence, Prevalence, and Years Lived with Disability for 301 Acute and Chronic Diseases and Injuries in 188 Countries, 1990–2013: A Systematic Analysis for the Global Burden of Disease Study 2013.” The Lancet 386 (9995): 743–800. ———. 2015c. “Global, Regional, and National Comparative Risk Assessment of 79 Behavioural, Environmental and Occupational, and Metabolic Risks or Clusters of Risks in 188 Countries, 1990–2013: A Systematic Analysis for the Global Burden of Disease Study 2013.” The Lancet ­ 396 (10010): 2287–2323. Gibson, John, Sandra Barns, Michael Cameron, Steven Lim, Frank Scrimgeour, and John Tressler. 2007. “The Value of Statistical Life and the Economics of Landmine Clearance in Developing Countries.” World Development 35 (3): 512–31. Golub, Elena, Irina Klytchnikova, Gerardo Sanchez-Martinez, Juan Carlos Belausteguigoitia, and Centro Mario Molina. 2014. “Environmental Health Costs in Colombia: The Changes from 2002 to 2010.” Latin America and Caribbean Region, Environment and Water Resources Occasional Paper Series 92956, World Bank, Washington, DC. Guan, Dabo, and Zhu Liu. 2013. Tracing Back the Smog: Source Analysis and Control Strategies for PM2.5 Pollution in Beijing-Tianjin-Hebei. Beijing: China Environment Press. Guo, Xiaoqi, Timothy Haab, and James Hammitt. 2006. “Contingent Valuation and the Economic Value of Air-Pollution–Related Health Risks in China.” Paper presented at 2006 annual meeting of American Agricultural Economics Association, July 23–26, Long Beach, CA. Haagsma, Juanita, Suzanne Polinder, Allessandro Cassini, Edoardo Colzani, Arie Havelaar. 2014. “Review of Disability Weight Studies: Comparison of Methodological Choices and Values.” Population Health Metrics 12 (20). Hammitt, James. 2002. “QALYs Versus WTP.” Risk Analysis 22 (5): 985–1001. ———. 2007. “Valuing Changes in Mortality Risk: Lives Saved Versus Life Years Saved.” Review of Environmental Economics and Policy 1 (2): 228–40. ———. 2013. “Admissible Utility Functions for Health, Longevity, and Wealth: Integrating Monetary and Life-Year Measures.” Journal of Risk and Uncertainty 47: 311–25. Methodology for Valuing the Health Impacts of Air Pollution 53 Hammitt, James, and Kevin Haninger. 2011. “Valuing Morbidity Risk: Willingness to Pay per Quality- Adjusted Life Year.” LERNA Working Paper 11.09.343, University of Toulouse, France. https://ideas​ .repec.org/p/ler/wpaper/24322.html. Hammitt, James, Jin-Tan Liu, and Jin-Long Liu. 2000. “Survival Is a Luxury Good: The Increasing Value of a Statistical Life.” Discussion paper for the NBER Summer Institute Workshop on Public Policy and the Environment, National Bureau of Economic Research, Cambridge, MA. Hammitt, James, and Lisa Robinson. 2011. “The Income Elasticity of the Value per Statistical Life: Transferring Estimates Between High and Low Income Populations.” Journal of Benefit-Cost Analysis 2 (1): 1–27. Hammitt, James, and Ying Zhou. 2006. “The Economic Value of Air-Pollution–Related Health Risks  in China—A Contingent Valuation Study.” Environmental and Resource Economics 33: 399–423. Hirway, Indira, and Sunny Jose. 2011. “Understanding Women’s Work Using Time-Use Statistics: The Case of India.” Feminist Economics 17 (4): 67–92. Hausman, Jerry. 2012. “Contingent Valuation: From Dubious to Hopeless.” Journal of Economic Perspectives 26 (4): 43–56. Hoffmann, Sandra, Ping Qin, Alan Krupnick, Burmaajav Badrakh, Suvd Batbaatar, Enkhjargal Altangerel, Lodoysamba Sereeter. 2012. “The Willingness to Pay for Mortality Risk Reductions in Mongolia.” Resource and Energy Economics 34 (4): 493–513. Holland, Mike. 2014. “Cost-Benefit Analysis of Final Policy Scenarios for the EU Clean Air Package, Version 2, Corresponding to IIASA TSAP Report #11, Version 2a.” Report for European Commission Directorate-General for the Environment, October. Holland, Mike, Alistair Hunt, Fintan Hurley, Ståle Navrud, and and Paul Watkiss. 2005. “Methodology for the Cost-Benefit Analysis for CAFE, Vol. 1: Overview of Methodology.” Report to the European Commission Directorate-General for the Environment, Didcot, UK. Hubbell, Bryan. 2002. “Implementing QALYs in the Analysis of Air Pollution Regulations.” Draft QALY Paper for Environmental and Resource Economics, Office of Air Quality and Planning Standards, US Environmental Protection Agency, May. Hunt, Alistair, and Julia Ferguson. 2010. “A Review of Recent Policy-Relevant Findings from the Environmental Health Literature.” Report ENV/EPOC/WPNEP(2009)9/FINAL, Working Party on National Environmental Policies, Organisation for Economic Co-operation and Development, Paris. Hunt, Alistair, Julia Ferguson, Fintan Hurley, and Alison Searl. 2016. “Social Costs of Morbidity Impacts of Air Pollution.” OECD Environment Working Paper 99, Organisation for Economic Co-operation and Development, Paris. http://dx.doi.org/10.1787/5jm55j7cq0lv-en. Hurley, Fintan, Alistair Hunt, Hilary Cowie, Mike Holland, Brian Miller, Stephen Pye, and Paul Watkiss. 2005. “Methodology for the Cost-Benefit Analysis for CAFE, Vol. 2: Health Impact Assessment.” Report to the European Commission Directorate-General for the Environment, Didcot, UK. Johannesson, Magnus, and Per-Olov Johansson. 1996. “To Be, or Not to Be, That Is the Question: An Empirical Study of the WTP for an Increased Life Expectancy at an Advanced Age.”Journal of Risk and Uncertainty 13: 163–74. Jones-Lee, Michael, Susan Chilton, Hugh Metcalf, and Jytte Seested Nielsen. 2015. “Valuing Gains in Life Expectancy: Clarifying Some Ambiguities.” Journal of Risk and Uncertainty 51 (1): 1–21. Jorgenson, Dale, and Barbara Fraumeni. 1992. “The Output of the Education Sector” in Output Measurement in the Service Sectors, edited by Zvi Griliches, 303–341. Chicago: University of Chicago Press. Kaplow, Louis. 2005. “The Value of a Statistical Life and a Coefficient of Relative Risk Aversion.” Journal of Risk and Uncertainty 31 (1): 23–34. 54 Methodology for Valuing the Health Impacts of Air Pollution Kling, Catherine, Daniel Phaneuf, and Jinhua Zhao. “From Exxon to BP: Has Some Number Become Better Than No Number?” Journal of Economic Perspectives 26 (4): 3–26. Klytchnikova, Irina, Juan Carlos Belausteguigoitia, Elena Strukova, Anjali Acharya, Francisco Carranza, Nelson Antonio Medina Rocha, Gerardo Sánchez. 2013. “Environmental Health in Nicaragua: Addressing Key Environmental Challenges.” LAC Environment and Water Resources Occassional Paper Series 76894, World Bank, Washington, DC. Kniesner, Thomas J., and W. Kip Viscusi. 2005. “Value of a Statistical Life: Relative Position vs. Relative Age.” American Economic Review 95 (2): 142–46. Kniesner, Thomas, W. Kip Viscusi, James Ziliak. 2006. “Life-Cycle Consumption and the Age-Adjusted Value of Life.” Contributions to Economic Analysis and Policy 5 (1): 1–36. ———. 2010. “Policy Relevant Heterogeneity in the Value of Statistical Life: New Evidence from Panel Data Quantile Regressions.” Journal of Risk and Uncertainty 40 (1): 15–31. Kochi, Ikuho, Bryan Hubbell, and Randall Kramer. 2006. “An Empirical Bayes Approach to Combining and Comparing Estimates of the Value of a Statistical Life for Environmental Policy Analysis.” Environmental and Resource Economics 34: 385–406. Krupnick, Alan. 2007. “Mortality-Risk Valuation and Age: Stated Preference Evidence.” Review of Environmental Economics and Policy 1 (2): 261–82. Krupnick, Alan, Anna Alberini, Maureen Cropper, Nathalie Simon, Bernie O’Brien, Ron Goeree, and Martin Heintzelman. 2002. “Age, Health, and the Willingness to Pay for Mortality Risk Reductions: A Contingent Valuation Survey of Ontario Residents.” Journal of Risk and Uncertainty 24 (2): 161–86. Krupnick, Alan, Susan Hoffmann, Bjorn Larsen, Xizhe Peng, R. Tao, and C. Yan. 2006. “The Willingness to Pay for Mortality Risk Reductions in Shanghai and Chongqing, China.” RFF working paper, Resources for the Future, Washington, DC. Krupnick, Alan, Sandra Hoffmann, and Ping Qin. 2010. “Willingness to Pay for Mortality Risk Reduction in China.” RFF working paper. Resources for the Future, Washington, D.C. Larsen, Bjorn. 2004. “Colombia: Cost of Environmental Damage, A Socio-Economic and Environmental Health Risk Assessment.” Report for Ministry of Environment, Housing and Land Development, Government of Colombia. ———. 2015. “Updating Methodologies for Estimating Health Impacts and the Economic Costs of Air Pollution.” Draft working paper, e-mail communication with authors, April. Lindhjem, Henrick, and Ståle Navrud. 2015. “Reliability of Meta-Analytic Benefit Transfers of International Value of Statistical Life Estimates: Tests and Illustrations” in Benefit Transfer of Environmental and Resource Values: A Guide for Researchers and Practitioners, edited by Robert Johnston, John Rolfe, Randall Rosenberger, and Roy Brouwer, 441–64. Dordrecht, the Netherlands: Springer. Lindhjem, Henrick, Ståle Navrud, Nils Axel Braathen, and Vincent Biausque. 2011. “Valuing Mortality Risk Reductions from Environment, Transport and Health Policies: A Global Meta-Analysis of Stated Preference Studies.” Risk Analysis 31 (9): 1381–1407. Mahmud, Minhaj. 2009. “On the Contingent Valuation of Mortality Risk Reduction in Developing Countries.” Applied Economics 41 (2): 171–81. Matus, Kira, Trent Yang, Sergey Paltsev, John Reilly, Kyung-Min Nam. 2008. “Toward Integrated Assessment of Environmental Change: Air Pollution Health Effects in the USA.” Climatic Change 88: 59–92. Matus, Kira, Kyung-Min Nam, Noelle Selin, Lok Lamsal, John Reilly, and Sergey Paltsev. 2012. “Health Damages from Air Pollution in China.” Global Environmental Change 22 (1): 55–66. McMahon, Kate, and Said Dahdah. 2008. “The True Cost of Road Crashes: Valuing Life and the Cost of a Serious Injury.” Report for International Road Assessment Programme, March, http://www​ .irap.net/about-irap-3/research-and-technical-papers?download=45:the-true-cost-of-road-crashes​ -valuing-life-and-the-cost-of-a-serious-injury-espaol. Meisner, Craig, Dragan Gjorgjev, Fimka Tozija. 2015. “Estimating Health Impacts and Economic Costs of Air Pollution in the Republic of Macedonia.” SEEJPH 2015, April, DOI 10.12908/SEEJPH-2014-45. Methodology for Valuing the Health Impacts of Air Pollution 55 Miller, Ted. 2000. “Variations between Countries in Values of Statistical Life.” Journal of Transport Economics and Policy 34 (2): 169–88. Milligan, Craig, Andreas Kopp, Said Dahdah, and Jeannette Montufar. 2014. “Value of a Statistical Life in Road Safety: A Benefit-Transfer Function with Risk-Analysis Guidance Based on Developing Country Data.” Accident Analysis and Prevention 71: 236–47. Mofadal, Adam, Kunnawee Kanitpong, Piyapong Jiwattanakulpaisarn. 2015. “Analysis of Pedestrian Accident Costs in Sudan Using the Willingness-to-Pay Method.” Accident Analysis and Prevention 78: 201–11. Mohd Faudzi Mohd Yusoff, Nuura Addina Mohamad, Nahdiya Zainal Abidin, Nor Ghani Md. Nor, Husaini Salleh. 2013. “The Value of Statistical Life in Fatal Injury Among Drivers and Riders in Malaysia: Conjoint Analysis Method.” Malaysian Institute of Road Safety Research (MIROS) Internal Report, Kajang, Selangor Darul Ehsan, Malaysia. Moore, Michael and W. Kip Viscusi. 1990. “Models for Estimating Discount Rates for Long-Term Health Risks Using Labor Market Data.” Journal of Risk and Uncertainty 3: 381–401. Morgan, Granger, Maureen Cropper, and US EPA Science Advisory Board. 2007. “SAB Advisory on EPA’s Issues in Valuing Mortality Risk Reductions.” EPA-SAB-08-001, Memorandum to Stephen John, administrator, US Environmental Protection Agency, October 12. Morris, Jill, and James Hammitt. 2001. “Using Life Expectancy to Communicate Benefits of Health Care Programs in Contingent Valuation Studies.” Medical Decision Making 21 (6): 468–78. Murray, Christopher, Theo Vos, Rafael Lazano, et al. 2012. “Disability-Adjusted Life Years (DALYs) for 291 Diseases and Injuries in 21 Regions, 1990–2010: A Systematic Analysis for the Global Burden of Disease Study 2010.” The Lancet 380 (9859): 2197–223. Nam, Kyung-Min, Noelle Selin, John Reilly, Sergey Paltsev. “Measuring Welfare Loss Caused by Air Pollution in Europe: A CGE Analysis.” MIT Joint Program on the Science and Policy of Global Change Report 178, Massachusetts Institute of Technology, Cambridge, MA, August 2009. Nor Ghani Md. Nor, and Mohd Faudzi Mohd Yusoff. 2003. “Value of Life of Malaysian Motorists: Estimates from a Nationwide Survey.” Journal of the Eastern Asia Society for Transportation Studies 5 (October): 3031–37. NRC (National Research Council). 2008. Estimating Mortality Risk Reduction and Economic Benefits from Controlling Ozone Pollution. Washington, DC: National Academies Press. OECD (Organisation for Economic Co-operation and Development). 2012. Mortality Risk Valuation in Environment, Health, and Transport Policies. Paris: OECD. ———. 2014. The Cost of Air Pollution: Health Impacts of Road Transport. Paris: OECD. Pak, Jennifer. 2013. “Pollution Threatens Hong Kong’s Competitiveness,” BBC News, Hong Kong, January 16, http://www.bbc.com/news/business-21038495. Pliskin, Joseph, Donald Shepard, and Milton Weinstein. 1980. “Utility Functions for Life Years and Health Status.” Operations Research 28 (1): 206–24. Robinson, Lisa. 2007. “How US Government Agencies Value Mortality Risk Reductions.” Review of Environmental Economics and Policy (1): 283–299. Robinson, Lisa and James Hammitt. 2009. “The Value of Reducing Air Pollution Risks in Sub-Saharan Africa.” Report for Sub-Saharan Africa Refinery Study, World Bank, Washington, DC. ———. 2014. “Research Synthesis and the Value per Statistical Life.” Working Paper RPP-2014-14, Regulatory Policy Program, Mossavar-Rahmani Center for Business and Government, Harvard Kennedy School, Cambridge, MA. ———. 2015. “Valuing Reductions in Fatal Illness Risks: Implications of Recent Research.” Health Economics (June). doi:10.1002/hec.3214. Rojas-Bracho, Leonora, Verónica Garibay-Bravo, Gretchen Stevens, Georgina Echániz-Pellicer. 2013. “Environmental Fuel Quality Improvements in Mexico: A Case Study of the Role of Cost- Benefit Analysis in the Decision-Making Process” in The Globalization of Cost-Benefit Analysis 56 Methodology for Valuing the Health Impacts of Air Pollution in Environmental Quality, edited by Michael Livermore and Richard Revesz, 161–77. New York: Oxford University Press. Ryen, Linda and Mikael Svensson. 2015. “Willingness to Pay for a Quality Adjusted Life Year: A Review of the Empirical Literature.” Health Economics 24 (10): 1289–301. Salomon, Joshua, Theo Vos, Daniel Hogan, et al. 2012. “Common Values in Assessing Health Outcomes from Disease and Injury: Disability Weights Measurement Study for the Global Burden of Disease Study 2010.” The Lancet 380 (9859): 2129–43. Sander, Klas, Daniel Mira-Salama, Arndt Feuerbacher. 2015. The Cost of Air Pollution: A Case Study for the City of Cuenca, Ecuador. Washington, DC: World Bank Group. Scharff, Robert, and W. Kip Viscusi. 2011. “Heterogeneous Rates of Time Preference and the Decision to Smoke.” Economic Inquiry 49 (4): 959–72. Shepard, D.S., and R.J. Zeckhauser. 1984. “Survival versus Consumption.” Management Science 30: 423–39. Tan, Jijun, and Jinhua Zhao. 2014. “The Value of Clean Air in China: Evidence from Beijing and Shanghai.” Frontiers of Economics in China 9 (1): 109–37. Tan-Torres Edejer, Tessa, R. Baltussen, T. Adam, R. Hutubessy, A. Acharya, D.B. Evans, C.J.L. Murray. 2003. Making Choices in Health: WHO Guide to Cost-Effectiveness Analysis. Geneva: World Health Organization. Tekeşin, Cem, and Shihomi Ara. 2014. “Measuring the Value of Mortality Risk Reductions in Turkey.” International Journal of Environmental Research and Public Health 11 (7): 6890–922. Tsevat, Joel, Neal Dawson, Albert Wu, Joanne Lynn, Jane Soukup, E. Francis Cook, Humberto Vidaillet, and Russell Phillips. 1998. “Health Values of Hospitalized Patients 80 Years or Older.” Journal of the American Medical Association 279 (5): 371–75. U.K. Defra (Department for Environment, Food and Rural Affairs, United Kingdom). 2007. An Economic Analysis to Inform the Air Quality Strategy, Vol. 3: Updated Third Report of the Interdepartmental Group on Costs and Benefits, London: U.K. Defra. https://www.gov.uk/government/uploads/system​ /uploads/attachment_data/file/221088/pb12637-icgb.pdf. ———. 2008. “Defra Flood and Coastal Defence Appraisal Guidance, Social Appraisal, Supplementary Note to Operating Authorities: Assessing and Valuing the Risk to Life from Flooding for Use in Appraisal of Risk Management Measures.” https://www.gov.uk/government/uploads/system​ /­uploads/attachment_data/file/181441/risktopeople.pdf. ———. 2013. “Impact Pathway Guidance for Valuing Changes in Air Quality.” https://www.gov.uk​ /­government/publications/air-quality-impact-pathway-guidance. U.K. DFT (Department for Transport, United Kingdom). 2012. “Reported Road Casualties in Great -casualties​ Britain: Annual Report 2011.” https://www.gov.uk/government/statistics/reported-road​ -great-britain-annual-report-2011. ———. 2015. Accident and Casualties Costs (RAS60) tables, updated September 24, 2015. https://www​ .gov.uk/government/statistical-data-sets/ras60-average-value-of-preventing-road-accidents. U.K. Treasury (HM Treasury, United Kingdom). 2011. The Green Book: Appraisal and Evaluation in Central Government. Treasury Guidance, TSO, London. UN-HABITAT (United Nations Human Settlements Programme). 2012. State of the World’s Cities Report 2012/2013: Prosperity of Cities. Nairobi, Kenya: UN-HABITAT. UNU-IHDP and UNEP. 2012. Inclusive Wealth Report 2012: Measuring Progress Toward Sustainability, produced by United Nations University International Human Dimensions Programme on Global Environmental Change and United Nations Enivornment Programme. Cambridge, U.K.: Cambridge University Press. U.S. DOT (Department of Transportation, United States). 2015.”Guidance on Treatment of the Economic Value of a Statistical Life (VSL) in U.S. Department of Transportation Analyses—2015 Adjustment.” Memorandum, Office of the Secretary of Transportation, June 17. https://www.transportation.gov​ /­regulations/economic-values-used-in-analysis. Methodology for Valuing the Health Impacts of Air Pollution 57 U.S. EPA (Environmental Protection Agency, United States). 2003. Children’s Health Valuation Handbook. Office of Children’s Health Protection, National Center for Environmental Economics, U.S. EPA, October. ———. 2014. “Guidelines for Preparing Economic Analyses.” Report issued December 2010, updated May 2014, National Center for Environmental Economics, Office of Policy, U.S. EPA. U.S. OMB (Office of Management and Budget, United States). 2003. “Regulatory Impact Analysis: A Primer.” Circular A-4. September 17. ———. 2014. “2014 Report to Congress on the Benefits and Costs of Federal Regulations and Unfunded Mandates on State, Local, and Tribal Entities,” available at https://www.whitehouse.gov/sites/default​ /files/omb/inforeg/2014_cb/2014-cost-benefit-report.pdf. ———. 2015. “Draft Report to Congress on the Benefits and Costs of Federal Regulations and Agency Compliance with the Unfunded Mandates Reform Act,” available at https://www.whitehouse.gov​ /­sites/default/files/omb/inforeg/2015_cb/draft_2015_cost_benefit_report.pdf. Vásquez Cordano, Arturo. 2006. “El Valor de la Vida Estadística y sus aplicaciones a la Fiscalización de la Industria de Hidrocarburos (The Value of Statistical Life and its Applications in Regulating the Hydrocarbon Industry).” Working Paper 18, Office of Economic Studies, Supervisory Organization for Energy Investment (OSINERG), Government of Peru, Lima, March. Vassanadumrongdee, Sujitra, and Shunji Matsuoka. 2005. “Risk Perceptions and Value of a Statistical Life for Air Pollution and Traffic Accidents: Evidence from Bangkok, Thailand.” Journal of Risk and Uncertainty 30 (3): 261–87. Veisten, Knut, Stefan Flügel, Luis Rizzi, Juan de Dios Ortúzar, and Rune Elvik. 2013. “Valuing Casualty Risk Reductions from Estimated Baseline Risk.” Research in Transportation Economics 43 (1): 50–61. Viscusi, W. Kip. 1993. “The Value of Risks to Life and Health.” Journal of Economic Literature 31: 1912–46. ———. 2004. “The Value of Life: Estimates with Risks by Occupation and Industry.” Economic Inquiry 42: 29–48. ———. 2010. “The Heterogeneity of the Value of Statistical Life: Introduction and Overview.” Journal of Risk and Uncertainty (40): 1–13. ———. 2014. “The Value of Individual and Societal Risks to Life and Health.” In Handbook of the Economics of Risk and Uncertainty, Vol. 1, edited by Mark Machina and W. Kip Viscusi, 385–451. Amsterdam: Elsevier. Viscusi, W. Kip, and Joni Hersch. 2008. “The Mortality Cost to Smokers.” Journal of Health Economics 27: 943–958. Viscusi, W. Kip, Joel Huber, and Jason Bell. 2014. “Assessing Whether There Is a Cancer Premium for the Value of a Statistical Life.” Health Economics 23 (4): 384–96. Voigt, Kristin, and Nicholas King. 2014. “Disability Weights in the Global Burden of Disease 2010 Study: Two Steps Forward, One Step Back?” Bulletin of the World Health Organization 92: 226–28. Wang, Xiaoping, and Denise Mauzerall. 2004. “Characterizing Distributions of Surface Ozone and Its Impact on Grain Production in China, Japan, and South Korea: 1990 and 2020.” Atmospheric Environment 38 (26): 4383–402. Whitehead, Sarah, and Shehzad Ali. 2010. “Health Outcomes in Economic Evaluation: The QALY and Utilities.” British Medical Bulletin 96: 5–21. WHO (World Health Organization). 2008. WHO Guide for Standardization of Economic Evaluations of Immunization Programmes. Geneva: WHO. WHO-OECD (World Health Organization and Organisation for Economic Co-operation and Development). 2015. “Economic Cost of the Health Impact of Air Pollution in Europe: Clean Air, Health and Wealth.” Copenhagen: WHO Regional Office for Europe. World Bank. 2011. The Changing Wealth of Nations: Measuring Sustainable Development for the New Millennium. Washington, DC: World Bank. 58 Methodology for Valuing the Health Impacts of Air Pollution ———. 2013a. India: Diagnostic Assessment of Select Environmental Challenges, vol. 1., An Analysis of Physical and Monetary Losses of Environmental Health and Natural Resources. Washington, DC: World Bank. ———. 2013b. India: Diagnostic Assessment of Select Environmental Challenges, vol. 2., Economic Growth and Environmental Sustainability: What are the Tradeoffs? Washington, DC: World Bank. ———. 2013c. Kosovo—Country Environmental Analysis: Cost Assessment of Environmental Degradation, Institutional Review, and Public Environmental Expenditure Review. Washington, DC: World Bank. ———. 2015a. An Economic Assessment of Environmental Degradation in Argentina. Draft report. ———. 2015b. Georgia—Country Environmental Analysis: Institutional, Economic, and Poverty Aspects of Georgia’s Road to Environmental Sustainability. Washington, DC: World Bank. ———. 2015b. World Development Indicators 2015. Washington, DC: World Bank. ———. 2016. Leveraging Urbanization in South Asia: Managing Spatial Prosperity and Livability. Washington, DC: World Bank. World Bank–SEPA (World Bank and State Environmental Protection Agency, China). 2007. Cost of Pollution in China: Economic Estimates of Physical Damages. Washington, DC: World Bank. Yaari, M. 1965. “Uncertain Lifetime, Life Insurance, and the Theory of the Consumer.” Review Economic Studies 32: 137–50. Zheng, Siqi, Jing Cao, Matthew Kahn, and Cong Sun. 2014. “Real Estate Valuation and Cross-Boundary Air Pollution Externalities: Evidence from Chinese Cities.” Journal of Real Estate Finance and Economics (48): 398–414. Zheng, Siqi, Cong Sun, and Matthew Kahn. 2015. “Self-Protection Investment Exacerbates Air Pollution Exposure Inequality in China.” NBER Working Paper 21301, National Bureau of Economic Research, Cambridge, MA. http://www.nber.org/papers/w21301.pdf. Methodology for Valuing the Health Impacts of Air Pollution 59 SKU K8849