Environmental Degradation: Estimating the Health Effects of Ambient PM2.5 Air Pollution in Developing Countries
Environmental Degradation: Estimating the Health Effects of Ambient PM2.5 Air Pollution in Developing Countries
- Ernesto Sánchez-Triana, Ernesto Sánchez-TrianaThe World Bank
- Bjorn Larsen, Bjorn LarsenThe World Bank
- Santiago EnriquezSantiago EnriquezThe World Bank
- and Andreia Costa SantosAndreia Costa SantosLondon School of Hygiene and Tropical Medicine and The World Bank
Air pollution of fine particulates (PM2.5) is a leading cause of mortality worldwide. It is estimated that ambient PM2.5 air pollution results in between 4.1 million and 8.9 million premature deaths annually. According to the World Bank, the health effects of ambient PM2.5 air pollution had a cost of $6.4 trillion in purchasing power parity (PPP) adjusted dollars in 2019, equivalent to 4.8% of global gross domestic product (PPP adjusted) that year.
Estimating the health effects and cost of ambient PM2.5 air pollution involves three steps: (1) estimating population exposure to pollution; (2) estimating the health effects of such exposure; and (3) assigning a monetary value to the illnesses and premature deaths caused by ambient air pollution.
Estimating population exposure to ambient PM2,5 has gone from predominantly using ground level monitoring data mainly in larger cities to estimates of nationwide population weighted exposures based on satellite imagery and chemical transport models along with ground level monitoring data. The Global Burden of Disease 2010 (GBD 2010) provided for the first time national, regional and global estimates of exposures to ambient PM2.5. The GBD exposure estimates have also evolved substantially from 2010 to 2019, especially national estimates in South Asia, the Middle East and North Africa, Sub-Saharan Africa and Latin America and the Caribbean.
Estimation of health effects of ambient PM2.5 has also undergone substantial developments during the last two decades. These developments involve: i) going from largely estimating health effects associated with variations in daily exposures to estimating health effects of annual exposure; ii) going from estimating all-cause mortality or mortality from broad disease categories (i.e., cardiopulmonary diseases) to estimating mortality from specific diseases; and iii) being able to estimate health effects over a wide range of exposure that reflect ambient and household air pollution exposure levels in low- and middle-income countries.
As to monetary valuation of health effects of ambient air pollution, estimates in most low- and middle-income countries still rely on benefit transfer of values of statistical life (VSL) from high-income countries.
- Environmental Economics
There are differing estimates of the number of people worldwide who die as a result of air pollution. The World Health Organization (WHO) estimates that seven million people die each year from exposure to fine particles, including 4.2 million from ambient (outdoor) PM2.5 air pollution (WHO, 2018). Using the Global Exposure Mortality Model, Burnett et al. (2018) estimated that ambient PM2.5 air pollution caused 8.9 million deaths worldwide in 2015.
The 2019 Global Burden of Disease (GBD) Study estimated that 6.4 million people died from PM2.5 air pollution in 2019, making it the fourth leading cause of mortality worldwide. Ambient (outdoor) air pollution was the cause of 4.1 million deaths and household air pollution from the use of solid fuels for cooking accounted for the remaining 2.3 million deaths (GBD 2019 Risk Factors Collaborators, 2020). PM2.5 is particulate matter with a diameter of less than 2.5 micrometer (also called fine particulate matter), a diameter that is about 30 times smaller than that of a single human hair. Ozone, a harmful gas produced via atmospheric reactions from other gases, such as nitrogen oxides, carbon monoxide, and atmospheric methane, caused an additional 365,000 premature deaths (GBD 2019 Risk Factors Collaborators, 2020).
These estimates of deaths from PM2.5 differ because they use different methods and data. For instance, the estimates differ in the number of air pollution-related causes of death considered, in the number of deaths attributed to a particular air pollution exposure level, and on whether they were based on air quality data from local monitoring networks or satellite data (Alvarado et al., 2019; Ostro et al., 2018).
Fine particulate matter (PM2.5) is the air pollutant that affects human health the most because it is more toxic and is breathed more deeply into the lungs than coarser particles, such as PM10, a particulate matter that has a diameter of less than 10 micrometers (Pope & Dockery, 2006). The WHO has also noted that both ambient (outdoor) and household air pollution are the biggest environmental risks for noncommunicable diseases (NCDs), which constitute the largest cause of death and disease worldwide. Air pollution is responsible for almost 80% of NCDs caused by environmental risks. In addition, growing evidence indicates that early life exposure to environmental risks, including air pollutants, might increase NCD risk throughout an individual’s course of life (WHO, 2017).
Deaths caused by ambient air pollution—related mostly to PM2.5 but also to ozone, PM10 and other pollutants—are likely to increase in the coming decades. The Organisation for Economic Co-operation and Development (OECD) estimates that the global number of premature deaths caused by ambient air pollution could increase to between six and nine million by 2060 (OECD, 2016). Despite its widespread and significant health effects, ambient air pollution has not been a priority for policymakers in many countries, as evidenced by the 8% increase in pollution levels observed among cities that monitored air pollution for the period 2008–2013 (WHO, 2016). The GBD 2019 compared the evolution of 87 health risks and found that, while health risks from sources such as inadequate water, sanitation and hygiene, and household air pollution have fallen over time, health risks from ambient air pollution have grown. In fact, ambient air pollution was the health risk that had the second highest annualized rate of increase between 2010 and 2019 (GBD 2019 Risk Factors Collaborators, 2020).
Quantifying the health effects of ambient air pollution and assigning them a monetary value can help to communicate the severity of this environmental health risk in a language that policymakers can easily understand. Expressing these impacts in health or monetary units enables a comparison of the severity of different categories of environmental degradation and can inform efforts to set environmental priorities. The quantification of these impacts can also be used, along with other information, to evaluate the costs and benefits of alternative interventions that may be adopted to reduce ambient air pollution.
Estimating the cost of ambient air pollution involves three steps: (a) estimating population exposure to pollution; (b) estimating the health effects of such exposure; and (c) assigning a monetary value to the illnesses and premature deaths caused by ambient air pollution. The objective of this article is to review the main methodological approaches used for these three main steps, with particular focus on developing countries, and to provide some estimates of health effects and costs.
The first section explores the methods used for estimating population-weighted exposure to ambient air pollution. This is followed by a presentation of the methods for assessing the health effects of such exposure and discussions on the cost of environmental health risks and the causes of uncertainty in quantifying the effects and costs of ambient air pollution. The final section concludes.
In 2006, the WHO reduced its Ambient Air Quality Guideline (AQG) limits to an annual average ambient concentration of 10 μg/m3 of PM2.5 and 20 μg/m3 of PM10, in response to increased evidence of health effects of exposure to very low concentrations of PM (WHO, 2006). The development of the 2006 AQG was largely based on available epidemiological research of long-term exposure to PM2.5 conducted in the United States. Since then, research in Asia, Europe, and North America has documented a broader range of health effects caused by PM2.5 air pollution, many of which occur at even lower levels than the AQG limits. In addition to providing evidence of the effects of long-term exposure to PM2.5 on mortality, mostly as a result of cardiovascular and pulmonary disease and lung cancer, available evidence suggests that air pollution may also have effects on diabetes, neurological development in children, and neurological disorders in adults (WHO, 2013). Scientific research has found strong linkages between PM2.5 pollution and diabetes mellitus (Bowe et al., 2018; GBD 2019 Risk Factors Collaborators, 2020). Based on growing scientific evidence on the health effects of exposure to PM2.5 at very low concentrations, the WHO started the process of updating and revising the AQGs in 2016 (WHO, 2015).
Population exposure to ambient air pollution is typically defined as the concentration of a pollutant in the outdoor air environment to which a population is exposed. To capture nationwide health effects of ambient air pollution it is important to use nationwide population exposure. This has been undertaken in the Global Burden of Disease (GBD) studies since 2010. Estimating population exposure to ambient air pollution is, however, a challenge in low- and middle-income countries because many of these countries lack reliable air monitoring networks to systematically collect data on air quality.
Several studies have been completed since the 1990s to estimate the GBD. The first was commissioned by the World Bank to inform the preparation of the World Development Report in 1993. Comprehensive evaluations on ambient air pollution were also supported by WHO and the World Bank in 2000 and 2004. In the GBD 2010, the scope of the GBD was updated with support from the Bill and Melinda Gates Foundation to include 235 causes of death and disability from diseases and injuries and 67 risk factors (including ambient air pollution) in 187 countries and territories in 21 regions of the world. The Institute for Health Metrics and Evaluation (IHME) became the main provider for a broad range of GBD estimates that underpinned the preparation of GBD reports for 2010, 2013, 2015, 2016, 2017, and 2019. GBD 2019 included estimates of deaths and disability from 369 diseases and injuries and 87 risk factors in 204 countries and territories (GBD 2019 Diseases and Injuries Collaborators, 2020; GBD 2019 Risk Factors Collaborators, 2020).
The estimates of the effects of ambient air pollution from these studies have increased from 3.4 million deaths in 2010 to 4.5 million deaths in 2019, largely as a result of improved understanding of the health consequences of air pollution, but also due to more comprehensive population exposure assessments. As much as 92%–95% of the deaths were from ambient PM2.5 and 5%–8% from ambient ozone pollution.
The GBD 2010 provided national, regional, and global estimates of health effects of ambient PM2.5 air pollution based on estimates of nationwide exposures to ambient PM2.5. Estimates of health effects prior to GBD 2010 were generally limited to PM2.5 estimates for cities with a population greater than 100,000, thus missing the rural population and a substantial share of the urban population. The GBD 2010 accomplished the task of providing nationwide estimates by employing a combination of satellite imagery, chemical transport modeling, and ground-level PM2.5 and PM10 measurements to arrive at nationwide population-weighted ambient PM2.5 exposure estimates.1
Ground-level measurements of PM2.5 or PM10 that were used to estimate population exposure in GBD 2010 covered fewer than 700 locations (Brauer et al., 2012). Two-thirds of the locations were in the high-income countries of Western Europe, North America, and Asia Pacific/Australasia. Low- and middle-income countries in the East Asia and Pacific region had 133 locations, most of them in China. The lowest number of locations per region was in sub-Saharan Africa (8) and the Middle East and North Africa (9).
The GBD 2013 expanded the number of ground-level measurements of PM2.5 and PM10 to 4,073 data points from 3,387 unique locations (Brauer et al., 2016). This included measurement data from the information sources used by the GBD 2010 and new data, especially from China and India, including data compiled from a literature survey (van Donkelaar et al., 2015) and the WHO ambient air pollution database.
The GBD 2015 updated the ground-level measurement database to include more recent data and additional locations. This involved collaboration with WHO and contributions to the WHO Ambient Air Quality Database 2016. Thus, the GBD 2015 and the GBD 2016 employed data from 6,003 ground monitors in about 3,000 human settlements ranging in size from a population of a few hundred to over 10 million (GBD 2015 Risk Factors Collaborators, 2016; GBD 2016 Risk Factors Collaborators, 2017; WHO, 2016).
The GBD 2017 used the WHO Ambient Air Quality Database 2018 Update with PM10 and PM2.5 from about 9,690 stations in nearly 4,400 locations (defined geographic areas) in 108 countries (GBD 2017 Risk Factors Collaborators, 2018). The GBD 2019 also used this updated database, along with additional measurement data mainly from the United States, Canada, European Union, Bangladesh, and China, and PM measurement data from U.S. embassies and consulates. Thus measurement data from 10,408 ground monitors from 116 countries were used by the GBD 2019 (GBD 2019 Risk Factors Collaborators, 2020). Nevertheless, ground monitoring remains particularly scarce in low-income countries and sub-Saharan Africa.
Analysis of the WHO Ambient Air Quality Database 2018 Update finds that PM10 is presented from well over 5,500 monitoring stations in about 3,500 locations in 99 countries and PM2.5 from over 4,100 stations in about 2,600 locations in 90 countries.2 The number of PM2.5 monitoring stations in relation to population varies greatly across regions and country income classification (figure 1). There are 0.4–0.5 million people per PM2.5 monitoring station in Europe and Central Asia and North America. In contrast, there are 15–28 million people per station in South Asia and sub-Saharan Africa. By income classification, there are 0.37 million people per station in high-income countries while there are nearly 65 million people per station in low-income countries.
As a result of the increase in ground-level monitoring data used by the GBD over time, estimated population exposure to PM2.5 in GBD 2010, GBD 2013, GBD 2015, GBD 2016, GBD 2017, and GBD 2019 often show large differences in PM2.5 at national and regional levels for the same year. These differences are not due to actual changes in ambient PM2.5 when estimates are for the same year, but rather due to the increase in the number of ground-level monitoring locations as well as the evolution in satellite imagery/chemical transport model (CTM) estimation techniques and the method of calibrating the satellite imagery/CTM estimates with the ground-level measurements. (These issues are discussed in detail in Brauer et al., 2012, 2016; Ostro et al., 2018; Shaddick et al., 2018; van Donkelaar et al., 2015, 2016.)
Figure 2 presents the global average annual population weighted ambient PM2.5 for the year 2010 as estimated by the GBD 2010–2019. GBD 2010 and GBD 2013 estimated substantially lower ambient PM2.5 than the subsequent GBD studies.3
Regional variations in estimated annual exposure from GBD 2010 to GBD 2019 are substantially larger than the global variation. Figure 3 presents regional exposure from GBD 2010 to GBD 2019 for the year 2010. Absolute variations are largest in South Asia, the Middle East and North Africa, and sub-Saharan Africa. The percent variation in Latin America and the Caribbean is also large. The largest variations were between GBD 2010 and GBD 2015 while variations were substantially smaller between GBD 2015 and GBD 2019 as a substantially larger number of ground-level monitoring data were used starting with the GBD 2015.
The GBD also provides time trends in annual ambient PM2.5 exposure estimates. Global population weighted exposure increased by about 10% from 2010 to 2014 and declined by about 10% from 2014 to 2019 according to GBD 2019 (figure 4). The increase to 2014 and subsequent decline took place primarily in South Asia and East Asia and Pacific. Over the period 2010–2019, regional exposure levels increased in South Asia by 8% and in sub-Saharan Africa by 3%, declined by 8%–11% in Latin America and the Caribbean, the Middle East and North Africa, and East Asia and Pacific, and by 16% in Europe and Central Asia and North America (figure 5).
Global population weighted exposure to ambient PM2.5 was 43 μg/m3 in 2019 (figure 4). This is over four times higher than WHO’s AQG of 10 μg/m3 for annual PM2.5. Population weighted ambient PM2.5 ranged from less than 10 μg/m3 in North America to nearly 80 μg/m3 in South Asia. Ambient PM2.5 in Latin America and the Caribbean and East Asia and Pacific were half the levels in sub-Saharan Africa, East Asia and Pacific, and the Middle East and North Africa, and less than one-third the level in South Asia.
By country income level, estimated ambient PM2.5 exposures in 2019 was highest in lower middle-income countries followed by low-income, high-income non-OECD, and upper middle-income countries. Exposures in high-income OECD countries were one-fifth to one-third the level in the former groups. Exposures in low-income countries were constant from 2010 to 2019, increased by 5% in lower middle-income countries, declined by 3% in high-income non-OECD countries, and declined by 11%–12% in upper middle-income and high-income OECD countries (figure 6). The high-income non-OECD country group is dominated by arid gulf countries with high ambient concentrations of desert dust. Figure 7 shows the population weighted annual ambient PM2.5 exposures across the world.
Satellite data and CTM have been used to estimate PM2.5 concentrations and population exposure in countries with limited or no monitoring of PM2.5. However, it is important to highlight that the use of these approaches is not a reliable substitute for actual ground level monitoring of PM2.5. Alvarado et al. (2019) evaluated the use of CTM and statistical methods using satellite data in nine cities in low- and middle-income countries. They found that the uncertainty in the satellite-based estimates of the daily average PM2.5 concentrations tended to be 21%–77% for the statistical methods and 48%–85% for the CTM-based methods. The satellite-based methods seemed to work best in low altitude, inland cities like Hanoi and Delhi, but still had errors of 43%–60% in predictions of daily average PM2.5 concentrations at sites within these cities. When satellite data is calibrated using ground-level data, it may actually reduce the number of monitoring stations needed on the ground to characterize PM2.5 concentrations with a city. Thus, satellite-based methods might be considered as a complement, rather than as a substitute for ground-level networks, but the reliability of satellite-based methods needs to be assessed on a case-by-case basis.
Estimating the Health Effects of Ambient Air Pollution
Once population exposure to ambient PM2.5 has been established, estimating the health effects of exposure involves the use of relative risk functions that relate exposure levels and risk of health outcomes. The evolution of these risk functions is discussed in this section.
Initial Efforts to Assess the Health Effects of Exposure to Ambient Air Pollution
Pope et al. (2002) found elevated risk of cardiopulmonary (CP) and lung cancer (LC) mortality from long-term exposure to ambient PM2.5 in a study of a large population of adults aged 30 years or more in the United States. CP disease included heart disease, stroke, and acute and chronic respiratory diseases. The WHO used this study when it estimated global mortality from outdoor ambient air pollution (WHO 2004, 2009).
Ambient PM2.5 concentrations are, however, substantially higher in many locations in low- and middle-income countries than in the United States. Ostro (2004) therefore recommended a log-linear function for estimating CP and LC mortality that could be applied in low- and middle-income countries. The log-linear function, primarily based on Pope et al. (2002), is an exposure-response function expressed as relative risk (RR) of CP and LC mortality from exposure to PM2.5. Exposure to PM2.5 was expressed as annual average ambient concentration (X), with a specified lower threshold value (X0), for which any value below X0 was assumed to have none or minimal health effects (e.g., levels PM2.5 <5 μg/m3; table 1; figure 8). The β coefficient in the function gives the rate of increase in RR from increasing concentrations of PM2.5. Ostro (2004) also provided a relative risk function for acute lower respiratory infections mortality among children from exposure to PM10.
Table 1. PM Exposure-Response Functions
Suggested β Coefficient
Adults 30+ years of age
RR = [(X + 1/(X0 + 1]β
Long-term exposure to PM2.5
Adults 30+ years of age
RR = [(X + 1/(X0 + 1]β
Long-term exposure to PM2.5
Children under 5 years of age
Short-term exposure to PM10
Source: Ostro (2004).
Subsequent studies by Pope et al. (2009, 2011) supported a log-linear function relating health effects and PM2.5 exposure. These studies derived the shape of the PM2.5 exposure-response curve based on studies of mortality from outdoor ambient PM2.5 air pollution, second-hand cigarette smoke, and active cigarette smoking.
Integrated Exposure Response Function
The GBD 2010 study developed an integrated exposure response (IER) function by taking the findings of Pope et al. (2009, 2011) a few steps further. The IER has two main advantages over the methods that were previously used to estimate the health effects of ambient air pollution. First, it allows for prediction of relative risk (RR) of health outcomes over a very large range of PM2.5 concentrations. Second, it enables the estimation of health effects based on a country’s specific baseline structure of mortality, instead of the one-size-fits-all function that was previously available and that ignored intercountry differences in the structure of mortality (Burnett et al., 2014; Shin et al., 2013).
Ischemic heart disease;
Cerebrovascular disease (stroke);
Chronic obstructive pulmonary disease;
Lung cancer; and
Acute lower respiratory infections
Two additional health outcomes were added in GBD 2017 and GBD 2019, respectively:
Diabetes Type II
The GBD also includes cataract for morbidity from household air pollution. The IER functions from the GBD 2019 for the six major mortality and morbidity health outcomes are presented in figure 9.
The shape of the relative risk curve or rate of increase in relative risk from increasing concentrations of PM2.5 has changed quite substantially for some of the health outcomes from the GBD 2010 to GBD 2019. These changes have occurred as new evidence becomes available. The relative risks applied in GBD 2010 to GBD 2019 are presented below in figures 10-14 as a continuous curve for PM2.5 exposure ranging from 10 to 100 μg/m3, and for the discrete exposures of 20, 40, and 80 μg/m3. These exposure levels correspond approximately to the average population-weighted PM2.5 ambient exposure in 2019 in Latin America and the Caribbean and Europe and Central Asia (approx. 20 μg/m3), East Asia and Pacific, sub-Saharan Africa and Middle East and North Africa (approx. 40 μg/m3), and South Asia (approx. 80 μg/m3).
Ischemic Heart Disease and Stroke
Ischemic heart disease (IHD) and stroke generally contribute the largest share of mortality from ambient PM2.5. The IER functions for these two disease outcomes are specific for each 5-year age group for adults 25 years of age and older. Relative risks for age group 75–79 years are presented in figures 10 and 11. The relative risk of IHD and stroke from PM2.5 in the GBD 2010 was substantially larger than the risk applied in GBD 2013 to GBD 2017. The relative risk applied in GBD 2019 is almost of similar magnitude as in GBD 2010.
Chronic Obstructive Pulmonary Disease
The IER function for chronic obstructive pulmonary disease (COPD) is applied to adults 25 years of age and older as a group. The relative risk of COPD applied in GBD 2015 to GBD 2017 was substantially larger than in GBD 2010 and GBD 2013. The risk is largest in GBD 2019 at higher exposure levels (figure 12).
The IER function for lung cancer (LC) is applied to adults 25 years of age and older as a group. The relative risk of LC was largest in the GBD 2013 and was similar in GBD 2010 and GBD 2015–2017. The risk applied in GBD 2019 is somewhat larger for PM2.5 exposures up to about 90 μg/m3 (figure 13).
Acute Lower Respiratory Infections
The IER function for ALRI is now applied to all ages in the GBD. It was, however, initially only applied to children under 5 years of age. The relative risk of ALRI applied in the GBD 2019 is somewhere between the risk in GBD 2013 and the risk in GBD 2010 and GBD 2015–2017 (figure 14).
Diabetes type II from PM2.5 exposure was added in GBD 2017 and GBD 2019. The relative risk increases rapidly up to a PM2.5 exposure level of 30 μg/m3 and flattens out for higher exposure levels (figure 9). Neonatal mortality from PM2.5 exposure was added in the GBD 2019. However, this cause of mortality from ambient PM2.5 accounts for only 3% of total deaths from PM2.5 according to the GBD 2019 estimates.
An Application to Two Countries
The changes in the relative risk of health effects in GBD 2010 to GBD 2019 have impacts on estimated mortality from ambient PM2.5. This is illustrated for Mexico and India in figure 15 by using ambient PM2.5 exposure from GBD 2019 and relative risks from GBD 2010 to GBD 2019.
In Mexico, with a population-weighted annual ambient PM2.5 exposure of 20 μg/m3 in 2019 according to the GBD 2019, annual deaths from ambient PM2.5 are estimated at 21,000 to 29,000 in 2019 from the five health outcomes common to GBD 2010 to GBD 2019 (i.e., IHD, stroke, COPD, LC, and ALRI). For relative risks from GBD 2015 to GBD 2019, estimated deaths are within a relatively narrow range of 24,000–26,000. The narrow range is because some relative risks increased while others declined. More significant, in the case of Mexico, is the addition of diabetes type II. Diabetes type II mortality constituted 30% of deaths from ambient PM2.5 based on the relative risks from GBD 2019.
In India, with a population-weighted annual ambient PM2.5 exposure of 83 μg/m3 in 2019 according to the GBD 2019, the variation in estimated deaths from ambient PM2.5 is more pronounced. This is mainly due to larger differences in relative risks across GBD 2010 to GBD 2019 for the higher PM2.5 exposure level in India. Estimated annual deaths ranged from 484,000 to 889,000 in 2019 from the five health outcomes common to GBD 2010 to GBD 2019. For relative risks from GBD 2015 to GBD 2019 estimated deaths are within a somewhat narrower but still large range of 629,000–889,000. The difference in annual deaths using relative risks from GBD 2017 and GBD 2019 is mainly due to the difference in risk of IHD and stroke at higher PM2.5 exposure levels. Diabetes type II mortality constitutes a very minor share of total deaths from ambient PM2.5.
The method for estimating the health effects of ambient PM2.5 is explained in GBD 2019 Risk Factors Collaborators (2020, Supplementary Appendix 1, pp. 78–115). The GBD 2019 first estimates the total joint health effects of ambient PM2.5 and household air pollution PM2.5 from household use of solid fuels for cooking, and then apportions the health effects to ambient PM2.5 and household PM2.5 air pollution. The total annual health effects of annual PM2.5 exposure are estimated as follows:
where is annual cases of deaths or illness from PM2.5 among the exposed population, is baseline annual cases of deaths or illness among the same population, and (population attributable fraction) is the fraction of baseline cases attributable to PM2.5 exposure among this population. and are calculated for each type of health effect covered by the GBD 2019.
The for each health effect included in the GBD 2019 is calculated as follows:
where is the share of the population that is exposed only to ambient PM2.5 (i.e., the population not using solid fuels for cooking), is the share of the population that uses solid fuels for cooking, is the relative risk of health effects from ambient PM2.5 among the population exposed only to ambient PM2.5, and is the relative risk of health effects from PM2.5 among the population exposed to household air pollution PM2.5. As the whole population is exposed to at least some level of ambient PM2.5, then .
The size of the relative risks of health effects among the population only exposed to ambient PM2.5 is:
where is the relative risk of health effects at annual PM2.5 = and is the relative risk at PM2.5 = (theoretical minimum-risk exposure level). These relative risks are reported by the GBD 2019 for each type of health effect for a range of annual PM2.5 from 0.01 to 2500 μg/m3 (GBD 2019 Risk Factors Collaborators, 2020, Supplementary Appendix). TMREL may be chosen within the range of 2.4–5.9 μg/m3 that is used by the GBD 2019 or may be chosen to be larger or smaller. may also be chosen as zero in which case .
The size of the relative risks of health effects among the population exposed to household air pollution PM2.5 is:
where is the relative risk at exposure level . This exposure level includes both exposure to PM2.5 from the use of solid fuels () and exposure to ambient is a so-called personal exposure level that is typically measured over a 24- to 48-hour period by a measurement device attached to a person (and assumed to reflect long-term exposure). The exposure level is generally different for each household member due to differences in activity patterns. Most personal exposure measurement studies have been for adult women, who usually are exposed to the highest level of PM2.5 in households cooking with solid fuels. The GBD 2019 uses a fraction of adult women exposure equal to 0.64 for adult men and 0.85 for children. Because of these differences in exposure levels, is calculated separately for adult females, adult males, and children under the age of 5 years.
The GBD 2019 then apportions the to ambient and household air pollution as follows:
where and are the population attributable fractions of deaths and illnesses due to ambient PM2.5 and PM2.5 household air pollution from the use of solid fuels, respectively; is annual ambient PM2.5; is personal PM2.5 exposure from the use of solid fuels; and is the share of the population using solid fuels for cooking.
This approach to apportioning the health effects of PM2.5 ensures that:
The GBD 2019 estimates , , and at small geographic units over which health effects are summed to national, state, province, or city level. Each geographic area is a 0.1° × 0.1° grid (corresponding to 11 × 11 km at equator).
Equations 5 and 6 are, however, less accurate if the are estimated for large geographic units, such as a city or subnational region with a single population-weighted PM2.5 exposure level. In this case, an alternative approach to apportioning the health effects is outlined in the following three steps:
which are “partial” for the population exposed only to ambient PM2.5 and the population exposed household air pollution (the population using solid fuels). in the “partial” for household air pollution is for exposure level . The next step involves separating the effects of AAP and HAP in Equation 9 and adding the effects to the in Equations 8 and 9.
The result in Step 2 is such that:
The final step involves adjusting the two downward so that the sum is equal to .
The two approaches give identical for and . But for , the GBD approach results in a smaller for ambient PM2.5 and a larger for household air pollution PM2.5 than the alternative approach described in steps 1–3. The difference can be quite large when PM2.5 exposure from household air pollution is substantially higher than from ambient PM2.5 and increases as approaches 0.5 from 0 and from 1.0. The approach in steps 1–3 is therefore recommended if estimation of health effects of ambient air pollution is undertaken with exposure data reflecting relatively large geographic units, such as city by city, state by state, province by province, or urban and rural.
Global and Regional Health Effects of Ambient Air Pollution
The GBD 2019 estimated a global population-weighted exposure to ambient PM2.5 of 43 μg/m3 (HEI, 2020). Based on this exposure and the relative risk functions from GBD 2019, an estimated 4.1 million people died from ambient PM2.5 air pollution in 2019 (GBD 2019 Risk Factors Collaborators, 2020). Nearly three million of these deaths were in East Asia and Pacific and South Asia, of which 2.4 million occurred in China and India. By income group, over 3.6 million deaths were in lower and upper middle-income countries (figure 16). Additionally, ambient PM2.5 globally caused over 13 million years lost to illness.4
The estimated annual deaths from ambient PM2.5 ranged from 14 per 100,000 population in North America to 76 per 100,000 in East Asia and Pacific. By income group, the death rates are highest in lower middle-income and upper middle-income countries. The rates reflect population-weighted ambient PM2.5 exposure and population age distribution and baseline health status. Thus, while regional ambient PM2.5 exposure is relatively low in the Europe and Central Asia region, the death rate from ambient PM2.5 in this region is about twice as high as in sub-Saharan Africa and Latin America and the Caribbean due to high baseline mortality rates and high vulnerability of the aging population to ambient PM2.5 in Europe and Central Asia (figure 17).
Estimated deaths from ambient PM2.5 constitute less than 2% of total deaths from all causes in North America to 10%–12% of all deaths in Middle East and North Africa, East Asia and Pacific, and South Asia. Estimated death rates from ambient PM2.5 are highest in middle-income countries and high-income non-OECD countries (figure 18).
Economic Valuation Methods
Economists have most often used two distinct methods of valuation of mortality to estimate the social cost of premature deaths: the human capital approach (HCA) and the value of statistical life (VSL). The HCA measures the cost of mortality as the total lost future earnings of an individual, but ignores less tangible elements such as leisure, unpaid work, and social engagement. The VSL approach substituted the HCA in policy assessments because it allows such elements—which are hard to quantify—to be integrated and assessed in an evaluation, by participants placing a value on reduced risk of death (Andersson & Treich, 2009). Although VSL methods are widely used for policy evaluations, the approach is not free of criticism and potential bias. For instance, the Diamond and Hausman (1994) criticism of contingent valuation as a methodology talks about the kinds of biases one might find in CVM methods that could be present in VSL estimates based on contingent valuation.
Human Capital Approach
The HCA is based on the economic contribution of an individual to society over the individual’s lifetime. Death involves an economic loss that is approximated by the loss of all future income of the individual. Future income is discounted to reflect its value at the time of death. The discount rate commonly applied is the rate of time preference. Thus, according to the HCA, the social cost of mortality is the discounted future income of an individual at the time of death. If the risk of death, or mortality risk, is evenly distributed across income groups, average expected future income is applied to calculate the social cost of death (Dixon et al., 1994). Mathematically, the present value of future income, or human capital value (), is expressed as follows:
where is present value of income in year 0 (year of death), is annual growth in real income, and is the discount rate (rate of time preference). Equation 15 allows for income to start from year and end in year at retirement. Usually, for adults while for children.
An important issue that was often raised regarding the is the application of this valuation approach to individuals who did not participate in the economy (e.g., individuals that did not have an income, such as the elderly, family members taking care of the home, and children). One may think of an extension of the that recognizes the value of nonpaid household work at the same rate as the average income earner, or at a rate equal to the cost of hiring a household helper. In this case, the was applied to the death of nonincome earners (whether or not children would become income earners or take care of the home during their adult life). In the case of the elderly, the would assign zero economic value to old individuals who had either retired from the workforce or did not make significant contributions to household work. This obviously was a serious shortcoming of the approach. An alternative option was to calculate the for adults based on the number of years lost to premature death.
An estimate of average annual income is needed to estimate the and may be estimated using the following equation:
where is gross domestic product (GDP) per capita in the year of death, is the labor income share of GDP, and is the labor participation rate (percentage of total population).
The human capital approach has been largely rejected as an inappropriate measure of the value of reducing mortality risks because it does not capture individuals’ willingness to pay for small risk reductions and as such does not capture the value associated with avoided pain and suffering, dread, and other risk factors that are thought to affect value (Viscusi, 1993).
Value of Statistical Life
The value of statistical life (VSL) is calculated based on individuals’ valuation of changes in mortality risk. Although nonspecialists might think that the VSL provides a monetary judgment on the worth of an individual life, it in fact aggregates the value that multiple individuals would accept as compensation for mortality risk increases or be willing to pay for reductions in mortality risk (Bosworth et al., 2017). Everyone in society is constantly facing a certain risk of dying. Examples of such risks are occupational fatality risk, risk of traffic accident fatality, and environmental mortality risks. It has been observed that individuals adjust their behavior and decisions in relation to such risks. For instance, individuals demand a higher wage (a wage premium) for a job that involves a higher occupational risk of fatal accident than in other jobs, individuals may purchase safety equipment to reduce the risk of death, and/or individuals and families may be willing to pay a premium or higher rent for properties (land and buildings) in a cleaner and less polluted neighborhood or city.
VSL is calculated as follows: For instance, it may be observed that a certain health hazard has a mortality risk of 2.5/10,000. This means that one individual dies from this hazard for every 4,000 individuals exposed. If each individual on average is willing to pay $40 for eliminating this mortality risk, then every 4,000 individuals are collectively willing to pay $160,000. This is the VSL, or the value that individuals collectively are willing to pay to avoid one death. Mathematically this can be expressed as follows:
where is the average willingness to pay per individual for a mortality risk reduction of magnitude . In the numerical example above, and . Thus, if 10 individuals die from the health risk in this example, the cost to society would be:
The main approaches to estimating VSL are through revealed preferences and stated preferences of people’s WTP for a reduction in mortality risk or their willingness to accept (WTA) an increase in mortality risk. Since mortality risk reduction is a nonmarketed good (i.e., there is no market price for this reduction), we must rely on nonmarket valuations of this WTP. The valuation methods can be split into two camps: stated preferences and revealed preferences.
Revealed preference studies examine data where individuals have, either explicitly or implicitly, made a trade-off decision. This is usually preferred by economists, since the studies consider repeated real-life market interactions. Thus, individuals can be expected to correctly identify the choices in front of them and can be expected to be acting on their true underlying preferences regarding the nonmarketed good. This type of analysis requires data of markets where individuals are assumed to be informed while making choices over a set of alternatives with varying amounts of risk. In practice, with the goal of estimating VSLs, this has usually boiled down to using labor markets to estimate the trade-off between wages and occupational risk employing hedonic regressions. These hedonic regressions try to break down a certain choice/outcome into the many composite factors resulting in said choice/outcome, with the end goal of identifying the magnitude of wage premium associated with an elevated occupational fatality risk.
While in theory revealed preferences are preferred in practice for air pollution externalities, this is usually not the most apt method for various reasons. First and foremost, there is a lack of data in which individuals are actively making a choice between exposure to PM2.5 pollution and lack of exposure. At the same time, if we do have such data, we must also assume that individuals are correctly and accurately judging the risks of air pollution, a rather strict assumption. Finally, even with the previous two assumptions met, we are forced to cast our scope on the specific markets under analysis; that is, external validity to the general population is not guaranteed, but is a desired characteristic of any VSL study (Andersson & Treich, 2009).
In the face of these challenges, the most common solution is to resort to the other main method of estimating WTP, which is stated preferences and choice models. This method consists of building surveys and directly asking individuals about their preferences. The method’s flexibility has allowed its utilization in a wide range of areas (Hammitt & Graham, 1999). The method usually estimates preferences in one of two ways: (a) directly asking respondents for their WTP for a decrease in risk (Alberini & Krupnick, 2000; Istamto et al., 2014) or (b) through choice modeling surveys that ask respondents to choose between discrete choices to form an implicit ranking of preferences (Kan & Chen, 2004; Yoo et al., 2008).
While flexible, these studies face a main difficulty of using data that is based on choices made by survey takers responding to hypothetical situations. Thus, these surveys cannot guarantee that respondents are answering truthfully, and the respondents may not accurately reflect their underlying preferences because of lack of knowledge or because they are not faced with the actual trade-off in real life. Therefore, stated preference studies usually pay great attention to the survey design in an effort to elicit as accurate a trade-off calculation by respondents as possible. As Kniesner and Viscusi (2019) explained, the main characteristics of the survey design are “the sample composition, a description of the health outcome, the starting risk level, the change in the risk, the mechanism by which the risk is altered, the payment mechanism, and the nature of the trade-off elicitation approach.” Potential issues notwithstanding, particularly in country contexts where revealed preference data is not readily available, stated preference studies are the principal approach for estimating WTP for mortality risk reduction.
Studies of VSL have been carried out in many countries and several meta-analyses of these studies have been conducted. Meta-analyses assess characteristics that determine VSL, such as household income, size of risk reduction, other individual and household characteristics, and often characteristics of the methodologies used in the original WTP studies (Masterman & Viscusi, 2018; OECD, 2012).
Most of the early meta-analyses of VSL are entirely or predominantly based on hedonic wage studies. A meta-analysis prepared for the OECD was, however, exclusively based on stated preference studies, arguably of greater relevance for valuation of mortality risk from environmental factors such as air pollution than hedonic wage studies. These stated preference studies are from a database of more than 1,000 VSL estimates from multiple studies in over 30 countries, including in developing countries (OECD, 2012).
VSL studies are, however, only available from a minority of countries globally. A commonly used approach to estimate VSL in a country that lacks such studies is to use a benefit transfer approach based on meta-analyses of VSL studies from other countries. The World Bank (2016) presented a benefit transfer methodology for valuing mortality from air pollution that draws on the empirical literature of VSL, especially OECD (2012). The methodology was applied in the publication by the World Bank and IHME (2016) on the global cost of air pollution. The proposed benefit transfer function is:
where is the estimated VSL for country in year , is the average base VSL in the sample of OECD countries with VSL studies (US$ 3.83 million), is GDP per capita in country in year , and is the average GDP per capita for the sample of OECD countries ($37,000), and ɛ an income elasticity of 1.2 for low- and middle-income countries and 0.8 for high-income countries. All values are in purchasing power parity (PPP) prices. must, therefore, be converted to local currency using PPP exchange rates, available in the World Development Indicators by the World Bank. The study by the WHO and OECD (2015) mentioned that the comparatively lower value assigned to the risk reductions captured in the VSL by individuals in lower-income countries is not a normative judgment by economists, but is a recognition of the fact that citizens from a specific country execute their trade-offs based on their own income level without references to the resources of other countries.
Key criticisms of VSL include individuals’ biases and perceptual limitations, which can lead to overestimation of small and catastrophic risks, and to uncertain probabilities. In addition, the type of study used, including both revealed and stated preferences, can result in wide variations of VSL estimates (Bosworth et al., 2017).
VSL and Income Elasticities
The VSL literature lacks consensus on the choice of income elasticity to use when transferring VSL estimates from one country to another for countries that lack such estimates. In adopting a benefit transfer approach, three main parameters are taken into consideration: a baseline VSL—usually based on estimates from the United States or European countries—the income level for the country of interest, and the income elasticity of the VSL (OECD, 2012). The income elasticity that should be used when transferring VSL estimates to countries with much lower income levels is still subject to debate.
Divergent values of VSL have been observed in studies that use different samples and methods, leading to different suggestions of the level of VSL that should be used. Hammitt and Robinson (2011) observed that, although U.S. regulatory agencies assume that a 1% change in real income over time will lead to a 0.4%–0.6% change in the VSL, this estimate does not apply to countries with lower income. When transferring values between high- and lower-income countries at an elasticity less than 1.0, VSL estimates may appear large in comparison to income. They also called attention to using elasticities greater than 1.0. Although supported by research, by cross-country comparisons, and by research that estimates the VSL by income quintile, caution is needed when applying these higher elasticities because the resulting VSLs appear smaller than expected future earnings or consumption in some cases, contrary to theory.
Suggestions have been made for elasticities in the context of lower-income economies. Masterman and Viscusi (2018) conducted a meta-analysis of stated preference estimates of VSL to analyze the variation in the income elasticity across countries. About 40% of the VSL estimates were from middle-income countries, while none were from low-income countries. They concluded an income elasticity of VSL of approximately 1.0 for low-income and lower-middle income countries and 0.55–0.85 for upper middle-income and high-income countries. Similar meta-analyses were conducted by Lindhjem et al. (2011) and OECD (2012),5 with results from the latter being widely used by countries to transfer VSL estimates to their contexts.6 Masterman and Vascusi (2018), however, were the only ones among these publications to include estimates of willingness to accept in their analysis and controlled for possible sample selection bias and the influence of covariates, such as the type of risk.
Viscusi and Masterman (2017) estimated the VSL for 189 countries across the globe (figure 19), using the gross national income per capita as estimated by the World Bank as a measure of income. All VSLs were calculated using an income elasticity of 1.0 and based on a U.S. VSL of $9.6 million (2015 prices).
Estimating the Cost of Morbidity
The cost of morbidity includes work absenteeism and medical treatment. The willingness to pay to avoid pain and suffering and loss of leisure time can also be added to this cost, but estimates are generally not available for most countries. The most frequently used methods to estimate morbidity from ambient air pollution, especially in low and middle-income countries, are the cost of illness (COI) approach and willingness to pay based on contingent valuation methods or discrete choice experiments.
Studies that have valued both mortality and morbidity from ambient PM2.5 have generally found that morbidity counts for around 10%–20% of total health costs. Most of these studies were conducted in high-income countries, where health system facilities and life standards are usually better than those in low- and middle-income countries. Thus these cost shares do not necessarily reflect shares in developing countries. A study in Turkey designed to estimate the willingness to pay (WTP) for gains in health and life expectancy due to ambient air quality, using country-specific value of life year (VOLY) and the value of the healthier and longer life (VHLL), found that about 80% of the WTP was allocated to VHLL and only 20% for VOLY. The authors linked these results to religious beliefs of the respondents (Ara & Tekesin, 2017). Another explanation, although speculative, is that in many low- and middle-income countries a considerable percentage of the working population is in the informal market and therefore place a greater value on their current health condition (as a commodity to generate income) than on valuations about their death in the future.
Using data from the U.S. Census of Fatal Occupational Injuries, Gentry and Viscusi (2016) suggested a way to disentangle the mortality and morbidity components. Although the authors concluded that the mortality component is still the dominant component in the valuation of risks in a high-income country, it is possible that the share in costs in low- and middle-income countries is different. The authors proposed disaggregating the valuation of risk into two components, a value of fatality risk (VFR) and a value of morbidity risk (VMR) associated with traumatic job injuries (not illnesses), where the total valuation of risk would equal VFR plus VMR. They demonstrated the construction of the VFR and VMR based on wages of employers, supply and demand for risky jobs, and number of days survived in each category of job. The VMR component is thus represented as the number of days away from work and, conditional on death, by the number of days after the injury and until death. The final equation is then defined as:
where is the annual frequency of fatalities and is the average number of days after the injury and until death per worker, by industry and occupation .
Using this approach, Gentry and Viscusi (2016) estimated that morbidity accounts for around 6%–25% of the total morbidity and mortality risk effect. Equation 20 suggests the use of data from workers in the formal market, using revealed data and not stated labor market data. This clearly reflects the context of high-income and some upper middle-income countries; however, it also gives an opportunity to explore other parameters, such as work absenteeism and medical treatment expenses, as the independent variable is wage. In other settings, wage can be translated as income, although the challenge of defining income for those in the informal market can be considerable.
Work absenteeism and medical treatment expenses—not only from the perspective of patients and their families but also from the perspective of the healthcare system—are well explored in COI models. For example, Gao et al. (2015) assessed the health and economic impact of a severe haze event in Beijing, China, by using epidemiological modeling, a COI model, and a WTP study. Results from their study are presented in table 2. Mortality accounts for $189 million and morbidity accounts for $65 million, with hospitalizations and clinic visits accounting for more than half of these costs. From the total economic loss ($254 million), the morbidity component accounted for about 26% of the total costs.
Table 2. Economic Loss Estimates by Type of Study, Beijing, China, 2013 prices
Health Endpoints by Type of Study
Economic Cost (million US$)
188.7 (134; 243.4)
31.9 (18; 44.5)
7.6 (4.3; 10.8)
Willingness to pay (WTP)
Source: Adapted from Gao et al. (2015).
For ozone, Selin et al. (2009) estimated that health costs due to global ozone pollution above preindustrial levels by 2050 would be $580 billion (in 2000 prices) and that mortality from acute exposure would exceed two million deaths.
Cost of Health Effects of Ambient Air Pollution
The health effects of ambient PM2.5 have a global cost estimated at nearly $3.7 trillion in 2019, or $6.4 trillion in purchasing power parity (PPP) adjusted dollars. This was equivalent to 4.8% of global GDP in 2019 (PPP adjusted).7 The cost ranges from an equivalent of 1.7% of GDP in North America to an equivalent of 7.3% of GDP in East Asia and Pacific. The cost rises with income from an equivalent of 1.3% of GDP in low-income countries to an equivalent of 7.1% of GDP in upper middle-income countries while 2.8%–4.3% in high-income OECD and high-income non-OECD countries (figure 20).
Although morbidity is an important component of health-related costs of ambient air pollution in low- and middle-income countries, most of the literature and estimates of COI and WTP for avoiding illness are from high-income countries, with the bulk of these studies coming from the United States and the United Kingdom. In developing economies, most existing studies are largely limited to China, with few other examples from Brazil, Chile, and India.
Using an amended human capital (AHC) and a COI model, Yin et al. (2015) estimated the value of economic loss of PM10-related health risks in 16 districts and four zones of Beijing, and healthcare cost estimates for cardiovascular and respiratory diseases. For the period 2008–2012, economic losses related to health damage of PM10 pollution increased from $23–$31 billion dollars. Using a similar methodology, but for estimates of PM2.5 in 190 Chinese cities, Yang et al. (2018) found that the average economic loss was 0.3% (AHC) to 1% (VSL) of the total GDP for 190 cities from 2014 to 2016. Maji et al. (2018) assessed the PM2.5-related health and economic loss for 338 Chinese cities and found that PM2.5 exposure caused an economic loss of $101 billion, which was 0.91% of the national GDP in 2016. In the total economic loss, premature deaths accounted for $95 billion (CI 95%; 43.88, 133.18), approximately 94% of the total loss.
Ortiz et al. (2011) conducted a contingent valuation survey in Sao Paolo, Brazil, to estimate the WTP to avoid one hospital admission and one emergency room visit due to respiratory diseases in adults and children younger than 5 years old, and cardiovascular diseases in adults only, both associated with ambient air pollution. The annual mean WTP estimates were $106.37 (adult) and $179.30 (child) for hospital admission; $62.92 (adult) and $117.86 (child) for emergency room visits due to respiratory diseases and $117.10 (hospital admissions) and $69.64 (emergency room visits) for cardiovascular diseases. In Chile, the WTP to avoid emergency rooms visits due to respiratory illness and improved air quality ranged from $2,800 to $13,000 (Rizzi et al., 2014).
Uncertainties on Health and Costs Analyses: What Are The Main Issues?
Some of the issues related to uncertainties with the estimation of the impact of ambient air pollution on health and costs have already been discussed in this article. A major source of uncertainty in estimates of health effects in some countries and regions regards population exposure levels to ambient air pollution due to lack of data from ground monitoring stations. And a major challenge in estimating the economic cost of health effects relates particularly to the issues of income elasticities used for estimating VSL and thus the cost of mortality.
Ambient air pollution levels vary seasonally, daily, by time of day, and by the time of exposure. In short-term evaluations, time-series studies have been used to assess how adverse health events (e.g., deaths, hospital admissions, etc.) and daily variation in ambient air pollution concentrations co-vary overtime. In such short-term studies, risk factors, sociodemographic factors, or smoking status, for example, are not considered confounders because they do not co-vary with pollution over relatively short periods when averaged over large populations (Dominici et al., 2003). However, over years, socioeconomic and demographic factors, as well as other risk factors (such as diet and smoking) can play an important role in long-term exposure to air pollution.
These issues have been explored in cohort studies (Jerrett et al., 2010; Krewski et al., 2009) and in studies that include variation in exposure generated at several spatial levels (Miller et al., 2007; Puett et al., 2008). It is well known that long-term exposure to ambient air pollution is a risk factor for several diseases and mortality (WHO, 2016). However, mortality and air pollution levels vary in space, and other mortality risk factors, such as sociodemographics, should be collected and related simultaneously with mortality trends and account for cofounding in the models for more reliable and robust estimates of health and economic impacts. Sheppard et al. (2012) stated that spatial patterns in mortality can persist even after adjusting for individual risk factors, contextual risk factors, and air pollution. The authors suggested that, in assessing spatial mortality patterns, multiple levels of location of each individual subject should be assessed, such as the community and neighborhood in which the individual lives, as well as the spatial dependencies of these locations. More research is required to define the impact of long-term exposure to air pollution on mortality and, consequently, its effect on economic costs.
Discussion and Conclusions
This article has described the steps used to quantify main health effects and costs of ambient air pollution, focusing on fine particulate matter (PM2.5), the air pollutant that caused 95% of deaths from air pollution in 2019 according to the GBD 2019. It has described how research has evolved to provide increasingly comprehensive estimates of the health effects of air pollution, as evidenced in the methodological improvements introduced first by Pope et al. (2002, 2009, 2011) and subsequently in research conducted by the GBD project. However, the wide variation in estimates of pollution population exposure and its health effects in the GBD 2010, 2013, 2015, 2016, 2017, and 2019 raises questions about the precision and reliability of such estimates. The main source of uncertainty for these estimates is the scarcity of ground-level monitoring data in many countries, particularly low- and middle-income countries.
VSL is the main approach used to estimate the economic cost of air pollution–related mortality. Unfortunately, there are only a few low- and middle-income countries in which methodologically rigorous research has been conducted to estimate VSL. In countries where such studies do not exist, the general approach has been to transfer the results of studies conducted in other countries, adjusting for differences in income levels. However, there is an ongoing debate on the income elasticities that should be used when transferring VSLs across countries. There is a need for further research to better assess key differences between high-income countries and low- and middle-income countries.
The health effects and costs of ambient air pollution, particularly from PM2.5, are staggering. The World Bank estimates that ambient PM2.5 air pollution had a cost of $6.4 trillion in 2019, equivalent to 4.8% of GDP that year (PPP adjusted; World Bank, 2021). The impacts of air pollution are most severe in middle-income countries, both in terms of the number of deaths from PM2.5 and percentage of total deaths, and in terms of the cost of PM2.5 pollution.
Developing comprehensive air quality monitoring networks in countries that do not have them, as well as robust methods to estimate the economic impact of morbidity and mortality of ambient air pollution, is urgently needed to address existing uncertainties and build an even stronger case for tackling ambient air pollution as a development priority. Without decisive action, ambient air pollution will likely continue to increase and claim millions of lives every year.
Funding: This research was funded by the World Bank's Korea Green Growth Trust Fund and the Pollution Management and Environmental Health (PMEH) Program.
We are grateful for the insightful comments provided by Raul Figueroa Diaz, María del Pilar García Velázquez, Francisco Guillén Martin, Raúl Dávila Pérez, César Cabrera Cedillo y José Federico González Medrano from Mexico’s National Institute of Statistics and Geography (INEGI).We also gratefully acknowledge the helpful comments provided by Yewande Awe, Marcelo Mena, Dan Biller, and Katharina Siegmann.
- Alberini, A., & Krupnick, A. (2000). Cost-of-illness and willingness-to-pay estimates of the benefits of improved air quality: Evidence from Taiwan. Land Economics, 76(1), 37–53.
- Alvarado, J. A., McVey A. E., Hegarty, J. D., Cross, E. B., Hasenkopf C. A., Lynch, R., Kennelly, E., Onasch, T., Awe, Y., Sanchez-Triana, E., & Kleiman, G. (2019). Evaluating the use of satellite observations to supplement ground-level air quality data in selected cities in low- and middle-income countries. Atmospheric Environment, 218, 1–14.
- Andersson, H., & Treich, N. (2009). The value of a statistical life. Swedish National Road and Transport Research Institute.
- Ara, S., & Tekesin, C. (2017). The monetary valuation of lifetime health improvement and life expectancy gains in Turkey. International Journal of Environmental Research and Public Health, 14, Article 1151.
- Bosworth, R. C., Hunter, A., & Kibira, A. (2017). The value of a statistical life: Economics and politics. Strata.
- Bowe, B., Xie, Y., Li, T., Hong Xian, Y.-Y., & Al-Aly, Z. (2018). The 2016 global and national burden of diabetes mellitus attributable to PM2.5 air pollution. Lancet Planetary Health, 2(7), E301–E312.
- Brauer, M., Amann, M., Burnett, R. T., Cohen, A., Dentener, F., Ezzati, M., Henderson, S. B., Krzyzanowski, M., Martin, R. V., Van Dingenen, R., van Donkelaar, A., & Thurston, G. D. (2012). Exposure assessment for estimation of the global burden of disease attributable to outdoor air pollution. Environmental Science and Technology, 46(2), 652–660.
- Brauer, M., Freedman, G., Frostad, J., van Donkelaar, A., Martin, RV., Dentener, F., Dingenen, RV., Estep, K., Amini, H., Apte, JS., & Balakrishnan, K. (2016). Ambient air pollution exposure estimation for the Global Burden of Disease 2013. Environmental Science and Technology, 50(1), 79–88.
- Burnett, R., Chen, H., Szyszkowicz, M., Fann, N., Hubbell, B., Pope, C. A., III, Apte, J., Brauer, M., Cohen, A., Weichenthali, S., Coggins, J., Di, Q., Brunekreef, B., Frostad, J., Lim, S., Kan, H., Walker, K., Thurston, G., Hayes, R., Lim, C., . . ., Spadaro, J. (2018). Global estimates of mortality associated with long-term exposure to outdoor fine particulate matter. Proceedings of the National Academy of Sciences, 115(38), 9592–9597.
- Burnett, R. T., Pope, C. A., III, Ezzati, M., Olives, C., Lim, S. S., Mehta, S., Shin, H. H., Singh, G., Hubbell, B., Brauer, M., Anderson, H., Smith, K., Balmes, J., Bruce, N., Kan, H., Laden, F., Prüss-Ustün, A., Turner, M., Gapstur, S., Diver, W. R., & Cohen, A. (2014). An integrated risk function for estimating the global burden of disease attributable to ambient fine particulate matter exposure. Environmental Health Perspectives, 122, 397–403.
- Diamond, P. A., & Hausman, J. A. (1994). Contingent valuation: Is some number better than no number? Journal of Economic Perspectives, 8(4), 45–64.
- Dixon, J., Fallon Scura, L., Carpenter, R. A., & Sherman, P. B. (1994). Economic analysis of environmental impacts (2nd ed.). Earthscan.
- Dominici, F., Daniels, M., McDermott, A., Zeger, S. L., & Samet, J. (2003). Mortality among residents of 90 cities. In: Health Effects Institute (2003). Revised Analyses of Time-Series Studies of Air Pollution and Health, pp. 9-24. Special Report. Health Effects Institute, Boston MA.
- Gao, M., Guttikunda, S. K., Carmichael, G. R., Wang, Y., Liu, Z., Stanier, C. O., Saide, P. E., & Yu, M. (2015). Health impacts and economic losses assessment of the 2013 severe haze event in Beijing area. Science of the Total Environment, 511, 553–561.
- GBD 2019 Diseases and Injuries Collaborators. (2020). Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: A systematic analysis for the Global Burden of Disease Study 2019. Lancet, 396, 1204–1222.
- GBD 2015 Risk Factors Collaborators. (2016). Global, regional, and national comparative risk assessment of 79 behavioural, environmental and occupational, and metabolic risks or clusters of risks, 1990–2015: A systematic analysis for the Global Burden of Disease Study 2015. Lancet, 388, 1659–1724.
- GBD 2016 Risk Factors Collaborators. (2017). Global, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks, 1990–2016: A systematic analysis for the Global Burden of Disease Study 2016. Lancet, 390, 1345–1422.
- GBD 2017 Risk Factors Collaborators. (2018). Global, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks for 195 countries and territories, 1990–2017: A systematic analysis for the Global Burden of Disease Study 2017. Lancet, 392, 1923–1994.
- GBD 2019 Risk Factors Collaborators. (2020). Global burden of 87 risk factors in 204 countries and territories, 1990–2019: A systematic analysis for the Global Burden of Disease Study 2019. Lancet, 396, 1223–1249.
- Gentry, E. P., & Viscusi, W. K. (2016). The fatality and morbidity components of the value of statistical life. Journal of Health Economics, 46, 90–99.
- Hammitt, J. K., & Graham, J. D. (1999). Willingness to pay for health protection: Inadequate sensitivity to probability? Journal of Risk and Uncertainty, 18, 33–62.
- Hammitt, J. K., & Robinson, L. A. (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), 1–29.
- Hausman, J. (2012). Contingent valuation: From dubious to hopeless. Journal of Economic Perspectives, 26, 43–56.
- Health Effects Institute. (2020). State of global air 2020: Data from Global Burden of Disease Study 2019. Institute for Health Metrics and Evaluation (IHME).
- Institute for Health Metrics and Evaluation. (2020). Global burden of disease results tool.
- Istamto, T., Houthuijs, D., & Lebret, E. (2014). Willingness to pay to avoid health risks from road-traffic-related air pollution and noise across five countries. Science of the Total Environment, 497–498, 420–429.
- Jerrett, M., Gale, S., & Kontgis, C. (2010). Spatial modeling in environmental and public health research. International Journal of Environmental Research and Public Health, 7, 1302–1329.
- Kan, H., & Chen, B. (2004). Particulate air pollution in urban areas of Shanghai, China: Health-based economic assessment. Science of the Total Environment, 322, 71–79.
- Kniesner, T. J., & Viscusi, W. K. (2019). The value of a statistical life. Legal Studies Research Paper Series. Working Paper No. 19-15. Vanderbilt University Law School.
- Krewski, D., Jerrett, M., Burnett, R. T., Ma, R,, Hughes, E., Shi, Y., Turner, M. C., Pope, C. A., III., Thurston, G., Calle, E. E., Thun, M. J., Beckerman, B., DeLuca, P., Finkelstein, N., Ito, K., Moore, D. K., Newbold, K. B., Ramsay, T., Ross, Z., Shin, H., & Tempalski, B. (2009). Extended follow-up and spatial analysis of the American Cancer Society study linking particulate air pollution and mortality. Research Report Health Effects Institute (140):5–114; discussion 115–136. PMID: 19627030.
- Lim, S. S., Vos, T., Flaxman, A. D., Danaei, G., Shibuya, K., Adair-Rohani, H., Amann, M., Anderson, H., Andrews, K., Aryee, M., Atkinson, C., Bacchus, L., Bahalim, A., Balakrishnan, K., Balmes, J., Barker-Collo, S., Baxter, A., Bell, M., Blore, J., Blyth, F., . . . , Ezzati, M. (2012). A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990–2010: A systematic analysis for the Global Burden of Disease Study 2010. Lancet, 380, 2224–2260.
- Lindhjem, H., Navrud, S., Braathen, N. A., & Biausque, V. (2011). Valuing mortality risk reductions from environmental, transport, and health policies: A global meta-analysis of stated preferences studies. Risk Analysis, 31(9), 1381–1407.
- Maji, K. J., Ye, W.-F., Arora, M., & Nagendra, S. M. S. (2018). PM2.5-related health and economic loss assessment for 338 Chinese cities. Environment International, 121, 392–403.
- Masterman, C. J., & Viscusi, W. K. (2018). The income elasticity of global values of a statistical life: Stated preference evidence. Journal of Benefit-Cost Analysis, 9(3), 407–434.
- Mehta, S., Shin, H., Burnett, R., North, T., & Cohen, A. (2013). Ambient particulate air pollution and acute lower respiratory infections: A systematic review and implications for estimating the global burden of disease. Air Quality, Atmosphere, & Health, 6, 69–83.
- Miller, K. A., Siscovick, D. S., Sheppard, L., et al. (2007). Long-term exposure to air pollution and incidence of cardiovascular events in women. New England Journal of Medicine, 356, 447–458.
- Organisation for Economic Co-operation and Development. (2012). Mortality risk valuation in environment, health, and transport policies.
- Organisation for Economic Co-operation and Development. (2016). The economic consequences of outdoor air pollution.
- Ortiz, R. A., Hunt, A., Motta, A. S., & MacKnight, V. (2011). Morbidity costs associated with ambient air pollution exposure in Sao Paulo, Brazil. Atmospheric Pollution Research, 2, 520–529.
- Ostro, B. (2004). Outdoor air pollution: Assessing the environmental burden of disease at national and local levels. Environmental Burden of Disease 5. World Health Organization.
- Ostro, B., Spadaro, J. V., Gumy, S., Mudu, P., Awe, Y., Forastiere, F., & Peters, A. (2018). Assessing the recent estimates for the global burden of disease of ambient air pollution: Methodological changes and implications for low- and middle-income countries. Environmental Research, 166, 713–725.
- Pope, C. A., III, Burnett, R. T., Krewski, D., et al. (2009). Cardiovascular mortality and exposure to airborne fine particulate matter and cigarette smoke: Shape of the exposure-response relationship. Circulation, 120, 941–948.
- Pope, C. A., III, Burnett, R. T., Thun, M. J., Calle, E., Krewski, D., Ito, K., & Thurston, G. (2002). Lung cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution. Journal of the American Medical Association, 287, 1132–1141.
- Pope, C. A., III, Burnett, R. T., Turner, M., et al. (2011). Lung cancer and cardiovascular disease mortality associated with ambient air pollution and cigarette smoke: Shape of the exposure-response relationships. Environmental Health Perspectives, 119(11), 1616–1621.
- Pope, C. A., III, & Dockery, D. W. (2006). Health effects of fine particulate air pollution: Lines that connect. Journal of the Air and Waste Management Association, 56, 709–742.
- Puett, R. C., Schwartz, J., Hart, J. E., Yanosky, J. D., Speizer, F. E., Suh, H., Paciorek, C. J., Neas, L. M., & Laden, F. (2008). Chronic particulate exposure, mortality, and coronary heart disease in the Nurses’ Health Study. American Journal of Epidemiology, 168, 1161–1168.
- Rizzi, L. I., De La Maza, L., Cifuentes, L. A., & Gomez, J. (2014). Valuing air quality impacts using stated choice analysis: Trading off visibility against morbidity effects. Journal of Environmental Management, 146, 470–480.
- Selin, N. E., Wu, S., Nam, K. M., Reilly, J. M., Paltsev, S., Prinn, R. G., & Webster, M. D. (2009). Global health and economic impacts of future ozone pollution. Environmental Research Letters, 4(4), 044014.
- Shaddick, G., Thomas, M., Jobling, A., Brauer, M., van Donkelaar, A., Burnett, R., Chang, H., Cohen, A., Van Dingenen, R., Dora, C., Gumy, S., Liu, Y., Martin, R., Waller, L., West, J., Zidek, J., & Prüss-Ustün, A. (2018). Data integration model for air quality: A hierarchical approach to the global estimation of exposures to ambient air pollution. Applied Statistics, 67(1), 231–253.
- Sheppard, L., Burnett, R. T., Szpiro, A. A., Kim, S. Y., Jerrett, M., Pope, C. A., & Brunekreef, B. (2012). Confounding and exposure measurement error in air pollution epidemiology. Air Quality, Atmosphere, & Health, 5(2), 203–216.
- Shin, H., Cohen, A., Pope, C. A., III, Ezzati, M., Lim, S., Hubbell, B., & Burnett, R. (2013). Critical issues in combining disparate sources of information to estimate the global burden of disease attributable to ambient fine particulate matter exposure (Working Paper). Harvard Center for Risk Analysis.
- Smith, K. R., Bruce, N., Balakrishnan, K., Adair-Rohani, H., Balmes, J., Chafe, Z., Dherani, M., Hosgood, D. H., Mehta, S., Pope, D., Rehfuess, E.; others in the HAP CRA Risk Expert Group. (2014). Millions dead: How do we know and what does it mean? Methods used in the comparative risk assessment of household air pollution. Annual Review of Public Health, 35, 185–206.
- Van Donkelaar, A., Martin, R., Brauer, M., & Boys, B. (2015). Use of satellite observations for long-term exposure assessment of global concentrations of fine particulate matter. Environmental Health Perspectives, 123, 135–143.
- Van Donkelaar, A., Martin, R. V., Brauer, M., Hsu, N. C., Kahn, R. A., Levy, R. C., Lyapustin, A., Sayer, A. M., & Winker, D. M. (2016). Global estimates of fine particulate matter using a combined geophysical-statistical method with information from satellites, models, and monitors. Environmental Science and Technology, 50, 3762–3772.
- Viscusi, W. K. (1993). The value of risks to life and health. Journal of Economic Literature, 31(4), 1912–1946.
- Viscusi, W. K., & Masterman, C. J. (2017). Income elasticities and global values of a statistical life. Journal of Benefit Cost Analysis, 8(2), 226–250.
- World Bank. (2016). Methodology for valuing the health impacts of air pollution: Discussion of challenges and proposed solutions.
- World Bank. (2020). World Development Indicators 2020.
- World Bank. (forthcoming). The global health cost of PM2.5 air pollution in 2019. The World Bank.
- World Bank & Institute for Health Metrics and Evaluation. (2016). The cost of air pollution: Strengthening the economic case for action.
- World Health Organization. (2004). Comparative quantification of health risks: Global and regional burden of disease attributable to selected major risk factors.
- World Health Organization. (2006). WHO air quality guidelines for particulate matter, ozone, nitrogen dioxide and sulfur dioxide: Global update 2005—Summary of risk assessment.
- World Health Organization. (2009). Estimated deaths and DALYs attributable to selected environmental risk factors, by WHO member states, 2004.
- World Health Organization. (2013). Review of evidence of health aspects of air pollution: REVIHAAP Project [Technical report]. WHO Regional Office for Europe.
- World Health Organization. (2015). WHO expert consultation: Available evidence for the future update of the WHO Global Air Quality Guidelines [Report].
- World Health Organization. (2016). Ambient air pollution: A global assessment of exposure and burden of disease.
- World Health Organization. (2017). Preventing noncommunicable diseases (NCDs) by reducing environmental risk factors.
- World Health Organization. (2018). 9 out of 10 people worldwide breathe polluted air, but more countries are taking action.
- World Health Organization & Organisation for Economic Co-operation and Development. (2015). Economic cost of the health impact of air pollution in Europe: Clean air, health and wealth. WHO Regional Office for Europe.
- Yang, Y., Luo, L., Song, C., Yin, H., & Yang, J. (2018). Spatiotemporal assessment of PM2.5-related economic losses from health impacts during 2014–2016 in China. International Journal of Environmental Research and Public Health, 15, 1278.
- Yin, H., Xu, L., & Cai, Y. (2015). Monetary valuation of PM10-related health risks in Beijing, China: The necessity for PM10 pollution indemnity. International Journal of Environmental Reseach and Public Health, 12, 9967–9987.
- Yoo, S.-H., Kwak, S.-J., & Lee, J.-S. (2008). Using a choice experiment to measure the environmental costs of air pollution impacts in Seoul. Journal of Environmental Management, 86(1), 308–318.