Chapter 4. Measuring Pollution Damage from Fuel Use
  • 1 0000000404811396 Monetary Fund
  • | 2 0000000404811396 Monetary Fund
  • | 3 0000000404811396 Monetary Fund
  • | 4 0000000404811396 Monetary Fund


This chapter begins with a brief review of the literature on valuing climate change damage from carbon dioxide (CO2) emissions. The heart of the chapter is about measuring damage from the most important harm from local air pollution: human mortality risk.

This chapter begins with a brief review of the literature on valuing climate change damage from carbon dioxide (CO2) emissions. The heart of the chapter is about measuring damage from the most important harm from local air pollution: human mortality risk.

CO2 Damage

The future climate change damage from a ton of CO2 emissions is the same regardless of the fuel combustion process or where emissions are released. In principle, therefore, each ton should be priced the same in different countries. If charges are imposed on fuel suppliers, the appropriate charge per unit of fuel is CO2 damage times the CO2 emissions factor (i.e., CO2 emissions released per unit of fuel combustion). The first component is discussed here, and emissions factors are discussed in the second section.

Two economic approaches have been used to assess appropriate CO2 emissions prices—the benefit-cost and cost-effectiveness approaches.1

Benefit-Cost Approach

The benefit-cost approach assesses damage from future global climate change caused by additional emissions using “integrated assessment models” that incorporate the following:

  • Links between current emissions and the future global time-path of atmospheric greenhouse gas (GHG) concentrations

  • Impacts of changes in that time-path on global temperature and other climate variables in future years

  • Worldwide monetized damage from those climate changes (e.g., agricultural impacts, costs of sea-level protection, health impacts from altered climate and possible spread of vector-borne diseases, ecological impacts)

  • Discounting of damage at different future dates to the present, to obtain a single summary statistic, or damage per ton of CO2, known as the “social cost of carbon” (SCC).

Although many uncertainties surround these relationships, damage values are especially sensitive to discounting (CO2 emissions have very long range impacts because they reside in the atmosphere for many decades and the climate adjusts gradually to higher atmospheric concentrations) and the treatment of extreme risks.

One view on discounting is that the future benefits of emissions mitigation policies should be discounted using market interest rates (usually about 3–5 percent in advanced countries) because this is the standard way to evaluate the future benefits of any private and many public investments. Studies using market discount rates typically estimate the SCC to be about $10/ton to $50/ton of CO2; for example, US IAWG (2013) puts the SCC at $35/ton (in their central case for 2010 in 2010 dollars).

Others argue that for ethical reasons, below market rates should be used to evaluate policies if the benefits accrue to future generations (as opposed to the current generation), to avoid discriminating against people who are not yet born. Under this approach, SCC estimates are much higher, for example, Stern (2007) puts the SCC at $85 per ton (in 2004 dollars), with the difference compared with earlier studies largely reflecting different discount rates (Nordhaus, 2007).2

These SCC estimates often include a component for catastrophic risks by postulating probabilities—based on judgment, given that the risks are unknown—that future climate change may result in very large world GDP losses. However, the appropriate way to treat these risks remains very contentious (Pindyck, 2013): some studies (e.g., Weitzman, 2009) suggest they warrant dramatically higher CO2 prices.3

Typically, SCC estimates in the benefit-cost approach increase about 1.5–2.5 percent a year in real terms, primarily reflecting the growth rate in output potentially affected by climate change.

Cost-Effectiveness Approach

Rather than explicitly valuing environmental damage (the approach taken elsewhere in this volume), the cost-effectiveness approach assesses least-cost pricing paths for CO2 emissions that are broadly consistent with long-term climate stabilization goals.

Numerous climate-economy models have been developed, with particular detail on the global energy sector and links between emissions, atmospheric GHG concentrations, and future climate outcomes. Projecting future emissions prices needed to meet long-range climate targets is inherently imprecise, however, given considerable uncertainty about future emissions baselines (which depend on future population, per capita income, the energy-intensity of GDP, the fuel mix, and so on) and the emissions impact of pricing (which depends on the future costs of low-emission fuels and technologies and other factors).

A global CO2 price starting at about $30/ton (in current dollars) in 2020 and rising about 5 percent a year would be roughly in line with ultimately containing mean projected warming to 2.5°C (Nordhaus, 2013, p. 228) at least cost. A substantially higher global price would be needed to contain mean projected warming to 2°C (the goal identified in the 2009 Copenhagen Accord), though this target might now be beyond reach because it would likely require the future development and global deployment of technologies that, on net, remove GHGs from the atmosphere, to help lower future GHG concentrations to current levels. For any given climate stabilization target, substantially higher starting prices will also be needed in the absence of full participation by all the major emitting countries.

Illustrative Value Used Here

Tremendous uncertainty and controversy surround the appropriate CO2 emissions price—and country governments may have their own perspectives. A value of $35/ton (based on US IAWG, 2013) is used here for illustrative purposes. This chosen value should not be construed as a recommendation for one SCC value over another or one climate stabilization target over another. The implications of alternative values for corrective taxes are easily inferred by proportionally scaling the component for carbon damages.

A Note on Equity

A key principle in the UN Framework Convention on Climate Change is that developing countries have “common but differentiated responsibilities,” meaning (given their relatively low income and small contribution to historical GHG accumulations) that they should bear a disproportionately lower burden of mitigation costs than wealthier nations. This principle implies either their receiving compensation or their imposing lower emissions prices than others, or perhaps no price at all. Application of this principle need not hinder international mitigation efforts however, at least for the vast majority of low-income countries whose emissions constitute a tiny fraction of the global total (Gillingham and Keen, 2012).

Local Air Pollution Damage

Although local air pollution causes a variety of other harmful environmental effects, the central issue is premature human mortality, which is, by far, the most important category in previous damage assessments (Chapter 2).

The pollution-mortality impacts from fuel combustion can be valued using the following steps:

  • Determining how much pollution is inhaled by exposed populations, both in the country where emissions are released and, for emissions released from tall smokestacks, in countries to which pollution may be transported

  • Assessing how this pollution exposure affects mortality risks, accounting for factors, such as the age and health of the population, that affect vulnerability to pollution-related illness

  • Monetizing the health effects

  • Expressing the resulting damage per unit of fuels.

The focus here is damage from an incremental amount of pollution rather than damage from the total amount of pollution because the incremental amount is relevant for setting efficient fuel taxes.

For a very limited number of countries, previous studies have estimated local air pollution damage, and major modeling efforts are ongoing at the global level.4 This volume is the first attempt to provide an assessment of fossil fuel emissions damage across a broad range of developed and developing countries, using a consistent methodology.5

Although insofar as possible key country-specific factors determining environmental damage is captured, not all potentially significant factors, most notably cross-country differences in meteorological conditions affecting pollution formation, can feasibly be included. The corrective tax estimates in this chapter may also become outdated as evidence and data evolve. Nonetheless, some broad sense of how missing factors may affect the results is given by comparing the results for selected countries to those from a computational model of regional air quality. And accompanying spreadsheets,6 indicating corrective taxes by fuel product and country, are easy to update.

The discussion proceeds as follows: The first three subsections address, respectively, the first three steps in the bulleted list above. The fourth subsection summarizes the resulting cross-country estimates of local pollution damage. The fifth subsection compares the results with those from the computational model. The final subsection discusses procedures for converting emissions damage into corrective fuel taxes, the results of which are presented in Chapter 6.

Estimating Population Exposure to Pollution

As noted in Chapter 2, the main cause of mortality risk from pollution is particulate matter with diameter up to 2.5 micrometers (PM2.5), which is small enough to permeate the lungs and bloodstream. PM2.5 can be emitted directly as a primary pollutant from fuel combustion, but is also produced as a secondary pollutant from chemical reactions in the atmosphere involving primary pollutants, the most important of which is sulfur dioxide (SO2), but also nitrogen oxides (NOx).

“Intake fractions” are used to estimate how much pollution from stationary and mobile emissions sources in different countries is inhaled by exposed populations (see Box 4.1 for technical details). Specifically, these fractions, as used here, indicate grams of PM2.5 inhaled per ton of primary PM2.5, SO2, and NOx. Intake fractions are a powerful concept and are being used increasingly in pollution damage assessment (Apte and others, 2012; Bennett and others, 2002; Cropper and others, 2012; Humbert and others, 2011; Levy, Wolff, and Evans, 2002; Zhou and others, 2006), primarily because they circumvent the need to develop data-and computationally intensive air quality models.

Intake fractions depend on three main factors:

  • The height at which emissions are released: The most important distinction is between emissions from tall smokestacks, such as at power plants, which are more likely to be dispersed without harm but are also transported considerable distances, and emissions released at ground level, such as from cars and residential heating, which tend to stay locally concentrated.

  • The size of the population exposed to the pollution: For smokestack emissions, people living 2,000 kilometers or more from a plant can still intake some of the pollution (Zhou and others, 2006). Even if a plant were to be located away from an urban center, its emissions could still cause significant health damage elsewhere. Long-distance transportation of pollution also raises thorny issues about how one country should account for cross-border environmental damage when setting its own fuel taxes.

Intake Fractions: Some Technicalities

The intake fraction (iF) for a primary pollutant at a particular location is given by the formula (Levy, Wolff, and Evans, 2002):


in which Pi is the population residing in a region, indexed by i, defined by its distance from the emissions source; the region could be in the country where emissions are released or in some other country or some combination of both. The term ΔCi, is the change in the ambient concentration of pollution (PM2.5), perhaps defined by the daily average change in pollution per cubic meter, caused by emissions from the source; ΔCi is influenced by meteorology and other factors. BR is the average breathing rate, that is, the rate at which a given amount of ambient pollution is inhaled by the average person; for example, in Zhou and others (2006) the breathing rate is 20 cubic meters per day.

The numerator in the equation is, therefore, the total daily amount of pollution taken in by potentially exposed populations. In the denominator, Q is the emission rate of the primary pollutant in tons per day. The intake fraction is defined as average pollution inhaled per unit of emissions released, and is usually expressed as grams of PM2.5 inhaled per ton of primary emissions.

  • Meteorological conditions (most notably wind speed and direction), topography (e.g., proximity to mountain barriers that may block pollution dispersion), and ambient ammonia concentrations (which catalyze atmospheric reactions of SO2 and NOx).

For long-distance pollution, a strength of the approach used here is that it uses highly disaggregated data on population density (in different countries) up to 2,000 kilometers away from emissions sources. Therefore, the estimates of population exposure may be considerably more accurate than in other studies using much more spatially aggregated population data, or that only consider people living within shorter distances of the emissions source.

A weakness is that the intake fraction approach cannot easily account for cross-country differences in meteorological and related conditions, not least because emissions are transported across multiple climate zones and wind patterns. However, studies suggest that population exposure is usually, by far, the more important factor (Zhou and others, 2006).

The estimation of population exposure for coal plants, other stationary sources, and mobile emissions sources are discussed in turn below.

Exposure to coal plant emissions

Although intake fractions have been extensively estimated for emissions released at ground level for many different regions (see the discussion of mobile sources below), estimates are much more limited for emissions released from tall smokestacks because of the complexities involved in modeling long-distance pollution transport.

The approach in this chapter uses a widely cited study by Zhou and others (2006), which follows a two-step statistical procedure. Using a sophisticated model of regional air quality, they start by simulating how emissions are transported to different regions, then map the result to data on regional population density, to estimate intake fractions for a variety of primary pollutants from 29 coal plants in China.7 For example, for the average coal plant, they estimate that about 5 grams of PM2.5 ends up being inhaled for each ton of SO2 emitted. Zhou and others (2006) then use statistical techniques to obtain a set of coefficients indicating what fraction of an average plant’s emissions are inhaled by an average person residing within bands of 0–100 kilometers, 100–500 kilometers, 500–1,000 kilometers, and 1,000–3,300 kilometers from the emissions source.

These coefficients can be combined with data on the number of people living within the four distance classifications from the plant to extrapolate intake fractions for a coal plant in any country, without the need for developing a sophisticated model of regional air quality. To keep the calculations tractable, the last distance category is truncated for the analysis in this chapter (without much loss of accuracy) at 2,000 kilometers.8

For extrapolation purposes, the Carbon Monitoring for Action (CARMA) database9 is used to determine the geographical location of about 2,400 coal plants in about 110 different countries for 2009 (these data cover about 75 percent of the total electricity produced by coal power plants worldwide).

LandScan data are used to obtain 2010 population counts by grid cell for each of these 110 countries, as well as for countries without coal plants but where people are still at risk of inhaling cross-border emissions.10 These population data are extremely fine—each grid cell is 1 kilometer square or less.

Mapping these two data sets provides an extremely accurate estimate of the population living at the four distance classifications from each plant. Multiplying populations in these distance categories by the corresponding coefficient from Zhou and others (2006) for a particular pollutant, and then adding over the four distance categories, gives the estimated intake fraction for that pollutant for each coal plant.

Finally, the national average intake fraction for the pollutant is obtained for each country by taking a weighted sum of intake fractions for individual plants in that country, where the weights are each plant’s share in total coal use.11

There are some caveats to the intake fraction approach as applied to long-distance (but not ground-level) pollution. Most notably, adjustments are not made for meteorological or for topographical conditions or local ammonia concentrations. The last factor is relevant because SO2 and NOx form PM2.5 by reacting with ammonia—in fact, when SO2 and NOx are reduced substantially, ammonia is “freed up” for the remaining SO2 and NOx emissions to react with, making them more likely to form fine particulates (PM2.5). To the extent that all these factors vary (across a radius of 2,000 kilometers) for the average coal plant in another country relative to these conditions for the average coal plant in China, these estimates overstate or understate intake fractions for other countries. This issue is discussed further in a later section that reviews results for selected countries and compares them with those from a computational air quality model that does account for meteorological conditions, ammonia concentrations, and related factors.

Second, intake fractions may also depend on the precise height of the smokestack from which pollutants are emitted, with emissions from the tallest smokestacks having the greatest propensity to dissipate before they are inhaled. Again, data are not available on global variation in the height of smokestacks at power plants to allow an adjustment to be made. However, intake fractions do not appear to vary much with differences in smokestack height (Humbert and others, 2011).

Third, mortality risks to people living close to two or more power plants are assumed to be additive (or in other words, the intake fraction for one coal plant is the same, regardless of whether some of the people inhaling its pollution are also exposed to pollution from other plants). For the most part this seems reasonable, except, perhaps, for countries where air pollution is especially severe and people’s ability to inhale pollution starts to become saturated, but even then (see Box 3.3) there may not be much relevance for corrective fuel tax estimates.

Exposure to other stationary source emissions and vehicle emissions

Because of the lack of data—particularly in regard to geographical location—it is not feasible to estimate population exposure to emissions from other uses of coal (e.g., metals smelting). For purposes of calculating the impact of coal tax reform, environmental damage and corrective taxes for these other uses are assumed to be the same as for power plant coal use.

Essentially the same procedure and data sources as outlined above are used to estimate average population exposure to approximately 2,000 natural gas plants in 101 countries. Natural gas produces the same three primary pollutants as coal so the Zhou and others’ (2006) coefficients can be applied again.12

Intake fractions for each primary pollutant tend to be greater for natural gas than for coal because, on average, gas plants are located closer to population centers, but the differences are not large. In fact, the local pollution effects of natural gas combustion are far less severe than for coal because natural gas produces minimal amounts of SO2 and primary PM2.5.

With respect to natural gas use in homes, primarily for space heating, population exposure to outdoor pollution is far more localized, given that emissions are released and stay close to the ground. The same applies for vehicle emissions.

For both residential and vehicle emissions, estimates from Humbert and others (2011) and Apte and others (2012) are combined. Humbert and others (2011) report a global average intake fraction for ground-level sources of SO2, NOx, and primary PM2.5. Apte and others (2012) estimate, but only for primary PM2.5, intake fractions for 3,646 urban centers across the world, accounting for local population density and meteorology.13 The city-level intake fractions for primary PM2.5 (or a simple average of them for countries with more than one city in the data of Apte and others, 2012) are extrapolated to the country level by weighting them by the fraction of the population living in the relevant urban area.14 Intake fractions for SO2 and NOx by country are then derived from Humbert and others’ (2011) estimates, scaling them by the ratio of the intake fraction for PM2.5 for that country to the global average intake fraction for PM2.5 from Apte and others (2012).15

From Pollution Exposure to Mortality Risk

This section discusses the two steps needed to assess how additional pollution exposure increases mortality risk in different countries. The first step is to establish the baseline mortality rate for illnesses potentially aggravated by pollution. The second is to multiply these baseline mortality rates by estimates of the increased likelihood of mortality with extra pollution relative to mortality without extra pollution, and then aggregate over illnesses.

Much of the discussion relies on work by the World Health Organization’s Global Burden of Disease project, which provides the most comprehensive assessment to date of mortality and loss of health from pollution-related and other diseases, injuries, and risk factors for all regions of the world.16

Baseline mortality rates

The increased mortality risk from extra pollution inhaled by a population of given size will depend on the age and health of the population. Seniors, for example, are generally more susceptible to pollution-induced illnesses than younger adults. Health status also matters—someone already suffering from a heart or lung condition that is potentially aggravated by inhaling pollution is more vulnerable than a healthy person. And if people are more likely to die prematurely from other causes (e.g., traffic accidents, non-pollution-related illness), they are, by definition, less likely to live long enough to die from pollution-related illness.

The role of these factors can be summarized by calculating an age-weighted mortality rate for illnesses potentially worsened by pollution. The focus is on the four adult diseases—lung cancer, chronic obstructive pulmonary disease, ischemic heart disease (from reduced blood supply), and stroke—all of whose prevalence is increased when people intake pollution.

Annual mortality rates from these four illnesses were estimated for each country, taking into account the age structure of the population, as follows: Global Burden of Disease data provide mortality rates for the four diseases for 12 different age classifications at the regional level, with the world divided into 21 regions (the Annex Table 4.1.1 lists the countries within each region). These age classes are for people 25 and older (mortality risks for those younger than 25 are assumed to be zero—see below). Age-weighted mortality rates by disease at the country level are then obtained using the share of the country’s population in each age class.17

Figure 4.1 shows the results for the 21 regions. At a global level, the total mortality rate for diseases potentially worsened by pollution is 3.7 deaths per 1,000 people per year. (Most of these deaths, roughly 89 percent on average, would still occur with no pollution.) Eastern Europe has the highest mortality rate, 10.6 deaths per 1,000 people, in part because of the high prevalence of alcohol- and smoking-related illness. The lowest mortality rate is 1.3 deaths per 1,000 people in western sub-Saharan Africa, where people are more prone to die from other causes rather than surviving long enough to suffer pollution-related illness.18

Figure 4.1
Figure 4.1

Baseline Mortality Rates for Illnesses Whose Prevalence Is Aggravated by Pollution, Selected Regions, 2010

Source: Authors’ calculations.Note: Figure shows number of people (ages 25 and older) per 1,000 population dying from diseases whose incidence can be increased by outdoor air pollution. Only a minor portion should be attributed to pollution—the rest would occur anyway, even if there were no pollution.

For all regions, heart disease is the largest source of mortality—at a global level it accounts for almost half of total deaths from the four diseases, with pulmonary disease and stroke accounting for about 20 percent each, and lung cancer about 10 percent. These shares vary somewhat by region—in Eastern Europe, for example, heart disease accounts for 72 percent of total deaths.

Pollution damage estimated in this volume is understated in the sense that premature deaths of those younger than 25, most notably from infant mortality, are excluded. One reason for omitting these deaths is that the valuation of mortality risk for infants is even more unsettled and contentious than that for adults (see Box 4.3).19

Increased mortality from air pollution

A limited number of studies for the United States have estimated the relationship between pollution concentrations and increased mortality for pollution-related diseases—so-called concentration response functions.20 For example, Pope and others (2002) track the health status of a large cohort of adults in 61 U.S. cities over a long period to attribute health outcomes to PM2.5 concentrations as opposed to other factors such as age, gender, income, dietary habits, smoking prevalence. They estimate that each 10 microgram/cubic meter increase in PM2.5 concentrations increases annual mortality risks from all pollution-related illness in the United States by 6.0 percent. Until recently, the concentration-response functions underlying Pope and others (2002) were used in regulatory assessments by the United States Environmental Protection Agency (US EPA). However, based on more recent evidence (Krewski and others, 2009; Lepeule and others, 2012; Industrial Economics Incorporated, 2006), US EPA now assumes a 10 microgram/cubic meter increase in PM2.5 concentrations raises all pollution-related mortality risks by 10.6 percent (Chapter 5 of US EPA, 2011).

An important question is whether these findings—which are based on evidence for the United States, where PM2.5 concentrations vary geographically by about 5–30 micrograms/cubic meter—apply to other regions. The assumptions used here are based on a best statistical fit for each of the four pollution-related illnesses of various model runs for different regions and different types of studies in Burnett and others (2013).21 The resulting coefficients indicate that each 10 microgram/ cubic meter increase in PM2.5 concentrations increases the risk of all pollution-related mortality (averaged worldwide) by 9.8 percent. Although Burnett and others (2013) provide a state-of-the-art review of the limited number of studies, much more research is needed to improve the understanding of the complex relationship between pollution and mortality (especially for chronic illness); in the meantime, the health impacts calculated here should be viewed very cautiously.

A further caveat is that evidence suggests additional pollution exposure may, paradoxically, have significantly weaker impacts on mortality risk in regions where pollution concentrations are already very high, as the human body becomes progressively “saturated” with pollution (Burnett and others, 2013; Good-kind and others, 2012; Health Effects Institute, 2013). In other words, although the concentration response function appears to be approximately linear in pollution concentrations up to some point (an extra microgram/cubic meter of PM2.5 has the same impact on mortality rates regardless of the initial pollution concentration level), eventually it may flatten out, that is, an extra microgram/cubic meter of PM2.5 has a diminishing impact on elevating mortality rates, the higher the initial PM2.5 concentration. However, as discussed in Box 3.3 of Chapter 3, the corrective fuel tax calculations abstract from this complication, on the assumption that implementing efficient taxes would have a large enough impact on emissions to lower pollution concentrations into the region where the concentration response function is approximately linear.22

Valuing Mortality Risks

Health risk valuation is highly controversial. Many people are uncomfortable with the idea of assigning values to the lives saved from policy interventions. Nonetheless, policymakers should still consider methodologies that have been developed for this exact purpose, despite the implication—unpalatable to some—that people with lower incomes are willing to sacrifice a smaller amount of their consumption to reduce health risks than people with much higher incomes.

In reality, people are constantly trading off money and mortality risk in a variety of decisions on a daily basis (e.g., when deciding whether to pay extra for a safer vehicle or to accept a higher-paying but riskier job like cleaning skyscraper windows). Economic studies attempt to measure these trade-offs, and a consistent finding across a broad range of countries is that mortality risk values generally rise with per capita income (OECD, 2012).

Methodological approaches for valuing mortality risks—or more precisely, the value per premature death avoided—are discussed below, along with empirical evidence, and the possible implications for different countries. Although not all governments will endorse this approach, the implications for corrective fuel taxes of alternative risk values are easily inferred from the results and accompanying spreadsheets by appropriate proportionate changes in the local pollution damage.

Methodological approaches

Two distinct approaches are often used to assess people’s “willingness to pay” to reduce mortality risk. A third approach—generally less preferred by economists—based on valuing losses in human capital is discussed in Box 4.2.

The “revealed preference” method uses observed market behavior to assess mortality risk values, most usually by inferring a person’s willingness to accept lower wages in return for a job with lower fatality risk (given other characteristics of jobs and workers). In contrast, the “stated preference” method relies on responses to questionnaires, most usually contingent-valuation studies in which people are asked direct questions about their money and risk trade-offs.

A potential drawback of revealed preference studies based on labor market data is that they focus on relatively healthy, average-age workers and on immediate accidental death in the workplace. Risks from pollution-related mortality—which primarily affects seniors and results from longer-term risk exposure—might be valued somewhat differently.

The Human Capital Approach

The human capital approach to valuing mortality risk does not (unlike willingness-to-pay approaches) measure people’s own valuation of these risks—instead it focuses on measuring productivity losses from premature mortality. Traditionally this approach has been applied to lost years of working-age life, with a person’s annual productivity proxied by market wages or per capita GDP, and productivity losses across future years discounted back to the present.

However, the human capital approach may undervalue the full economic cost of premature mortality in several respects. For example, the value of lost nonwork time (i.e., time in retirement and leisure time while working age) is often excluded. And people’s valuation of pain and suffering before death are also excluded, as is grief to surviving family members. For these reasons, economists generally prefer willingness-to-pay approaches.1

1 For comparison, in World Bank and State Environmental Protection Agency of China (2007), the costs of air and water pollution in China are about twice as high using the willingness-to-pay measures of mortality risk compared with the human capital measure.

Stated preference studies can avoid these problems through choice of a sample that is more representative of the at-risk populations and through questions about specific hazards, such as cancer, posed by air pollution. The main concern with stated preference studies is that they are hypothetical—whether survey respondents would actually behave the way they say they would when confronted with risk-money trade-offs in the marketplace is unclear, leaving open the question of how accurately they describe people’s actual trade-offs.

Both approaches focus on the costs to the individual (and grief to family members) from mortality risk and omit broader costs borne by third parties, such as medical costs. However, these broader costs may be small relative to the value of mortality risks to individuals; for example, when avoided medical costs later in the life cycle (from premature mortality) are subtracted from higher short-term treatment costs, the net medical burden may be relatively modest.

Empirical evidence

The starting value for mortality risk valuation used in this analysis, and its extrapolation to other countries, is based on a widely peer reviewed study by OECD (2012). This extrapolation accounts for differences in per capita income across countries but not, for reasons discussed in Box 4.3, for other factors, such as age.

Determinants Other than Income of Mortality Risk Valuation

OECD (2012) discusses several non-income-related factors that might cause mortality valuation to differ across countries, but in each case concludes that available evidence is not sufficiently conclusive to make adjustments.

With regard to population characteristics, conceivably the average age of the at-risk population matters, but whether, on balance, this has a positive or negative effect on mortality risk valuation is unclear. On the one hand, older individuals should have lower willingness to pay to reduce mortality risk given that they have fewer years of life left. Offsetting this, however, is that they might be wealthier and therefore have higher willingness to pay, compared with younger people, to increase expected longevity by a given amount. Some studies suggest there is little or no net effect of age on people’s valuation of mortality risk, whereas others suggest a modest decrease at older ages (Krupnick, 2007; Chestnut, Rowe, and Breffle, 2004; Alberini and others, 2004; Hammitt, 2007). Two expert panels in the United States have recommended against age-related adjustments to mortality valuation (Cropper and Morgan, 2007; National Research Council, 2008), and the U.S. Environmental Protection Agency has, for now, abandoned analyses with these adjustments.

Even more unsettled is the appropriate value to apply to child mortality because children have not been the subject of revealed and stated preference studies. As noted, child mortality is excluded from the damage estimates in this book.

Evidence on whether healthier populations are willing to pay more to extend longevity than less healthy populations is similarly inconclusive (Krupnick and others, 2000). Unhealthy people may gain less enjoyment from living longer, but if they also gain less enjoyment from consumption, they may be willing to give up more consumption to prolong life. People in different countries may also have different preferences for trade-offs between consumption goods and mortality risks (perhaps because of cultural factors), but again there is no solid evidence on which to base an adjustment. Definitive evidence is also lacking on whether pollution-related risks (e.g., elevated cancer risk) are valued differently from accident risks, such as the risk of immediate death in a car accident.

Starting value for mortality risk reduction. In OECD (2012), the central case recommendation is to value mortality risks in OECD countries as a group at $3 million per life saved, in 2005 U.S. dollars.

This amount (which is updated below) was obtained from an extensive statistical analysis using several hundred stated preference studies applied to environmental, health, and traffic risks in a variety of countries (mostly Canada, China, France, the United Kingdom, and the United States). Stated preference studies were used because they have been conducted in numerous countries, while revealed preference studies have mainly been confined to the United States (which has ample labor market data). Stated preference studies tend to produce lower valuations than revealed preference studies; therefore, pollution damage estimates might be understated here.23

Income adjustment. The value for mortality risk per life for individual countries (denoted Vcountry) is extrapolated from that for the OECD as a whole (denoted VOECD), using the following equation:


In equation (4.1), Icountry and IOECD denote real income per capita in a particular country and for the OECD, respectively. Relative per capita income is appropriately measured using purchasing power parity rather than market exchange rates because purchasing power parity, which takes the local price level into account, more accurately reflects people’s ability to pay out of their income for local products or risk reductions. The income per capita figures are obtained from IMF (2013) and World Bank (2013).

The exponent ε in (4.1) measures how mortality risk values vary with income; specifically, it is the percentage change in the mortality value per 1 percent change in real per capita income. Based on OECD (2012), the illustrative calculations in this analysis assume ε is 0.8.24

The $3 million mortality value for the OECD is updated to 2010 for inflation (using the average consumer price index for the OECD) and real income (using equation (4.1) and the ratio of per capita income in the OECD in 2010 to that in 2005) to give VOECD = $3.7 million. This amount is then extrapolated to other countries, using equation (4.1) and the countries’ relative per capita incomes for 2010.

A tricky issue is how to value mortality risks for people across the border in other countries. To keep the exercise tractable, the same mortality risk value for these people is used as for people in the country in which emissions are released. An alternative, perhaps more appealing, approach would be to use a weighted average mortality risk valuation, whereby each affected country’s risk valuation is weighted by its share of deaths in total deaths caused by the source country’s emissions. If a source country has high per capita income relative to neighboring countries, this approach would imply somewhat lower pollution damage estimates than obtained in this analysis and vice versa for emissions from countries with relatively low income. However, the differences in emissions damage estimated by the two approaches may not be large; for example, if 40 percent of the affected population resides in other countries and mortality risks for these countries are 25 percent lower than the source country for the emissions, emissions damage will be 10 percent lower (a notable though not dramatic amount) compared with the approach taken here.

Implied mortality risk valuations

Figure 4.2 shows the implied mortality risk values for 20 selected countries. Mortality values per death are highest in the United States at $4.9 million and are less than $1 million in India, Indonesia, and Nigeria. To reemphasize, these values are purely illustrative—as shown below, if mortality values in all countries were set at the OECD average, the corrective tax estimates for relatively low-income countries would increase considerably.

Figure 4.2
Figure 4.2

Value of Mortality Risk, Selected Countries, 2010

Source: Authors’ calculations.Note: Figure shows the value assigned to an individual’s premature death caused by pollution.

The illustrated mortality values in this analysis differ quite a bit from values used at various points in different government studies. However, as shown by the examples in Table 4.1, there appears to be no systemic pattern to these differences. The values for the United States, Canada, and Germany used here are much lower than in government studies for these countries, but the converse applies in other cases. At any rate, the purpose is not to pass judgment on government practices but simply to obtain, for illustrative purposes, a consistently estimated set of cross-country mortality risk values.

Table 4.1

Examples of Mortality of Risk Valuations Used in Previous Government Studies

article image
Sources: Government websites and personal communications with government officials.

Air Pollution Damage Estimates

Selected estimates of local air pollution damage per ton of emissions are discussed in this section. (Damage per unit of fuel is discussed in Chapter 6.) Annex Table 4.2.1 provides the full set of estimates by emissions type, emissions source, and country.

Figure 4.3 shows estimated damage per metric ton for SO2 from coal plants for selected countries. The range of the damage estimates is striking.

Figure 4.3
Figure 4.3

Damage from Coal Plant Sulfur Dioxide (SO2) Emissions, Selected Countries, 2010

Source: Authors’ calculations.

The United States, with damage of about $17,000/ton (in 2010 dollars), is an intermediate case. Damage estimates are much higher (about $35,000–$39,000/ton) in Japan, Poland, Korea, and the United Kingdom, and higher still (about $53,000/ ton) in Germany, reflecting much higher population exposure to power plant emissions, which more than outweighs any influence of lower mortality values for these countries.

Conversely, as the result of a combination of lower population exposure and lower mortality risk values, Australia, Brazil, Chile, Kazakhstan, Mexico, and South Africa have dramatically lower damage values (about $1,500–$3,000/ton). For example, premature mortality per ton of emissions in Australia is just 15 percent of that for the United States.

Damage for China is about $22,000/ton. Although the illustrated mortality risk value for China is only 23 percent of that for the United States, this lower risk value is more than offset by an average population exposure to emissions that is six times as high.25

Figure 4.4 illustrates damage per ton of SO2 from coal combustion for all countries. Damage per ton is highest in European countries where both per capita income and population density are relatively high; for countries in North and South America, Asia, and Oceania damage per ton generally takes intermediate values. For Africa, many countries do not use coal, and for those that do, data limitations often preclude damage estimates.

Figure 4.4
Figure 4.4

Damage from Coal Plant Sulfur Dioxide (SO2) Emissions, All Countries, 2010

Source: Authors’ calculations.

On a per ton basis damage from primary coal plant emissions of PM2.5 is about 25 percent larger than damage from SO2. This is a broadly consistent finding across countries, so the relative pattern of damage across countries is similar to that for SO2. Damage for NOx from coal plants also follows the same broadly similar pattern across countries, though in absolute terms damage per ton from NOx is about 20–50 percent lower than for SO2 (mainly because NOx is less prone to forming PM2.5). Damage per ton from NOx emissions from natural gas plants (essentially the only type of local emissions from these plants) is generally similar to the NOx damage from coal plants.

Figure 4.5 shows the damage per ton from ground-level NOx emissions (estimated for vehicles but also applied to home heating). Again, there are significant cross-country differences; for example, estimated damage exceeds $5,000/ton in Germany, Japan, Korea, and the United States but is less than $1,000/ton in India, Indonesia, Nigeria, South Africa, and Thailand. The relative differences are, however, smaller than for power plant emissions. Ground-level emissions tend to remain locally concentrated, therefore the large average distance between cities, such as in the United States, or the coastal location of cities, such as in Australia, reduce population exposure to a much lesser extent than for power plant emissions. Consequently, damage per ton in the United States is much closer to that in Germany, and in Australia is closer to typical European countries, as shown in Figure 4.5 compared with the relative damage for power plant emissions in Figure 4.3.

Figure 4.5
Figure 4.5

Damage from Ground-Level Nitrogen Oxide (NOx) Emissions, Selected Countries, 2010

Source: Authors’ calculations.

Robustness Checks

The assumptions about how pollution exposure affects health are based on state-of-the-art evidence from the Global Burden of Disease project—though this evidence is far from definitive—and Chapter 6 notes how corrective tax estimates vary with alternative values for mortality risk.

This subsection focuses on other issues relevant for tall smokestack emissions that are dispersed over great distances. The first issue is the reasonableness of the air pollution model (based on Zhou and others, 2006) implicitly underlying the intake fractions for China, and from which intake fractions for other countries are extrapolated. The second issue is how and to what extent the failure to capture cross-country differences in meteorology and related factors might bias damage estimates from the intake fraction approach.

These issues are examined by comparing selected results from the intake fraction approach with those from the TM5-Fast Scenario Screening Tool (FASST).26 This tool (described in Annex 4.3) provides a simplified representation of how pollution concentrations in different regions change in response to additional emissions, and links these changes to population exposure and health impacts. The parameters underlying the air quality component of the model are chosen so that it yields predictions consistent with those from a highly sophisticated model of regional air pollution formation developed by the UN Environment Program (UNEP, 2011).

Unlike in the intake fraction approach, cross-country damage estimates from TM5-FASST capture regional differences in meteorology, ammonia concentration, and other factors. However, the estimation of population exposure is averaged over large areas—the world is divided into 51 regions—which understates population exposure if, as seems likely, power plants are located in areas with higher population density than the regional average.27 Insofar as possible, other inputs to TM5-FASST—particularly baseline mortality rates by region and disease, impacts of additional PM2.5 exposure on mortality rates, and the local valuation of mortality risks—are chosen to be consistent with the intake fraction approach, to facilitate a cleaner comparison of results.

With regard to the reasonableness of Zhou and others’ (2006) air pollution model, TM5-FASST estimates SO2 damage per ton for China to be about $12,000, or slightly more than half of the damage estimate from Zhou and others’ (2006) intake fraction approach. Some of this difference reflects, as just noted, differences in population exposure, but some also likely reflects differences in assumptions about the impact of emissions on air quality. Unfortunately, it is not possible to make a definitive judgment about which air quality model is the more realistic.

With regard to the meteorology issue, Figures 4.6 and 4.7 show SO2 damage per ton for selected countries expressed relative to damage per ton for China from the intake fraction approach and TM5-FASST, respectively. To the extent there are differences in relative damages for particular countries between the two figures, this suggests that differences in meteorological factors between that country and China play a potentially significant role. This does not appear to be a major concern for some countries—for example, the two approaches suggest damage per ton for Japan is 62–67 percent higher than for China, and for the United States damage is 22–24 percent lower than for China. But there are some exceptions; for example, relative damage for Israel, Poland, and the United Kingdom from the intake fraction approach is substantially higher than from TM5-FASST,28 and vice versa for Thailand and Turkey. In short, meteorological factors can significantly alter damage estimates in certain cases, though both the sign and scale of these effects are country specific.

Figure 4.6
Figure 4.6

Estimated SO2 Damage Relative to China Using the Intake Fraction Approach, 2010

Source: Authors’ calculations.
Figure 4.7
Figure 4.7

Estimated SO2 DamageRelativeto China Using the TM5-FASST Model, 2010

Source: Authors’ calculations.

Expressing Damage per Unit of Fuel

To assess efficient taxes on fuel use, the damage expressed per ton of emissions needs to be converted into damage per unit of fuel, or per unit of energy, using appropriate emissions factors. These factors relate the amount of emissions, such as SO2, released into the atmosphere to combustion of a particular fuel, such as natural gas, in a particular activity, for example, power generation. The Greenhouse Gas and Air Pollution Interactions and Synergies (GAINS) model, developed by the International Institute for Applied Systems Analysis (IIASA), was used to estimate these factors.29 See Box 4.4 for more details.

Emissions Factors from the GAINS Model

The Greenhouse Gas and Air Pollution Interactions and Synergies (GAINS) model estimates country-specific emissions factors for carbon and local air emissions for different fossil fuels used in different sectors of the economy. The estimates are reported in kilotons of pollutant per petajoule (heat content) of fuel input, though they could be expressed in emissions per unit of weight or volume (by multiplying heat content per unit of weight or volume using the GAINS data). Two calculations are performed.

First, uncontrolled emissions factors (denoted EFU) are calculated from the basic properties of the fuel and combustion processes (Amann and others, 2011; Cofala and Syri, 1998a, 1998b; Klimont and others, 2002). For example, as defined for SO2 emissions, the emissions factor is calculated by:


In equation (4.4.1) sc is the sulfur content per unit of weight, hv is the heat value per unit of weight, and sr is the sulfur retention fraction (the portion of sulfur that is retained in ash rather than released into the atmosphere).

Second, various controlled emissions factors (denoted EFC) are calculated, where applicable, for emissions sources employing an abatement technology denoted t (e.g., a particular type of scrubber, or hotter boiler that reduces NOx emissions), from the following formula:


In equation (4.4.2) ret represents the fraction of emissions that are abated (that would otherwise be released into the atmosphere) as a result of technology t. Where specific regulations (technology mandates, emission rate standards) exist and are enforced, GAINS calculates the controlled emissions factors based on the regulation. GAINS can be used to calculate three emissions factors—an average over sources with controls (taking into account the potential application rates of alternative control technologies t), an average over sources with and without controls, and an uncontrolled emissions rate.

The GAINS data for these calculations are detailed for some countries; for others, judgment is used to transfer estimates to countries for which data are not directly available.

Coal emissions factors (for CO2, SO2, NOx, and primary PM2.5) are defined relative to energy or heat content in petajoules (PJ) rather than tons of coal, given significant variation in energy content across different types of coal. Where relevant, the factors represent a weighted average across different coal types—in these cases, a more refined pricing system than the one estimated here would vary charges according to the emissions intensity of the particular coal type.

One emissions factor is obtained for carbon, for which opportunities for abating emissions at the point of combustion are presently very limited. For the local air pollutants from coal plants, three different emissions factors are distinguished for each pollutant. First is an uncontrolled emission rate. Second is an emission rate for a representative plant that has some control technology (e.g., an SO2 scrubber). Third is the average emission rate across all existing plants with and without emissions control technologies. These factors are used to estimate three different taxes, for coal plants without and with emissions controls, and for the average coal plant at present, though crediting would provide strong incentives for all plants to use control technologies. In each case, the corrective tax is the product of the emissions factor for a pollutant and the damage per ton for that pollutant, which is then aggregated over all pollutants.

Similar procedures are used to obtain emissions factors and corrective taxes for natural gas. For power plants for which the local air pollution damages are small relative to those from coal, the focus is just on the emissions factor averaged across all plants with and without control technologies, whereas for household use of gas only uncontrolled emissions are relevant. Damage and corrective taxes are again expressed per unit of energy because emissions per unit of volume can vary significantly depending on gas pressure.

For mobile sources, emissions factors for CO2, SO2, NOx, and PM2.5 per liter are obtained for gasoline vehicles and diesel vehicles (the latter representing an average of light- and heavy-duty vehicles using diesel), in each case averaging across vehicles on the road with and without control technologies (fuel taxes by themselves do not encourage the adoption of control technologies).

There are several noteworthy points about the emissions factors:

First, carbon emissions factors for a particular fuel vary little across countries. However, the fuel products themselves vary significantly: per PJ of energy, natural gas, gasoline, and motor diesel generate about 59 percent, 73 percent, and 78 percent, respectively, of the carbon emissions generated by one PJ of coal.

Second, uncontrolled, average, and controlled SO2 emissions factors for coal can vary greatly both within and across countries (see Figure 4.8). For example, on average, the SO2 emissions per PJ for Japanese coal plants with no control technologies is only 30 percent of that for comparable U.S. plants, while in Israel the emission rate is about 70 percent greater than for the United States. Control technologies can dramatically reduce emissions, however; for example, SO2 emission rates at U.S. coal plants with such technologies are 95 percent lower than the rates at plants without these technologies. Primary PM2.5 emission rates for uncontrolled plants follow a pattern similar to those for SO2, but the control technologies have an even more dramatic impact on reducing pollution.

Figure 4.8
Figure 4.8

SO2 Emissions Rates at Coal Plants, 2010

Source: Authors’ calculations.Note: PJ = petajoule. Controlled rate is the average emissions factor for plans with emissions controls. Uncontrolled rate is for plants with no controls. Average rate averages across plants with and without controls. For some countries (e.g., India, South Africa), no sulfur control technologies had been adopted as of 2010. In Germany the average and controlled rates are the same because all plants had some form of control.

Third, NOx emission rates from plants without controls differ from rates for ground-level sources (depending, for example, on combustion temperature, which can affect the amount of nitrogen and oxygen sucked in from the ambient air), but the differences are not large.


An illustrative value for CO2 damage is taken from a recent study, although this value is subject to much debate.

To assess air pollution damage from power plant combustion of coal and natural gas, average population exposure to these emissions (which can be transported long distances) is estimated and combined with data on local mortality rates for pollution-related illness and evidence about how changes in pollution exposure affect these mortality rates. The most controversial step is monetizing these health effects, which, for illustration are inferred from OECD (2012), though the implications of alternative assumptions are transparent from the results. Estimation of pollution from motor vehicle fuels and other ground-level sources are constructed from studies about how much of these emissions are inhaled by people in different urban centers. These pollution damage estimates are then combined with data on local emissions factors for different fuels to derive environmental damage from fuel use (though emission rates vary considerably with the extent of use of control technologies).

Annex 4.1. Regional Classifications for Mortality Rates

Baseline mortality rates are obtained by combining regional average mortality rates for four pollution-related illnesses by age group with country-level data on the age structure of the population. The regional mortality data are from Burnett and others (2013), with the countries grouped into regional classifications as shown in Annex Table 4.1.1.

Annex Table 4.1.1

Country Classifications for Baseline, Pollution-Related Mortality Rates

article image
Source: Burnett and others (2013).

Annex 4.2. Air Pollution Damage by Emissions and Country

Annex Table 4.2.1 summarizes, by country, estimated air pollution damage by the type of emissions, and the source of those emissions.

Annex Table 4.2.1

Damage from Local Air Pollution, All Countries, $/ton of Emissions, 2010

article image
article image
article image
article image
Source: See main text.Note: The table shows estimates of the local pollution health damage per ton from each of three pollutants, according to whether emissions are released from coal combustion, natural gas combustion at power plants, or natural gas and motor fuel consumption at ground level. Black #na = fuel not used; bold #na = data not available.

Annex 4.3. Details on the TM5-FASST Tool

The TM5-FASST tool estimates air pollution damage per ton for different types of emissions in several steps.

First, the baseline mortality rates for four pollution-related illnesses are calculated by region, according to equation (4.2):


RR denotes the risk of premature death from a particular illness relative to that in the baseline case (with current pollution concentration levels). RR–1 is therefore the proportionate change in the relative risk. ΔPM2.5 denotes the change in PM2.5 concentrations relative to the initial situation. α is a parameter that is calibrated separately for each of the four pollution-related diseases to be consistent with the evidence in Burnett and others (2013).

The change in premature deaths for a change in PM2.5 concentrations is given by equation (4.3):


in which mortality rate refers to the baseline rate and population is the exposed population (all those ages 25 and older). Both population data and mortality rate data are from the Institute for Health Metrics and Evaluation (IHME). Deaths are monetized using the same mortality values as discussed in the main text of this chapter.

Next, one ton of SO2 emissions from a particular source is added and processed through an air quality model that links all emissions sources to PM2.5 concentrations in the model’s 51 different regions. The air quality model in the FASST tool is a simplified version of a far more sophisticated air quality model in UN Environment Programme (2011). The change in PM2.5 concentrations in each region is then used to calculate changes in premature deaths based on equations (4.2) and (4.3), and the result is then monetized.

Averaged across the 20 countries considered in this chapter, the damage from SO2 is $17,640/ton. Individual country estimates, relative to those from the intake fraction approach used here, are discussed in the main text of this chapter.

As a check on the above results, simulations were also run with an alternative specification for relative risk given by equation (4.4):


in which parameters α, γ, and δ are calibrated for each of the four diseases to be consistent with Burnett and others (2013). However, the results are only moderately affected. For example, the average damage from SO2 across the 20 countries is 15 percent smaller than when the linear functional form is used.


  • Alberini, Anna, Maureen Cropper, Alan Krupnick, and Nathalie B. 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, Vol. 48, pp. 76992.

    • Search Google Scholar
    • Export Citation
  • Amann, Markus, Imrich Bertok, Jens Borken-Kleefeld, Janusz Cofala, Chris Heyes, Lena Höglund-Isaksson, Zbigniew Klimont, and others, 2011, “Cost-Effective Control of Air Quality and Greenhouse Gases in Europe: Modeling and Policy Applications,Environmental Modelling and Software, Vol. 26, pp. 1489501.

    • Search Google Scholar
    • Export Citation
  • Apte, Joshua S., Emilie Bornbrun, Julian D. Marshall, and William W. Nazaroff, 2012, “Global Intraurban Intake Fractions for Primary Air Pollutants from Vehicles and Other Distributed Sources,Environmental Science & Technology, Vol. 46, pp. 341523.

    • Search Google Scholar
    • Export Citation
  • Bennett, D.H., T.E. McKone, J.S. Evans, W.W. Nazaroff, M.D. Margni, O. Jolliet, and K.R. Smith, 2002, “Defining Intake Fraction,Environmental Science & Technology, Vol. 36, No. 9, pp. 20716.

    • Search Google Scholar
    • Export Citation
  • Bosetti, Valentina, Sergey Paltsev, John Reilly, and Carlo Carraro, 2012, “Emissions Pricing to Stabilize Global Climate,” in Fiscal Policy to Mitigate Climate Change: A Guide for Policymakers, edited by I.W.H. Parry, R. de Mooij, and M. Keen (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Burnett, Richard T., C. Arden Pope, Majid Ezzati, Casey Olives, Stephen S. Lim, Sumi Mehta, Hwashin H. Shin, and others, 2013, “An Integrated Risk Function for Estimating the Global Burden of Disease Attributable to Ambient Fine Particulate Matter Exposure” (Unpublished; Ottawa, Ontario, Canada: Health Canada).

    • Search Google Scholar
    • Export Citation
  • Chestnut, L.G., R.D. Rowe, and W.S. Breffle, 2004, “Economic Valuation of Mortality Risk Reduction: Stated Preference Approach in Canada,Report prepared for Health Canada (Boulder, Colorado: Stratus Consulting Inc).

    • Search Google Scholar
    • Export Citation
  • Cofala, J., and S. Syri, 1998a, “Nitrogen Oxides Emissions, Abatement Technologies and Related Costs for Europe in the RAINS Model Database” (Laxenburg, Austria: International Institute for Applied Systems Analysis).

    • Search Google Scholar
    • Export Citation
  • Cofala, J., and S. Syri, 1998b, “Sulfur Emissions, Abatement Technologies and Related Costs for Europe in the RAINS Model Database” (Laxenburg, Austria: International Institute for Applied Systems Analysis).

    • Search Google Scholar
    • Export Citation
  • Cropper, Maureen, Shama Gamkhar, Kabir Malik, Alex Limonov, and Ian Partridge, 2012, “The Health Effects of Coal Electricity Generation in India,Discussion Paper No. 12–15 (Washington: Resources for the Future).

    • Search Google Scholar
    • Export Citation
  • Cropper, Maureen, and M. Granger Morgan, 2007, “SAB Advisory on EPA’s ‘Issues in Valuing Mortality Risk Reduction,’ “ Memorandum from the Chair, Science Advisory Board, and the Chair, Environmental Economics Advisory Committee, to EPA Administrator Stephen L. Johnson. EPA-SAB-08–001.

    • Search Google Scholar
    • Export Citation
  • European Commission (EC), 2008, ExternE-Externalities of Energy: A Research Project of the European Commission (Brussels: European Commission).

    • Search Google Scholar
    • Export Citation
  • Gillingham, Robert, and Michael Keen, 2012, “Mitigation and Fuel Pricing in Developing Countries,” in Fiscal Policy to Mitigate Climate Change: A Guide for Policymakers, edited by I.W.H. Parry, R. de Mooij and M. Keen (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Goodkind, Andrew L., Jay S. Coggins, Timothy A. Delbridge, Milda Irhamni, Justin Andrew Johnson, Suhyun Jung, Julian Marshall, Bijie Ren, and Martha H. Rogers, 2012, “Prices vs. Quantities With Increasing Marginal Benefits,Discussion paper, Department of Applied Economics, University of Minnesota.

    • Search Google Scholar
    • Export Citation
  • Griffiths, Charles, Elizabeth Kopits, Alex Marten, Chris Moore, Steve Newbold, and Ann Wolverton, 2012, “The Social Cost of Carbon: Valuing Carbon Reductions in Policy Analysis,” in Fiscal Policy to Mitigate Climate Change: A Guide for Policymakers, edited by Ian W.H. Parry, Ruud de Mooij, and Michael Keen (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Hammitt, James K., 2007, “Valuing Changes in Mortality Risk: Lives Saved Versus Life Years Saved,Review of Environmental Economics and Policy, Vol. 1, pp. 22840.

    • Search Google Scholar
    • Export Citation
  • Health Effects Institute, 2013, “Understanding the Health Effects of Ambient Ultrafine Particles,HEI Review Panel on Ultrafine Particles (Boston: Health Effects Institute).

    • Search Google Scholar
    • Export Citation
  • Humbert, Sebastien, Julian D. Marshall, Shanna Shaked, Joseph V. Spadaro, Ryrika Nichioka, Philipp Preiss, Thomas E. McKone, Arpad Horvath, and Olivier Jolliet, 2011, “Intake Fraction for Particulate Matter: Recommendations for Life Cycle Impact Assessment,Environmental Science & Technology, Vol. 45, pp. 480816.

    • Search Google Scholar
    • Export Citation
  • Industrial Economics, Incorporated, 2006, “Expanded Expert Judgment Assessment of the Concentration-Response Relationship between PM2.5 Exposure and Mortality,Final Report (Cambridge, Massachusetts: Industrial Economics, Incorporated).

    • Search Google Scholar
    • Export Citation
  • International Institute for Applied Systems Analysis (IIASA), 2013, “Greenhouse Gas-Air Pollution Interaction and Synergies Model” (Laxenburg, Austria: International Institute for Applied Systems Analysis).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, 2013, World Economic Outlook Database (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Klimont, Zbigniew, Janusz Cofala, Imrich Bertok, Markus Amann, and Chris Heyes, 2002, “Modelling Particulate Emissions in Europe. A Framework to Estimate Reduction Potential and Control Costs” (Laxenburg, Austria: International Institute for Applied Systems Analysis).

    • Search Google Scholar
    • Export Citation
  • Krewski, Daniel, Michael Jerrett, Richard T. Burnett, Renjun Ma, Edward Hughes, Yuanli Shi, Michelle C. Turner, C. Arden Pope III, George Thurston, Eugenia E. Calle, and Michael J. Thun, 2009, “Extended Follow-Up and Spatial Analysis of the American Cancer Society Study Linking Particulate Air Pollution and Mortality,Research Report 140 (Boston, MA Health Effects Institute).

    • Search Google Scholar
    • Export Citation
  • Krupnick, Alan, 2007, “Mortality-Risk Valuation and Age: Stated Preference Evidence,Review of Environmental Economic Policy, Vol. 1, pp. 26182.

    • Search Google Scholar
    • Export Citation
  • Krupnick, Alan, Anna Alberini, Maureen Cropper, and Nathalie Simon, 2000, “What Are Older People Willing to Pay to Reduce Their Risk of Dying?Discussion paper (Washington: Resources for the Future).

    • Search Google Scholar
    • Export Citation
  • Lepeule, J., F. Laden, D. Dockery, and J. Schwartz, 2012, “Chronic Exposure to Fine Particles and Mortality: An Extended Follow-up of the Harvard Six Cities Study from 1974 to 2009,Environmental Health Perspectives, Vol. 120, pp. 96570.

    • Search Google Scholar
    • Export Citation
  • Levy J., S. Wolff, and J. Evans, 2002, “A Regression-Based Approach for Estimating Primary and Secondary Particulate Matter Intake Fractions,Risk Analysis, Vol. 22, pp. 895904.

    • Search Google Scholar
    • Export Citation
  • Lükewille, A., and M. Viana, 2012