Investments in air pollution control in Mexico City alone are likely to total more than $4.7 billion over the next five years. Even modest water and sewage projects cost hundreds of millions of dollars. Health improvements are often cited as the major justification for such investments. Consequently, one of the more troublesome problems, both practically and ethically, facing policy-makers is that of valuing the health impacts of pollution. While some argue that it is not possible (or morally ethical) to place monetary values on sickness or death, in many situations governments must decide on the appropriate level of health interventions or investments. Should available funds go to air pollution reduction, or would they be better spent on water supply and sanitation, or should priority be given to education and health care, or any other pressing concerns? Putting a value (even if underestimated) on morbidity and mortality due to pollution can be a powerful tool to demonstrate the costs of inaction.
The problems of valuing health impacts of pollution are twofold. First, it is the actual identification and measurement of health impacts; and, secondly, once impacts have been determined, it is often necessary to estimate monetary values for associated morbidity (illness) and mortality (death).
Measuring the Health Impacts of Pollution
Health impacts of air and water pollution are well recognized. Air pollution affects human health in a variety of ways - from itchy eyes and chest discomfort to chronic bronchitis and asthma attacks to premature death. There is ample evidence that inadequate water supply and sanitation can have a significant impact on the incidence of both mortality and morbidity associated with diarrhoea, intestinal nematodes and other tropical diseases.
The most accurate way to measure the health impacts of air pollution or lack of access to clean water and sanitation in a given area is to conduct epidemiologic studies for that area that establish dose-response relationships linking environmental variables to observable health effects. However, given the time and cost involved in such studies (as well as the likelihood of encountering problems of data availability) it may often be case that dose-response relationships established in other locations will have to be used instead. The note summarizes recent progress in quantifying air pollution dose-response functions and addresses major problems in applying these functions to other situations. In case of water pollution, establishing dose-response relationships is more complicated and far less advanced, since it is not ambient water quality per se that affects health, but access to clean drinking water and sanitation along with the household’s level of income and education. Box 1 describes an approach adopted in a recent study of pollution problems in Brazil.
Air Pollution Dose-Response Studies
Dose-response functions correlate mortality and morbidity outcomes of susceptible population groups with the ambient concentrations of a certain air pollutant. The majority of studies have focused on the mortality effects of exposure to particulates. Chronic exposure to particulates can lead to premature death by exacerbating respiratory illness, pulmonary disease and cardiovascular disease. Acute exposure (short-term peaks in the levels of particulates) can increase a chance that a person in a weakened state or an especially susceptible person will die. Detailed studies completed in the last years conclusively indicate that fine particulates (usually measured as PM2.5) are responsible for most of the excess mortality and morbidity associated with high levels of exposure to particulate matter (See Note 'Particulate Matter (Airborne)'). While one study that finds a statistically significant association between a health effect and a specific air pollutant does not prove causality, the inference of causation is strengthened if epidemiologic results are duplicated across several studies; if a range of effects is found for a given pollutant; and if these results are supported by human clinical and animal toxicology literature. An approach to reduce the uncertainty associated with individual study is to use meta-analytical techniques that, on the basis of the statistical pooling and aggregating of results from several studies, produce a 'best estimate' in which more confidence may be placed.
The reported epidemiologic studies involve two principal study designs: time-series and cross-sectional. More common time series studies correlate daily variations in air pollution with variation in daily mortality in a given city and primarily measure the effects of acute exposure to air pollution. The advantage of these studies is that they do not have to control for a large number of confounding factors, since the population characteristics (age, smoking, occupational exposure, health habits) are basically unchanged. Based on meta-analysis of acute mortality studies that measure the ambient levels of particulates of less than 10 microns (PM10), estimates of an average percent change in total mortality per 10 ug/m3 change in PM10 range from 0.74 (Maddison, 1997) to 1.23 (Ostro, 1996).
A cross-sectional analysis compares differences in health outcomes across several locations at a selected point or period of time and, in principal, allows to capture both acute and chronic effects of air pollution. Two types of cross-sectional long-term exposure studies can be distinguished. One is a retrospective (ecological) cross-sectional study design which correlates variations in air pollution levels with mortality rates across various locations at a single point in time. Such studies have consistently found measurably higher mortality rates in cities with higher average levels of particulate matter in the US. A common concern about these studies is, however, whether potential omitted and confounding variables have been adequately controlled.
A second type involves a prospective cohort design in which a sample of population is selected and followed over time in each location. These studies use individual level data so that other health risks factors can be better taken into account. Two of such studies conducted to date (Dockery et al., 1993; Pope et al., 1995), both in the US, reported a statistically significant correlation between exposure to particulates matter (measured as PM10 or PM2.5) and mortality, which was considerably higher by comparison with acute mortality studies (a 4.2 percent change in all-cause mortality per 10 ug/m3 change in PM10 in the Pope et al. study). Prospective cohort studies have potentially greater value for public health and environmental policies, and for estimating dose-response functions that can be applied elsewhere. However, these studies are most expensive, so their replication is most difficult.
Since cohort studies are few and cross-sectional studies are less reliable, the question remains how to factor long-term exposure to particulates into the results based on acute exposure mortality studies that seem to provide lower-bound estimates for health effects of air pollution.
Apart from mortality counts, dose-response functions can also be derived for many lesser health impacts, for example respiratory hospital admissions (RHA), emergency room visits (ERV), bed disability days (BDDs), restricted activity days (RAD), minor restricted activity days (MRAD), asthma attack(AAs), acute respiratory symptoms, chronic bronchitis, lower respiratory illness (LRI), and so on. Main results from meta-analyses of available studies for a number of key air pollutants (PM10, sulfur dioxide, nitrogen dioxide and ozone) are summarized in Attachment, Table A.1.
Valuation of Health Impacts
Several methods have been used in various studies to value health costs associated with environmental pollution. These methods can be grouped in two broader categories. The first one includes methods that measure only the loss of direct income (lost wages and/or additional expenditures). These measures, however, do not include inconvenience, suffering, losses in leisure and other less tangible impacts to individual and family well-being, and may seriously underestimate or completely ignore the health cost of people who are not members of the labor force. Therefore, these methods indicate only the lower bound of the social sosts and understate the total costs to individuals. The second category includes approaches that attempt to capture the Willingness-To-Pay (WTP) of individuals for avoiding or reducing the risk of dealth or ill-health. The principal techniques are summarized in Table 1 and discussed below.
Lower Bound of the Social Costs Estimates
The Human Capital (HK) approach that places a value on a premature death is the easiest but perhaps the least accurate and most ethically troubling. It considers individuals as units of human capital that produce goods and services for society. Just as the useful life of man-made capital can be calculated based on the discounted stream of future production, the human capital theory assumes that the value of each unit of human capital is equivalent to the present value of the future output, in the form of earnings, that might have been generated had the individual not died prematurely.
The values calculated are very dependent on the age of death (both the very young and the old would have small values when the HK approach is used) and on income, skill level and country of residence. For example, the use of HK approach in Mexico resulted in valuing each life lost due to exposure to TSP pollution at US$ 75,000. Another study, in Brazil, estimated the cost of premature death at US$ 7,700 for San Paulo in 1989 and US$ 25,000 for Cubatao in 1988. The big difference between San Paulo and Cubatao accounted for the difference in average age at which exposed people died.
Even based only on this minimal estimation of the cost of deaths, the economic benefits of investments that prevent the health effects of pollution are often apparent. For example, the minimal estimate of the 'worth' of the life of a child is sufficient to outweigh the costs of a major immunization program.
Cost of Illness (COI) approach applies to morbidity and is consistent with HK approach. The direct cost of morbidity can be divided into two categories: medical expenditure for treating illness (a large portion of hospital admissions cost and emergency room visits costs), and lost wages during days spent in bed, days missed from work, and other days when activities are significantly restricted due to illness.
An example of this method is an estimate of the cost of non-lethal diarrhea in Mexico. In 1990 these cases were estimated to number 3,360,000. The costs of treatment and laboratory analysis alone were considered, and came to US$30 million, or about $9.00 per person. (It is instructive to note that this figure represents less than 1% of the cost estimated for deaths from similar causes.)
Preventive Expenditures. By observing the amounts people living in polluted area spend on averting measures to reduce health risk, it is possible to make tentative inferences about the minimum amount they are willing to pay to reduce these risks. For example expenditures on bottled water can be used to infer the minimum value people are willing to pay to avoid water-borne diseases.
Willingness-To-Pay (WTP) approaches
If people's preferences are a valid basis upon which to make judgments concerning changes in human 'well-being', then it follows that changes in human mortality and morbidity should be valued according to what individuals are willing to pay for (or willing to accept as compensation to) the change in health status or the risks that they face. This is not the same as valuing an actual life, and should not be interpreted as such. Instead it involves valuing ex-ante changes in the level of risk people face and then aggregating them. Since the exact identity of those at risk is unknown, valuing ex-ante changes in the level of risk is the appropriate policy context.
The Wage Differential (WD) approach uses differences in wage rates to measure compensation people require for (perceived) differences in the chance of dying or falling ill from occupational hazards. Recent WD studies in the US have produced estimates of the value of statistical life (VOSL) in the range of US $1.9 - $10.7 million (1990 dollars).
Contingent Valuation (CV) approach uses survey information to determine what people are prepared to pay to reduce the likelihood of premature death or certain diseases. Contingent valuation studies have produced slightly lower estimates of US $1.2 - $9.7 million (1990 dollars) per statistical life.
A question being often asked is how a difference in age distribution of those involved in WTP studies and those primarily affected by pollution should change the respective estimate of VOSL. WD studies measure compensation for risk of instaneous death for people of about 40 years old and thus value approximately 35 years of life (Viscusi, 1993). Because death from air pollution reduces life-years by less than 35 years on average, labor market estimates should be adjusted accordingly. For instance, the relative numbers of over-65s versus under-65s who will die prematurely from air pollution in the US (estimated as 85 percent of over-65s) coupled with some evidence of a lower WTP for the former group (about 75 percent of the mean WTP), would adjust the respective VOSL downwards by 20 percent.
Another approach is to use the concept of DALYs (disability-adjusted life years) that is a standard measure of the burden of disease (WDR 1993; Murray and Lopez, 1996). DALYs combine life years lost due to premature death and fractions of years of healthy life lost as a result of illness or disability. A weighting function that incorporates discounting is used for years of life lost at each age to reflect the different social weights that are usually given to the illness and premature mortality at different ages. Thus, it is possible to link the VOSL obtained from WD/CV studies with the corresponding number of DALYs lost in order to estimate the implicit value per DALY, as well as to adjust the respective VOSL according to an average number of DALYs lost in a specific study. DALYs can also serve as an independent aggregate measure of health benefits (losses) in cost-effectiveness analysis of pollution control policies and options.
Although the valuation of morbidity is very important to cost-benefit of air pollution control programs as well as to many other areas of economic activity, these estimates are much more limited in scope and based entirely within the United States.
How Can it be Used in Developing Countries?
How appropriate is it to transfer the results from air pollution dose-response studies in industrial countries into the context of developing countries? Three issues should be carefully addressed here. First, measures of particulate matter. Using epidemiologic studies based on PM10 or PM2.5 and availability of PM10 or PM2.5 measurements in the country in question are essential for more reliable results from applying dose-response functions derived elsewhere. Secondly, disease-specific mortality profile. If the distribution of deaths by cause significantly differs between the country of interest and the countries where dose-response studies originate from, using dose-response functions for disease-specific mortality or adjusting for this difference may be needed to improve the accuracy of projections. For instance, exposure to particulates affects primarily respiratory and cardiovascular deaths that make half of all deaths in the US. In Delhi, India, fewer than 20 percent of all deaths are attributable to these causes. Therefore, even identical reaction by susceptible groups of population in Delhi and the US to the change in the levels of particulates would result in a lower total mortality in Delhi (Cropper and Simon, 1996). Third, age pattern of deaths from air pollution causes. Age profile of those affected by air pollution may be very different in developing countries than in rich countries. While peak effects were observed among people of 65 and older in the US (Schwartz and Dockery, 1992), in Delhi peak effects occur in the 15-44 age groups that implies more life years lost as a result of a death associated with air pollution (Cropper, et al., forthcoming).
The need to adjust the social costs of mortality and morbidity for income levels in different countries is obvious. In the US, for example, estimates of VOSL are typically five to ten times higher that the value of forgone earnings. If people in other countries were equally risk averse, then it would be appropriate to multiply the value of forgone earnings by the same factor. It is plausible to assume, however, that the risk aversion varies with living standards and that the value of a statistical life in developing countries is a smaller multiple of forgone earnings relative to the US. Unfortunately, the literature on the income elasticity of WTP for reducing the risk of insults to health is extremely limited, and empirical analyses in rich countries do not lend sufficient support to this assumption (Maddison, et al., 1997). Until further research is conducted, one possible approach is to simply adjust an average US value for the income difference between the countries. For a conservative estimate of the VOSL in a developing country, a lower bound US value, adjusted for the income difference, can be used.
Application of this approach to valuing a variety of health effects of exposure to particular matter in China produced estimates of the total health costs of urban air pollution country-wide in a range of US$ 27 - $ 32 billion; and the costs attributed to mortality in a range of US $11-$15 billion. It is important to stress that, under a number of assumptions on dose-response levels and base costs for these health effects drawn from different studies, the social costs of morbidity consistently appeared to be as significant as those of mortality.
The approaches to measuring physical impact and health cost of pollution outlined above represent the cutting edge of research in this area and, though carefully scrutinized in view of best available evidence from toxicological, epidemiological and economic work, is inevitably surrounded by some degree of uncertainty and controversy. It is presented to respond to the need of Bank staff and consultants to strengthen the economic analysis of pollution control projects and policies, and to help in advising policy makers on the necessary level of targets and interventions. Application of these approaches to a specific context of any particular project/study requires careful interpretation, and the limitations of the analysis should be fully understood before making conclusions and recommendations on its basis.
Box 1 Health Benefits of Water Supply and Sanitation in Brazil: A New Approach
Objective. Few studies have attempted to use epidemiologic results on water-related diseases as the basis for setting priorities in expanding water and sanitation services. A recent World Bank study in Brazil drew upon a detailed epidemiologic study of the impact of water and sanitation on infant and under-5 mortality to estimate the net benefits of improvements in water and sanitation.
Methodology. The analysis was carried out using a sample of 1533 municipalities from 4 states - Minas Gerais, Pernambuco, Rio de Janeiro, and San Paulo - which span the full range of incomes and living conditions in Brazil today. The main independent variables used in analysis were: average head of household income; percentage of population living in urban areas; percentage of females aged 5 or greater who are illiterate; percentage of urban population served by piped drinking water; and percentage of total population served by sewers and/or septic tanks. The analysis established that the coefficients for the variables characterizing income per person, the level of female education, and access to piped water and sanitation are highly significant and negative. The coefficient on the level of urbanization turned to be very significant and positive, so that infant and under-5 mortality rates are higher in urban areas than in rural areas if other factors are held constant. The relative importance of water and sanitation can be illustrated by the impact on mortality rates of 10 percent point increases in water and sanitation variables, shown below.
Impact of water and sanitation on mortality rates:
- Infant Mortality Under-5 mortality
- Change (reduction) per a 10 percent point rise in:
- Urban access to piped water 0.8 0.25
- Urban access to sewers 0.6 0.15
- Average mortality rate 39.4 8.8
Findings. The health benefits that would be generated by expanding urban and sewer services are large. For 4 states, the analysis shows that it should be possible to avoid nearly 3,000 deaths of babies and young children each year to reduce the burden of disease by nearly 220,000 DALYs (disability-adjusted life years). The largest impact would be achieved by ensuring that the entire urban population has access to piped water at average cost of $ 1,560 per DALY. The average cost per DALY saved by expanding urban sewers is much higher at $2,440, but it is still well below a reasonable estimate of WTP to save a DALY in Brazil. Even if the VOSL were set at only $ 1 million for the US, this would imply an average WTP per DALY saved for Brazil of about $5,500 in 1991. This is well above the annualized costs per DALY saved by expanding water supplies and sewers for all but a small number of municipalities.