An Economic Evaluation of the Health Effects of Reducing Fine Particulate Pollution in Chinese Cities

Air pollution, especially the ambient fine particulate matter known as PM_2.5, which are particulates with aerodynamic diameter ≤2.5 micrometers (μm), is a risk to human health (Dockery et al. 1993, Pope et al. 2002). The benefit of reduction in premature mortality risk attributable to lowering PM_2.5 comprises the vast majority of the overall benefit of air pollution control policies, and has been used to inform policy efficiency in various regulatory contexts (United States Environmental Protection Agency 2006, 2011). For example, the cost–benefit analysis of the Clean Air Act in the United States (US) suggests that in 2020 the total annual benefit will reach $2 trillion (in 2006 prices), more than 30 times the total compliance costs; of the total benefit, 85% is due to reductions in mortality attributable to ambient PM_2.5 (hereafter PM_2.5 mortality) (United States Environmental Protection Agency 2011).

Rapid economic growth powered by surges in fossil fuel consumption and urbanization in many developing countries in Asia, such as the People's Republic of China (PRC) and India, have led to increases in air pollution and adverse health outcomes that are much more severe than in developed countries. Of the 1.37 billion people living in the PRC, 83% live in areas where the PM_2.5 concentration exceeds the PRC's ambient air quality standard (AQS) of 35 micrograms per cubic meter (μg/m³) (Liu et al. 2016a). The northern, eastern, and central regions of the PRC have been exposed to annual PM_2.5 concentrations ranging from 40 μg/m³ to 100 μg/m³. Meanwhile, fine particulate pollution was associated with 4 million premature deaths worldwide in 2015, including 1 million in the PRC and 1 million in India (Global Burden of Disease Study 2015).

Recently, fine particulate pollution and its impacts have become national concerns and are now among the top political priorities in the PRC (Young et al. 2015; Jin, Andersson, and Zhang 2016). Stringent policies have been implemented on a national scale. Yet, the economic benefit of PM_2.5 control is unclear, and the effectiveness and efficiency of current policies is still being debated (detailed information is provided in section II). The targets of PM_2.5 concentration reduction, the prioritization of local interventions across sites and sectors, and the pace of the fine particulate pollution control process have been frequently questioned (Liu 2015; Jin, Andersson, and Zhang 2016, 2017). A sound economic analysis of the benefit of fine particulate pollution control is urgently needed to inform efficient policy design and implementation in the PRC.

In this study, we measure the benefit of fine particulate pollution control by quantifying PM_2.5 mortality and its welfare implications for the census-registered population in the urban areas of 300 cities at the prefecture level and above (hereafter major cities) in the PRC.¹ We develop an analytical framework that only requires publicly available data on PM_2.5 concentration and official statistics. We focus on analyzing three monetized benefits of PM_2.5 mortality reduction: (i) the theoretical maximum benefit, achieved by reducing PM_2.5 concentration from the current level to zero, which is identical to the total cost of PM_2.5 mortality;² (ii) the total benefit of mortality reduction achieved by meeting the World Health Organization's (WHO) Air Quality Guidelines (AQG) (TB_AQG) and three interim targets for fine particulate pollution (TB_IT-1, TB_IT-2, and TB_IT-3);³ and (iii) the benefit of mortality reduction achieved by a uniform 10 μg/m³ decrease in PM_2.5 concentration (UB₁₀).

Three main considerations motivate our research design. First, we focus on clearly defined city-level jurisdictions across the PRC rather than on provinces or grid-based maps with square cells. Most provinces in the PRC are vast territories, and the distributions of income, population, and pollution are highly uneven within each province. Therefore, the estimates aggregated at the national or provincial levels used in prior studies (see, for example, World Bank 2007, 2016; Xie et al. 2016a) may mask heterogeneity in PM_2.5 mortality. Furthermore, because it is the municipal governments that implement policies at the local level, it is more useful to provide estimates at the level of these micro jurisdictions. One may naturally think that the recent grid-based studies mapping PM_2.5 mortality distribution (see, for example, Lelieveld et al. 2015, Liu et al. 2016a, Xie et al. 2016b) are suitable for this purpose, but it is often burdensome to match the estimates of multiple cells with the borders of cities and districts.⁴ Therefore, in this study we directly estimate outcomes at the level of city jurisdictions.

Second, we focus on the urban areas of major cities.⁵ The greatest benefit of a reduction in PM_2.5 mortality will occur in urban areas because they have larger populations, more fine particulate pollution, and higher income levels, which are associated with a greater willingness to pay to reduce mortality risks. Further, in developing countries like the PRC, databases of monitored PM_2.5 concentrations and official statistics on socioeconomic status, which are essential for the validity of welfare estimates, are much more comprehensive for urban areas than rural areas. For simplicity, in the rest of this paper, we use “cities” to refer to the urban areas of major cities, unless indicated otherwise.

Third, in addition to considering the total cost and total benefit, we analyze the benefit of mortality reduction achieved by a uniform 10 μg/m³ decrease in PM_2.5 concentration in cities across the PRC. The distribution of UB₁₀ across cities is of both polity and research relevance. Most cities in the PRC are exposed to fine particulate pollution levels far above WHO's first interim target (35 μg/m³), and it may take decades of effort to meet this standard. Therefore, the economic benefit of a unit of pollution reduction, for example, a 10 μg/m³ PM_2.5 concentration reduction, is policy relevant. Efficient fine particulate pollution control requires that the cost of reducing PM_2.5 pollution by one unit should be lower than the estimated UB₁₀ for that unit change. However, the estimation of UB₁₀ is complicated by two issues. Recent evidence shows that the concentration response (CR) function between fine particulate pollution and mortality risks may be nonlinear (concave), especially at high PM_2.5 concentrations (Burnett et al. 2014, Pope et al. 2011). The concave CR function indicates that in cites with the same population density, the same reduction in PM_2.5 concentration will lead to less mortality risk reduction in cities with higher concentrations than in cleaner cities, which is counterintuitive and has raised environmental equity concerns (Pope et al. 2015). However, citizens in dirtier cities can sometimes have higher incomes and may therefore be more willing to pay to reduce health risks. These contradictory effects make the real distribution of UB₁₀ among different types of cities a complicated empirical question.

This study contributes to the growing literature on the economic analysis of the health impacts of air pollution in developing countries. In particular, our estimation framework uses only publicly available data; is based on clearly defined jurisdictional urban areas; and considers the total cost of PM_2.5 mortality, the benefit of meeting different air quality standards, and the benefit outcomes of a uniform reduction in PM_2.5 concentrations. Our study focuses on the outcomes of annual PM_2.5 exposure in 2016, but the same approach can be applied to other years and countries. This framework enables us to not only evaluate the environmental benefits of pollution control, but also to provide stakeholders with an easy-to-use tool for generating inputs for policy impact analyses.

Our analysis suggests that ambient PM_2.5 in the urban areas of major cities in the PRC caused 0.67 million premature deaths in 2016. The percentage of deaths attributable to fine particulate pollution in urban areas is twice the national average, indicating that urban residents face considerably higher mortality risks than the general population. For 2016, the aggregated total cost of PM_2.5 mortality in major PRC cities was about CNY1,172 billion.⁶ The average per capita cost of PM_2.5 mortality is CNY2,255. The aggregated benefit of mortality reduction from a uniform 10 μg/m³ decrease in PM_2.5 concentrations in these cities was about CNY141 billion in 2016, which is equal to a per capita UB₁₀ of CNY321. Our results show significant heterogeneity across cities with different characteristics. Cities with medium to high PM_2.5 concentrations, high per capita income, and large populations have the highest total cost. Cities with lower PM_2.5 concentrations, high per capita income, and large populations can realize the most benefit from a uniform 10 μg/m³ decrease in PM_2.5 concentrations. Meeting the PRC's current AQS of 35 μg/m³ will realize a very limited reduction in PM_2.5 mortality; therefore, all cities in the PRC should aim at reducing fine particulate pollution to levels below the stringent WHO third interim target and AQG.

The rest of the paper is organized as follows. Section II describes the available data and the challenges of fine particulate pollution in the PRC. Section III develops an analytical framework to quantify the premature deaths from fine particulate pollution and the welfare measures of total cost, total benefit, and the benefit of a uniform 10 μg/m³ decrease in PM_2.5 concentration. Section IV presents the results and section V discusses the implications for the literature and policy makers. Section VI concludes the paper. The list of technical terms used in this paper is found in the Appendix.

We collect three sets of publicly available data to form a city-level, cross-sectional data set for all the major cities in the PRC in 2016. They include the following data: (i) city-specific annual average PM_2.5 concentration, (ii) urban census registered population sizes, and (iii) urban per capita disposable income.⁷ The fine particulate pollution data are available from the Qingyue Open Environmental Data Center.⁸ The urban population and income data are collected from provincial statistical yearbooks.⁹ With this data set, we can assess PM_2.5 mortality, the various welfare indicators, and their heterogeneity across cities with different characteristics. Before presenting our formal estimation, we provide an overview of the PRC's ambient fine particulate pollution, recent policies, and how current pollution levels compare to various air quality targets.

The PRC's economy has been growing rapidly for decades; uneven regional development has concentrated most of the population, fossil fuel consumption, and vehicles in city clusters in the eastern and central PRC. As a result, citizens in these areas enjoy much of the benefit of economic development but have more severe environmental problems such as air pollution. Cities in the northern PRC are exposed to more severe air pollution than the national average, mainly because of the coal consumed by heavy industries and by urban central heating during the winter in these areas. Furthermore, rural households in the northern PRC use coal and biomass for winter space heating. Due to incomplete combustion, these fuels make substantial contributions to ambient fine particulate pollution throughout the region (Liu et al. 2016b). In addition to coal, vehicle emissions are important contributors to pollution in all cities in the PRC (Shao et al. 2006, Tsinghua University 2006). In megacities such as Beijing and Shanghai, vehicles make the greatest contribution to local fine particulate pollution, exceeding other sources such as coal combustion, road and construction site dust, and industry processes (Beijing MEPB 2014, Shanghai MEPB 2015). Finally, regional-scale transported PM_2.5 also affects cities in the plains (Xu, Wang, and Zhang 2013), whereas for the basin cities in the central PRC such as Chengdu and Chongqing, meteorological conditions that are unfavorable to pollutant diffusion worsen local pollution (Liang et al. 2016).

In January–February 2013, a winter-long haze caused by extremely high PM_2.5 concentrations enveloped the whole northern and eastern PRC. Fine particulate pollution and its impacts on health became, for the first time, a nationwide concern (Chris and Elizabeth 2016). The strong political will to control PM_2.5 pollution was seen in the quick and stringent response from the Government of the PRC. The China State Council (2013) issued the Air Pollution Prevention and Control Action Plan, 2013–2017 (hereafter the Action Plan), which contains compulsory regional air quality improvement goals and 10 key tasks. The goals were largely uniform within each region but varied between regions: the Action Plan required the Beijing–Tianjin–Hebei, Yangtze River Delta, and Pearl River Delta regions to reduce their annual average PM_2.5 concentration by 25%, 20%, and 15%, respectively, between 2013 and 2017, whereas the rest of the PRC was assigned the more relaxed goal of reducing particulates with aerodynamic diameter ≤10 micrometers (PM₁₀) by 10% over the same period. Faced with these mandatory air quality improvement goals, local authorities implemented a variety of regulations. Although improvements were observed at the end of the 5-year policy window, it was difficult for the northern region of Beijing–Tianjin–Hebei, which has much higher PM_2.5 levels (Tsinghua University and Clean Air Alliance of China 2014), to reach the policy goals even though regulations with very high compliance costs had been implemented (Jin, Andersson, and Zhang 2017; Wang 2017). However, the cleaner region of the Pearl River Delta reached its goals easily (China Environmental Protection Association 2016).

In the long term, the PRC's cities must meet the challenge of more stringent air quality targets. WHO AQG for annual average PM_2.5 concentration is 10 μg/m³.¹⁰ WHO also defines three interim targets (IT-3, IT-2, and IT-1) to gauge different countries’ progress in reducing fine particulate pollution. In 2012, the PRC updated the AQS and set 35 μg/m³ as the annual average PM_2.5 standard, equaling WHO's first interim target level of 35 μg/m³. Figure 1 shows that the PM_2.5 concentration of all major cities in the PRC in 2016 exceeded the AQG target. Two-thirds did not meet WHO's first interim and the PRC's AQS.

In sum, reducing PM_2.5 concentration in the PRC's cities will inevitably be a difficult and long-term process. This is evident from the severe and complex nature of the pollution itself, the policy dynamics of the Action Plan, and the large gap between current pollution levels and WHO's interim targets. We examine the economic value of various PM_2.5 concentration or reduction scenarios to provide a welfare perspective on this long-term PM_2.5 control process. As defined earlier, we define the values as the total cost, total benefit of WHO's AQG (TB_AQG) and three interim targets (TB_IT-1, TB_IT-2, and TB_IT-3), and the benefit of a uniform 10 μg/m³ decrease in fine particulate pollution concentration (UB₁₀). As discussed in the next sections, total cost, together with the physical estimates of premature deaths, can represent a major part of the welfare loss attributable to fine particulate pollution across cities. The results of UB₁₀ answer important questions such as how significant is the benefit of near-term achievable PM_2.5 control, and why relatively clean cities should further reduce their fine particulate pollution. The relative sizes of the total cost, total benefit, and the benefit of a uniform 10 μg/m³ decrease in fine particulate pollution concentration have implications for setting targets and choosing appropriate speeds for PM_2.5 pollution reduction across cities.

In this section, we build an estimation framework using an integrated exposure-response (IER) model and parameters suitable for the PRC context. We first note that, regardless of the estimation scales used, the economic value of the mortality (or mortality changes) attributable to PM_2.5 will eventually be composed of the following four sets of information: (i) the exposed population; (ii) the baseline mortality rate; (iii) the CR relationship, which is a monotonic increasing function of PM_2.5 concentration with its slope usually called the CR coefficient; and (iv) the value of a statistical life (VSL), which represents the willingness to pay for mortality risk reduction aggregated over the affected population. As discussed below, holding other factors unchanged, the economic estimate will grow with an increase in any of the four factors.

Recent evidence suggests that the CR function is nonlinear (concave), and the CR coefficient is lower when the PM_2.5 concentration is higher (Pope et al. 2011, Burnett et al. 2014). This implies that using the constant CR coefficient derived from studies of cleaner developed countries may overestimate the health impacts from fine particulate pollution in places with very high PM_2.5 concentrations. To better describe the concave characteristics of the CR function, an IER model was developed by Burnett et al. (2014) and Smith et al. (2014). It has been applied in the Global Burden of Disease project (Lim et al. 2013) and recently in PRC-specific studies (Liu et al. 2016a, Xie et al. 2016a, Xie et al. 2016b, and Mu and Zhang 2015). The IER simulates a set of disease-specific RR curves that cover the whole range of PM_2.5 exposure by integrating available evidence for the health impacts of ambient air pollution, indoor air pollution from household solid fuel use, second-hand smoking, and active smoking. It therefore can be used to estimate the PM_2.5 mortality risks for cities in the PRC with high concentrations.

The Global Burden of Disease Study (2013) provides the exact $RRkci$ value and its 95% confidence interval for disease k at each integer's concentration level for the IER model. Therefore, given city j’s 2016 annual average PM_2.5 concentration (⁠$ci$⁠), we know the corresponding $RRkci$⁠.

The premature deaths from PM_2.5 are monetized using the VSL estimates for the PRC. VSL is the marginal rate of substitution between income and micro mortality risk reduction (Hammitt 2000) and is nonconstant across income groups and risk contexts (Cameron and DeShazo 2013). It is obtained by aggregating individuals’ willingness to pay for mortality risk reductions over the affected population. Therefore, the higher the income level of the affected population, the higher their willingness to pay and VSL. Meta analyses of VSL empirical estimates show that it increases with the income of the sampled population and decreases with the level of mortality risk reduction in the research (Viscusi and Aldy 2003, Lindhjem et al. 2011). As a result, Cameron (2010), among others, advocates not using a one-size-fits-all VSL for populations across sites and regulatory contexts, and proposes adjusting it to fit the population's income level and a study's risk settings.

Empirical VSL estimates for the PRC are limited, and the data are more than 10 years out-of-date. Reviews of these studies are available in Huang, Andersson, and Zhang (2017); and Jin (2017). Due to the large differences in income and health risks between developed countries and the PRC, and the changes over the last decade, applying VSL values from United States-based or old PRC studies to current PRC data sets using benefit transfer methods would incur large uncertainties. Motivated by this gap, Jin (2017) designed a state-of-the-art discrete choice experiment (DCE) that incorporated the newest IER evidence on the health impacts of air pollution and constructed risk reduction scenarios that are realistic in the PRC setting. Based on online DCE surveys implemented in September 2016 on a representative sample of over 1,000 Beijing citizens, Jin reported the VSL for air pollution mortality impacts in Beijing to be CNY3 million (95% confidence interval from CNY2.2 million to CNY5 million, in 2016 prices).¹¹ More detailed information of this DCE survey, econometric analysis, and results are available in Jin (2017).¹² This VSL estimate of Beijing's air pollution fits this study's research objective, as it is up-to-date and the risk contexts in the two studies are the same.

We define UB₁₀ as based on a 10 μg/m³ decrease, rather than on a 1 μg/m³ decrease of $ci$ because a 10 μg/m³ change is more policy relevant. A stable and significant reduction in the PM_2.5 concentration of 10 μg/m³ could in most cases only be achieved through policy, whereas a 1 μg/m³ change may occur due to natural factors such as meteorological fluctuations. Furthermore, the PRC's Action Plan sets air quality improvement goals based on percentage changes that, if transformed into concentration changes, are on a 10 μg/m³ order of magnitude. Therefore, using a 10 μg/m³ change in our estimations provides ready-to-use references for local policy economic evaluations.

Our first set of results is for the estimated premature deaths attributable to 1 year of exposure to ambient PM_2.5 in 2016 in the major cities in the PRC. Table 1 reports the estimated total reductions and disease-specific reductions. It also compares our results with those of the newest estimates given in the Global Burden of Disease Study (2015) and in Liu et al. (2016a). In the latter's analysis, the premature deaths are also calculated based on the IER model, but for the entire population of the PRC.

Our results, as shown in Table 1, suggest that 0.67 million premature deaths are attributable to ambient PM_2.5 in the urban areas of major cities in the PRC cities in the 2016 data set. Although the urban-registered population in our study area is 0.45 billion, which is only 33% of the 1.37 billion people living in the PRC, the premature deaths due to ambient fine particulate pollution account for 60% of the deaths nationally. The rate of disease-related death for urban residents is almost the same as that of the whole population, suggesting that although there is no difference between the baseline mortality risks of urban residents and the national average, the percentage of deaths attributable to ambient PM_2.5 in urban areas is almost twice the national average. These results indicate that urban residents face considerably higher mortality risks associated with ambient fine particulate pollution than the general population.

The lower part of Table 1 compares disease-specific premature deaths. Stroke accounts for the majority of premature deaths, followed by ischemic heart disease; chronic obstructive pulmonary disease and lung cancer are less common. This is consistent with the fact that the baseline mortality risks for stroke and ischemic heart disease are higher. The relative magnitude of the four types of disease-specific PM_2.5 mortalities in our study are similar to the patterns in nationwide estimates.

Cities with higher estimates of premature deaths are those with larger populations and higher PM_2.5 concentrations. The magnitudes and distribution of our results are in accordance with recent grid-based studies such as Liu et al. (2016a) and Xie et al. (2016a).

In the second set of results, we determine welfare outcomes by analyzing the monetary terms of different PM_2.5 mortality impacts. We begin with the total cost estimates; that is, the total cost of PM_2.5 mortality. It represents a majority of the annual social welfare loss if ambient fine particulate pollution remains unchanged. In other words, total cost is the theoretical maximum benefit that could result from reducing the current PM_2.5 concentration to zero. In major cities in the PRC in 2016, it aggregated to about CNY1,172 billion. The spatial distribution of total cost is shown in Figure 2. Megacities in the Beijing–Tianjin–Hebei and Yangtze River Delta regions, as well as Chengdu and Chongqing in the central PRC, have the highest total cost. Income, population, and PM_2.5 concentration are all high in these cities, pulling up their total cost estimates. In the highly developed Pearl River Delta region, although PM_2.5 concentration is relatively low, populated cities such as Guangzhou and Shenzhen have very high total cost.

We then consider UB_10, that is, the benefit of mortality reduction from a uniform 10 μg/m³ decrease in PM_2.5 concentration in 2016 in each city. It should be noted that for most of the PRC's cities with currently high PM_2.5 concentration levels, UB₁₀ measures the near-term and achievable benefits of implementing effective fine particulate pollution control measures. In contrast, for cities with a very low PM_2.5 concentration, a further decline of 10 μg/m³ is a big reduction and may be difficult to achieve. The total UB₁₀ of all cities in 2016 was about CNY141 billion. Importantly, the spatial distribution of the UB₁₀, shown in Figure 3, is very different from that of the total cost, shown in Figure 2. Beijing, Tianjin, Chongqing, and provincial capitals with high PM_2.5 concentrations such as Chengdu and Shenyang have high UB₁₀ values. However, cities with medium and low PM_2.5 concentrations, high incomes, and high populations, such as those along the eastern and southern coastline, also have very high UB₁₀.

Next, we compare the values for total cost with total benefit, which include the benefit of a reduction of mortality achieved by meeting different air quality targets and a uniform 10 μg/m³ decrease in fine particulate pollution concentration for five representative cities and for all cities (Figure 4). Intuitively, it seems that more stringent air quality targets will have greater benefits. However, a disproportionately increasing trend can be seen from TB_IT-1 to TB_IT-3 and finally TB_AQG. This is due to the concavity of the CR relationship between PM_2.5 concentration and mortality. The current PRC's AQS of 35 μg/m³, which is equal to WHO's first interim target, represents a huge reduction in PM_2.5 concentration for some cities, but the associated benefit (TB_IT-1) is very small (less than one-third of the total cost). Cities will achieve a much higher total benefit (TB_IT-3) (more than one-half of the total cost) if the annual PM_2.5 concentration is lower than WHO's third interim target level of 15 μg/m³. When cities meet WHO's AQG, the TB_AQG is about 80% of the total cost. Figure 4 illustrates the welfare implications of the uniform 10 μg/m³ decrease in fine particulate pollution concentration for different cities. For dirty cities such as Beijing and Urumqi, UB₁₀ is much lower than the total benefit, whereas for clean cities such as Guangzhou and Kunming, it corresponds to a significant pollution reduction and is therefore comparable to the total cost.

In Table 2, we list the 20 cities with the highest estimates of attributable premature deaths along with their corresponding total cost, total benefit of WHO's third interim target (15 μg/m³), and the benefit of a uniform 10 μg/m³ decrease in PM_2.5 concentration. The cities’ rankings of total cost and total benefit of WHO's third interim target vary slightly from their ranking of premature deaths, but the ranking of UB₁₀ is totally different from the other three measures. In fact, among the cities with the most deaths, cleaner and dirtier cities swap rankings for the UB₁₀, and many new cities that are generally cleaner and wealthier appear on the UB₁₀’s top 20 list.

Our third set of results considers per capita welfare indicators. We use them to remove the influence of huge disparities in cities’ populations on our results. Due to space limitations, we focus on per capita total cost and per capita UB₁₀. Per capita total cost can be interpreted as the economic cost of the annual attributable mortality risk for an average individual in city i in 2016. Per capita UB₁₀ represents the economic benefit of this individual's mortality risk reduction from a 10 μg/m³ decrease in PM_2.5 concentration. We also study how income and PM_2.5 pollution levels influence individual welfare. Ideally, we would use subsamples of populations with different socioeconomic statuses within and across cities for this purpose. However, due to data limitations, we only approximately examine this issue by looking at the difference between low- and high-income cities. We equally divide the cities into five income groups with the first group having the lowest income and the fifth group having the highest income. Then we plot the per capita total cost and per capita UB₁₀ for these cities against the PM_2.5 concentration, as shown in Figures 5 and 6, respectively. The areas of the circles represent the city's population.

For per capita total cost in Figure 5, we see that cities with the highest per capita total cost are high-income cities with PM_2.5 concentrations between 40 μg/m³ and 60 μg/m³. Residents in low-income and high PM_2.5 concentration cities (see the circles for the first and second income groups with PM_2.5 > 60 μg/m³) bear a greater burden in terms of the economic costs of PM_2.5 mortality.

Figure 6 shows a clear trend of decreasing per capita UB₁₀ as PM_2.5 concentration increases. This is contrary to the conventional understanding that the first unit of pollution reduction has the highest benefit. As indicated in previous analyses, this is due to the concavity of the PM_2.5 CR function. In each income group, the per capita UB₁₀ spans a wide range, from about CNY100 to CNY900.

Our last exercise is to compare the results for cities in the PRC with those in other countries. We choose two megacities in developed countries, New York City (population 8.5 million; annual PM_2.5 concentration in 2016, 13.5 μg/m³) and Seoul (population 10.3 million; annual PM_2.5 concentration in 2016, 27 μg/m³). We also choose Monrovia (population 1.1 million; annual PM_2.5 concentration in 2016, 22 μg/m³), the capital city of Liberia, to represent cities in least developed countries. We perform the same total cost and UB₁₀ estimation process on the three cities. It should be noted that it is a huge step for cities with low PM_2.5 concentrations to further reduce them by 10 μg/m³ (e.g., for New York City, the final PM_2.5 concentration would be near zero).

Figure 7 shows the estimated per capita total cost for all these cities. The results for the PRC's cities are well below CNY5,000, whereas they are near CNY15,000 for New York City and CNY20,000 for Seoul. Similarly, Figure 8 indicates that the per capita UB₁₀ is much lower in cities in the PRC than in cities in developed countries. The huge differences are mainly due to the concavity of the CR function and the income disparity between cities in the PRC and those in developed countries.

Figure 9 further explains the difference in the benefit of a uniform 10 μg/m³ decrease in fine particulate pollution concentration between cities. In the right part of Figure 9, we show several typical cities along the per capita UB₁₀ curve, from those with low pollution levels (Shenzhen, Sanya, and Kunming), through middle pollution levels (Guangzhou, Shanghai, and Zhangye), to high pollution levels (Chongqing, Beijing, Urumqi, and Hengshui). These cities also vary significantly in per capita income and population. For example, Beijing and Urumqi have almost the same PM_2.5 concentration levels, but Beijing has a higher per capita income than Urumqi; therefore, the per capita UB₁₀ in Beijing is higher. On the left panel of Figure 9, we multiply per capita UB₁₀ by each city's population and get squares representing the size of the UB₁₀. In this figure, the difference in the populations of Beijing and Urumqi results in more widely differing UB₁₀. Similarly, all cities with high PM_2.5 concentrations have relatively lower UB₁₀, and cities with middle to high pollution levels, relatively small populations, and low per capita income have the lowest (e.g., Hengshui, Urumqi, and Zhangye). Figure 9 also suggests that when dirty cities reduce their fine particulate pollution enough to enter the lower PM_2.5 concentration ranges, the UB₁₀ of a further reduction significantly increases. For example, if Beijing's concentration decreases to the level of Shanghai's, they would have similar UB₁₀ values, and if they both become as clean as Shenzhen, they would have very large values.

The first and foremost implication of our results is that when PM_2.5 pollution is high, the benefit of controlling it is relatively low. Thus, the PRC and other emerging economies with high levels of fine particulate pollution should accelerate their pollution control processes. Although the optimal PM_2.5 control pathway and speed depend on factors such as projections of changes in a country's technology, economy, and population, the implications of the relative size differences of the total benefit, and the increasing trend of UB₁₀ are clear: given compliance costs and assuming steady economic development, the lower the PM_2.5 concentration, the more beneficial an extra pollution control effort is to society.

Second, our results highlight the importance of considering the cost effectiveness of PM_2.5 control policies. This is especially relevant for cities with high pollution levels. To reach the socially optimal pollution level, an efficient policy requires that the marginal cost of pollution reduction is smaller than the marginal benefit.¹⁴ Thus, for cities with a UB₁₀ that is close to the marginal benefit of pollution reduction, policies to reduce high pollution have to be very cost effective. However, in the PRC and many developing countries, the cost of controlling PM_2.5 across sites and sectors is unclear. Sound economic analyses of the available technologies and policy instruments are urgently needed to enhance the cost effectiveness of PM_2.5 control. Furthermore, as fine particulate pollution can be transported over long distances, regional control strategies that combine the efforts of neighboring cities and provinces are essential to cost-effective reductions in PM_2.5 concentrations (Wu, Xu, and Zhang 2015).

Third, our estimation scheme can be easily used by the general public and the government. Our approach mainly relies on official statistics and data on PM_2.5 annual concentrations. We focus on clearly defined jurisdictional urban areas rather than on cells in grids. As pollution control costs are usually calculated for specific programs and projects within certain jurisdictional areas, our approach helps to unify the cost–benefit analyses of PM_2.5 control. This helps local authorities to understand the trade-offs between the cost and benefit for their own administrative areas, and will help neighboring cities to coordinate their pollution control efforts through economic mechanisms (e.g., transfers).

Fourth, our results imply that current policy targets of uniform percentage reductions in regional PM_2.5 concentrations do not consider the significant heterogeneity in the welfare impacts of these reductions among cities with different characteristics. We show how income, population, and original fine particulate pollution levels can alter the welfare measures of PM_2.5 pollution reduction. More efficient pollution control policies could incorporate these factors and set differential targets for different cities. Our results also demonstrate the importance of equity considerations, which are a critical part of the policy-making process. We show that for low-income and high-pollution cities, the benefit of a uniform 10 μg/m³ decrease in fine particulate pollution concentration is much lower than that for high-income and low-pollution cities. For the former, the sooner they enter the low PM_2.5 concentration range, the greater their benefit. However, these less advantaged cities often have more tension between economic growth and pollution control, and are therefore more likely to continue to have high pollution levels. Equity considerations would support the allocation of extra resources to these cities to accelerate the PM_2.5 control process.

Air pollution, especially ambient PM_2.5, is one of the most pressing challenges for the PRC and many developing countries. It results in millions of premature deaths annually. Although a growing body of literature provides increasingly precise and high-resolution information on fine particulate pollution exposure and its health impacts at the regional, national, and provincial levels, the social welfare implications of PM_2.5 mortality in urban areas have remained unclear. Multiplying the estimated premature deaths by monetary values such as VSL does not generate accurate economic results due to the nonlinear concentration response relationship between PM_2.5 and mortality, disparities in income and population, and the mismatch between the grid cells commonly seen in the literature and the more policy-relevant jurisdictional areas.

This study develops an accessible estimation scheme to provide welfare implications for PM_2.5 mortality in urban areas of the PRC. Based on the integrated exposure-response model, our approach uses publicly available PM_2.5 concentration data and city-level socioeconomic statistics to estimate the total cost of PM_2.5 mortality, the benefit of a uniform PM_2.5 concentration reduction, and the benefit of meeting WHO AQG and interim targets for the urban areas of nearly 300 major PRC cities. The results suggest that in these cities the aggregated total cost of annual PM_2.5 mortality for 2016 is about CNY1,172 billion. The average per capita total cost is CNY2,255. The aggregated benefit of mortality reduction resulting from a uniform 10 μg/m³ decrease in PM_2.5 concentration (UB₁₀) in these cities is about CNY141 billion for 2016; the per capita UB₁₀ is CNY321. Cities with high incomes and large populations that are located in areas with severe air pollution suffer the highest welfare losses, and the UB₁₀ is lower than for cleaner cities. The benefit of meeting WHO's first interim target is very low relative to the benefit of meeting more stringent air quality targets. As most of the cities in the PRC still have PM_2.5 concentrations well above WHO's first interim target of 35 μg/m³, local authorities need to accelerate fine particulate pollution control in a cost-effective manner. Only in this way can greater benefits of PM_2.5 mortality reduction be achieved.