Market-based policies for air-pollution control
Three decades of rapid economic growth in China have been accompanied by severe environmental degradation. In July 2007, the Financial Times headlined an article about a World Bank report on this problem, “750,000 a year killed by Chinese pollution.” Our estimate of the number of lives shortened by air pollution in 2002, described below, is very similar: 710,000. The World Bank report put annual health damages from air pollution in China at 3.8 percent of gross domestic product (GDP) in 2003. The estimate of damages from rural water pollution alone was 1.9 percent of rural GDP. Both figures suggest huge pollution costs, in absolute terms and relative to other countries.
Current levels of air pollution in China far exceed international environmental standards. Particulate matter from smoke damages health through fine particles that lodge deep in the lungs. In 2003, the average concentration of PM10 (particulate matter smaller than 10 microns) for 52 northern Chinese cities was 140 μg/m3 (micrograms per cubic meter), compared to the World Health Organization’s healthy air guideline of 20 μg/m3.
Historically, estimates of pollution concentration are expressed in terms of total suspended particulates (TSP). The average TSP concentration in these northern cities, where coal is burned for heating, is 337 μg/m3. Although there has been improvement in many areas since the 1990s, the concentrations of particulate matter far exceed China’s own standards. The average TSP concentration of major cities in China in 1990 was 379 μg/m3 and by 2003 was still 256 μg/m3, exceeding the Chinese national standard of 200 μg/m3. By comparison, on the eve of the landmark Clean Air Act signed by President Richard Nixon in 1970, the average TSP concentration in the United States was 70 μg/m3. Even at the ninetieth percentile—the top tenth of most polluted areas—the U.S. concentration level in 1970 was only 106 μg/m3.
By any standard, then, China’s pollution problem is severe: costly in lives and in economic impact. We want to put the problem in context, and to detail an analysis—using modern social-science and public-health tools—of how that nation can best address it. We are impressed by China’s dynamism and economic strengths, which hold out the potential for making significant environmental progress. Though it is not directly the subject of our analysis, China is also a significant and growing contributor to greenhouse-gas emissions, the culprit in global warming. The policies we suggest could have a powerful positive impact there, too—and wider application throughout the world.
It is instructive to contrast China’s pollution problems with those of other nations. Thomas Rawski of the University of Pittsburgh has pointed out that comparable levels of air pollution were observed in the United States, Japan, and Korea in earlier stages of development. In Pittsburgh, TSP levels above 300 μg/m3 were recorded even as late as the 1940s, while Tokyo had levels above 400 μg/m3 in 1968.
One major difference between China and more developed countries is the degree of urbanization. Even today, China is much less urbanized than the United States, Japan, and Korea were in their periods of high pollution. As late as 2003, only 41 percent of China’s citizens resided in urban areas; 74 percent of the U.S. population was already living in cities by 1970. The area definitions are not identical, but these figures show the potential for much more urbanization in China, which could expose many more people to high levels of urban pollution.
Policy analysts assign a dollar value to environmental damages in order to compare these damages with the costs of controlling pollution. As noted above, air-pollution damages are estimated by the World Bank at 3.8 percent of China’s GDP. As its citizens’ incomes rise with the rapid economic growth, these valuations will rise. We project that air-pollution damages as a share of GDP in China would double in about 20 years under current policies, due to rising incomes, increased urbanization, and slow improvement in air quality.
In the face of truly daunting environmental challenges, the Chinese government has developed relatively sophisticated institutions to address the problems. The U.S. Environmental Protection Agency was not established until 1970, when the country’s per capita income was $21,000 (in 2005 dollars). By 2005, when China’s per capita income was only $4,100 in comparable terms, the 17-year-old State Environmental Protection Agency and other agencies had acquired a high level of technology for dealing with these environmental issues.
Traditional methods of controlling pollution in developed countries take the form of direct regulation and rigorous technology standards, such as mandatory scrubbing of power-plant emissions and pollution-control standards for automobiles. More recently, the United States has introduced market-based mechanisms, such as permit-trading programs for sulfur emissions, in order to reduce the cost of pollution abatement. Given the magnitude of China’s pollution problems and the expected surge in energy use, we believe that China should seriously consider market-based approaches to environmental protection.
This paper describes an international effort to study China’s air-pollution problem in an integrated fashion, incorporating the costs of reducing pollution as well as the benefits of pollution abatement. The 10-year study involved more than a dozen environmental engineers, epidemiologists, and economists from Tsinghua University in Beijing and Harvard University. The base year of 2002 was chosen in order to provide a comprehensive set of data on the Chinese economy and environment. The research resulted in models of air pollution, health effects, and the economy that have made it possible to analyze the impact of market-based approaches to environmental policy in China.
The Harvard-Tsinghua study considers the impact of pollution-control policies on emissions of greenhouse gases, such as carbon dioxide (CO2). But our research emphasizes local air pollution, a major problem that ranks high on the Chinese government’s agenda. We focus on how particulate matter, sulfur dioxide, and nitrogen oxides affect human health. Instead of examining specific technologies or traditional regulatory regimens to control local pollution, we consider “green taxes”—taxes proportional to the damage caused by pollution. As we will show, these measures have substantial effects on emissions of local air pollutants.
Describing the Damages
Formulating an integrated pollution-control program for China might appear to require a staggering amount of information about the formation of particulate matter, sulfur dioxide, and nitrogen oxides and the pattern of atmospheric dispersion and concentration of the pollutants. Information on human exposure to pollution and the resulting health impacts, a valuation of the damages, and the cost of pollution-control policies would also be necessary. Obviously, it is not practical to model the dispersion from every source, even if emissions inventories were available (and for many pollutants in China, they are not).
The Harvard-Tsinghua study is described in Clearing the Air: The Health and Economic Damages of Air Pollution in China, edited by Mun S. Ho and Chris P. Nielsen (MIT, 2007), and this essay provides updated information. The goal of the study was to develop a convenient methodology that links emissions and human exposures. Using this information, we estimate health damages for a given level of emissions from each industry. We then incorporate these estimates into a model of the economy to generate an assessment of the benefits and costs of pollution control. Our methodology involves the following steps:
Step 1. From economic activity and fossil-fuel use to pollutant emissions. We characterize the economic output of China’s 33 industrial sectors, plus the household sector, and the consumption of fossil fuels—coal, oil, and natural gas—by each. For each sector, we then estimate emissions of three pollutants—total suspended particulates (TSP), sulfur dioxide (SO2), and nitrogen oxides (NOx)—from fuel combustion and other production processes. Fossil-fuel combustion yields most of the SO2 emissions, while production-process emissions are mostly TSP from the cement industry.
As an example of the data problems we face, the damages caused by particulate matter depend crucially on the size of the particles (see step 3, below). Current epidemiological studies use data on PM2.5 (particles finer than 2.5 microns) where such information is available. For China, however, comprehensive data are available only for TSP, which includes larger particles. We therefore calibrate our estimates to the official national TSP data, and convert to PM10 equivalents, using data from six Chinese cities where both measures were available.
Step 2. From emissions to concentrations. The Harvard-Tsinghua study uses a relatively simple model for dispersion of emissions within 50 kilometers and a more sophisticated model for regional dispersion covering most of China. Researchers from the department of environmental science at Tsinghua calculated the dispersion of TSP and SO2 for a sample of more than 600 smokestacks and road segments in five cities, concentrating on cement, iron and steel, and chemicals plants.
We exploit a national database on electric power plants in China and estimate the dispersion from 160 smokestacks. For a smaller sample of power plants, we also calculate the concentration of “secondary particles” (sulfates and nitrates) formed in the atmosphere from SO2 and NOx. The measured concentration of TSP in any particular location is due to the sum total of these secondary particles and the primary particles emitted from smokestacks. Although it is obvious that different industries produce different levels of emissions per unit of output, it is less obvious that each ton of emissions produces a different level of health damages—reflecting differences in meteorology, smokestack characteristics, proximity to dense populations, and particle size distributions.
Step 3. From concentrations to human exposures. Given the concentration of pollutants, we need to estimate human exposure to each pollutant. The analysis is straightforward for a particular smokestack, but is impossible for millions of emission sources. We therefore use a methodology developed by researchers at the Harvard School of Public Health (HSPH) to approximate the emission-exposure relationship from a small sample of sources. This involves estimating the “intake fraction”: the fraction of a pollutant emitted from a particular source that is eventually inhaled by people before it is dissipated. For every kilogram of SO2 emitted by our sample of cement plants, for instance, 4.41 μg are inhaled within 50 kilometers of the source. On the other hand, analysis reveals that—given their higher stack heights and more remote locations—electric utilities have the lowest intake fractions, the least damage per ton emitted. Such results show the importance of taking secondary particles into account when calculating potential health damage, because the intake fraction for the sulfates produced by cement plants, for example, is the same order of magnitude as for the TSP the plants produce. Studies that fail to take this into account will miss, for instance, the rapidly growing emissions of nitrogen oxides from motor vehicles. Given the intake fractions from the sample of emissions sources, we estimate national intake fractions using information from national databases on enterprises.
Step 4. From exposures to health impact. Next, we use air-pollution epidemiology to link the pollution exposures with adverse health outcomes—premature mortality, chronic bronchitis, and asthma attacks. We use a conservative estimate of a 0.03 percent increase in “acute” mortality per each microgram per cubic meter increase of PM10 and SO2. This is equal to 1.95 excess deaths per million people annually per μg/m3 increase in concentration. We also calculate lesser effects (such as asthma, chronic bronchitis, and general respiratory symptoms) attributable to the increase in each pollutant. Our study does not examine longer-term “chronic” effects, as these data do not yet exist for China.
Step 5. Valuation of health effects. Finally, we monetize health damages using risk analysis in order to compare the benefits and costs of pollution reduction. We express the change in the number of cases of chronic bronchitis, premature mortality, and so on, in terms of yuan, the Chinese currency. We also briefly consider long-term chronic mortality effects, because these are not very well understood yet for the Chinese population.
Air Pollution Damages by Industry Output and Fuel Use
In our model, the intake fractions from step 3, combined with the industry emissions from step 1, yield the dosage of a given pollutant due to a particular industry. These dosages, including primary and secondary particulates, after being adjusted for the breathing rate, generate concentration equivalents. When linked with the coefficients from step 4, the concentrations yield the health effects due to emissions from each industry sector. Finally, the health effects, evaluated in monetary terms and aggregated across all the effects, give the value of damages attributable to that industry. Thus, the results from steps 1 through 5 enable us to estimate the damages per unit of output and per unit of fuel use in order to calculate green taxes.
We assume that the incremental damage from an additional unit of output from an industry is equal to the average damage. This is expressed in terms of yuans’ worth of damages per ton of cement or per kilowatt hour of electricity produced. Finally, we express this as damages per yuan of industry output in 2002 currency units.
The electricity, steam, and hot water industry has the highest incremental damage, 7.6 Chinese cents per yuan of electricity output. This industry does not have the highest intake fraction, but has very large emissions per yuan of output.
Electricity is followed by nonmetallic mineral products, such as cement (2.9 cents per yuan) and transportation and warehousing (1.9 cents per yuan). Some service industries, including commerce and real estate, have surprisingly high incremental damages because they still rely on coal heat. The total value of damages for all sectors is 213 billion yuan, equivalent to 1.8 percent of China’s GDP in 2002. Of this total, the electricity sector, with its large TSP and SO2 emissions, contributes 28 percent, followed by transportation with 12.6 percent and nonmetallic mineral products with 7.8 percent. Some 89 billion yuan of the 213 billion total is due to primary TSP, the remainder is due to SO2, or SO2 and NOx, transformed into secondary particles.
Beyond damages per unit of output, we are also interested in the damages per unit of fossil fuel consumed. The result is an enormous 53.5 cents of damage per yuans’ worth of coal burned. By contrast, the marginal damage from cleaner, more expensive, oil is only 2.9 cents per yuan. Gas is relatively clean and generates negligible amounts of PM and SO2, although it does generate CO2.
Calculating Green Taxes
Informed by estimates of the damages associated with various industrial sectors and fuel sources, we have the basis for implementing green-tax policies. Ideally, emissions of pollutants should be taxed directly. But it is infeasible to measure the emissions from millions of sources. Industry output and fuel consumption are much more easily measured. Accordingly, we consider taxes on industry output, based on the damages per unit of output, and taxes on fuels, based on the damages per unit of fuel consumed.
In our model, Chinese economic growth is driven by labor-force growth, capital accumulation, and growth in productivity (or output per unit of input). We assume productivity growth matches the current high rate of 3 percent per year before tapering off. Energy used per unit output depends on the price of energy and changes in technology. We project future changes in technology using information from the more mature U.S. economy.
In our model, energy demand consists of demand by enterprises and demand by households and the government. Household consumption depends on the price of energy and the changes in preferences that come with rising incomes. We project a rising share of total consumption allocated to automobile, gasoline, and electricity consumption, based on patterns observed in other countries. The model is calibrated to the 2002 benchmark economic data, supplemented by energy-use data from the China Statistical Yearbook 2006, and emissions data estimated by our partners at Tsinghua University. The environmental module incorporates the five steps outlined above.
Simulating the model, we obtain a base case with the economy growing at 6.8 percent per year during the next 30 years and energy use rising at 5.1 percent. We then simulate the two policy cases. For the output tax, we add tax rates equal to the incremental damage rates to existing taxes on outputs. For the fuel taxes, taxes are imposed on coal, oil, and gas, proportional to the estimated damages.
Because green taxes raise new government revenues, the economic outcome depends on how these revenues are used. Cutting taxes has effects quite different from giving transfer payments to households to compensate consumers for the higher costs of their purchases. We chose to cut taxes on enterprises, because personal income taxes are very low in China. The results of comparing the tax policies in the first year are given in the adjacent table (note that much of the specific industry-by-industry data has been omitted).
Output tax. The proposed output tax consists of a heavy tax on electricity, a modest tax on cement and transportation, and a small tax on the other commodities. The initial effect is to raise the price of electricity by 4 percent and the prices of the other polluting commodities by about 1.0-2.5 percent. These price changes lead to a fall in the output of electricity by 4.7 percent in the first year. The output of nonmetallic mineral products, transportation, and metal smelting falls by 1 to 3 percent. The labor and capital released from these sectors allow an expansion of the cleaner ones: trade, real estate, and electronic products rise by 0.1 to 1 percent. Coal consumption falls by 3.4 percent.
The fall in fossil-fuel use and the change in the composition of output lead to a reduction in primary TSP combustion emissions of 3.3 percent and a 4.0 percent reduction in SO2 emissions. NOx emissions from transportation fall by 1.7 percent. The emissions of the greenhouse gas CO2 fall by 2.7 percent—a little less than the change in coal consumption, because there is a shift to oil and gas. The effect of lower emissions is to lower health damages by 2.6 percent, which is worth 0.04 percent of GDP. This is a modest improvement in the environment compared to the revenues raised.
Consumption falls by 0.2 percent in the first year, because households face higher prices but do not get tax relief. The new revenues allow a large tax cut for enterprises, which leads to higher retained earnings and investment. In this scenario, higher investment leads to a higher GDP in future years. By the twentieth year, in fact, GDP is 0.6 percent higher, allowing both higher consumption and higher investment. There could be a “double dividend” in some situations—that is, reduced pollution and increased consumer welfare in all years. Although we do not have such a win-win situation here, we should note that current high taxes on enterprises are very distorting, so that reducing them contributes to the small welfare cost we have estimated.
Why does the large tax lead to such a small improvement in the pollution levels? This occurs because the tax on output does not encourage any pollution-reduction effort or fuel-switching. Changes in coal consumption and pollutant emissions come entirely from the shift in the composition of output from polluting commodities to cleaner ones.
Fuel tax. The proposed fuel tax consists of a large tax on coal and a modest one on oil. This does encourage the switch from dirty coal to cleaner oil and gas and the substitution of capital for energy. (Because this does not directly penalize emissions, a fuel tax does not replace pollution-reduction efforts like installing scrubbers on smokestacks or washing coal before combustion.)
As shown in the table, coal use falls by 12 percent in the first year and oil use by 0.4 percent, because the major users of these fossil fuels have to raise their output prices to compensate for the tax, causing a reduction in demand for their products and services. Electricity is the biggest user of coal and output falls by 2.2 percent. Petroleum processing, metals smelting, and nonmetal mineral products fall by 1.0, 0.5, and 0.5 percent respectively. With the resources released from this contraction of the coal-intensive industries, the less energy-intensive sectors expand, including agriculture, food products, trade, and construction.
Changes in output mix and fuel switching reduce both primary combustion TSP and SO2 emissions by 10 percent in the first year. The modest tax on oil reduces transportation output and NOx emissions by only 0.9 percent. Because this green tax is narrowly focused, the new revenue raised is only 1.8 percent of total revenues. These reductions in emissions lead to a large reduction in health effects of pollution. Premature mortality, for example, falls by 9 percent. A relatively small, well-targeted tax reduces health damages by 9 percent, triple the effect of the output tax. This reduction in health damages in the first year is worth 0.13 percent of GDP, compared to 0.04 percent in the output tax case.
With the fall in coal use, the emissions of CO2 fall by 9 percent. Such a reduction would be a substantial contribution to the global goal of limiting climate change, given that China in 2004 produced 18.2 percent of the world’s total CO2 emissions from fossil-fuel combustion.
Finally, in the first year of a fuel tax, consumption falls because households are not compensated for the higher prices. In the fuel-tax scenario, the new revenues are much smaller than those from an output tax, yielding a corresponding smaller tax cut for enterprises. As a result, the impact on investment and, hence, future GDP is small.
Because per capita incomes are rising rapidly in China, the valuation of health damages is also rising rapidly. This should lead to higher green-tax rates over time, inducing larger reductions in coal use and emissions. By the twentieth year, coal use is down 15 percent, compared to the base case, and health damages are down 11 percent. This environmental benefit is equivalent to 0.34 percent of GDP, at the cost of a modest 0.04 percent fall in consumption.
The results of our simulations are sensitive to the assumptions we have made about price responses, the estimated health effects, and valuations of these effects. Nonetheless, we conclude that the benefits of green taxes in China greatly exceed the costs. Several lessons stand out from our analysis.
First, the benefits of reducing air pollution far exceed the cost of reduced consumption. Although we have not modeled the short-run adjustment costs, such as relocating laid-off coal miners, these could be mitigated by gradually phasing in green taxes.
Second, there are trade-offs between the effectiveness of a tax instrument and the ease of implementing it. A broad-based tax may gain greater acceptance because the costs are shared by many, but it is unlikely to be effective. A narrow tax, targeted at the main polluters or fuels, is more efficient but requires larger adjustments. This calls for a careful consideration of compensating policies such as displacement and adjustment assistance.
Third, because of the link to coal use, efforts to reduce local pollution will substantially reduce China’s large contribution to greenhouse-gas emissions. This outcome argues for international efforts to help China improve energy efficiency and reduce local pollution. This benefit is, of course, in addition to reducing the dispersion of SO2 and secondary particles to neighboring countries.
Finally, our integrated methodology for studying the costs and benefits of air-pollution control could be adapted for other countries. The intake-fraction approach is most useful where data are limited and modeling air dispersion is costly. Although our integrated model retains all the uncertainties of its underlying components, our methodology allows for improvements as more and better data become available.