Understanding china's urban pollution dynamics

Journal of Economic Literature 2013, 51(3), 731–772 http://dx.doi.org/10.1257/jel.51.3.731

731

1. Introduction

In 1960, 33.0 percent of the world’s popu-lation lived in cities. In 2010, this share grew to 50.5 percent and will continue to rise as urbanization in the developing world takes place. Urbanization offers market

opportunities that rural areas cannot match. The potential to learn, specialize and trade in cities raises per capita income (Glaeser 1998, 1999, 2011).

Over the last thirty years, one quarter of the rural people who entered cities world- wide were in China.1 In China’s cities, tens

1 http://esa.un.org/unpd/wup/unup/index_panel1.html.

Understanding China’s Urban Pollution Dynamics†

Siqi Zheng and Matthew E. Kahn*

China’s ongoing urban economic growth has sharply increased the population’s per capita income, lowered the count of people living below the poverty line, and caused major environmental problems. We survey the growing literature investigating the causes and consequences of China’s urban pollution challenges. We begin by studying how urban population and industrial growth impacts local pollution levels and greenhouse gas production. As the urban population grows richer, its demand for private transportation and electricity sharply increases. Such privately bene!cial activity exacerbates urban pollution externalities. Facing these severe environmental challenges, China’s urbanites increasingly demand quality of life progress. We survey the emerging literature investigating the demand for environmental progress in China. Progress in mitigating externalities hinges on whether the powerful central and local governments choose to address these issues. We analyze the political economy of whether government of!cials have strong incentives to tackle lingering urban externalities. We conclude by discussing future research opportunities at the intersection of environmental and urban economics. (JEL O18, P25, P28, Q53, R23, R41, R58)

* Zheng: Department of Construction Management and Hang Lung Center for Real Estate, Tsinghua Univer- sity. Kahn: University of California, Los Angeles, National Bureau of Economic Research, and IZA. We thank Janet Currie (the editor) and +ve reviewers for constructive com- ments. We thank Cong Sun for excellent research assis- tance. We thank the UCLA Ziman Center for Real Estate and PKU-Lincoln Institute Center for Urban Develop- ment and Land Policy for generous funding. Siqi Zheng

thanks the National Science Foundation of China (No. 71322307, No. 70973065, and No. 71273154), Program for New Century Excellent Talents in University (NCET- 12-0313), and Tsinghua University Initiative Scienti+c Research Program for research support.

† Go to http://dx.doi.org/10.1257/jel.51.3.731 to visit the article page and view author disclosure statement(s).

http://esa.un.org/unpd/wup/unup/index_panel1.html
http://dx.doi.org/10.1257/jel.51.3.731
Journal of Economic Literature, Vol. LI (September 2013)732

of thousands of new housing and commer- cial of+ce towers are being built in months to accommodate such huge urban growth. Millions of new vehicles are being registered. A YouTube video shows the construction of a +fteen-story hotel erected in six days.2 China is gearing up to supply a massive amount of electricity to meet the demands of growing cities featuring higher income urbanites. China’s electricity consumption in 2011 was roughly 4.5 trillion kilowatt hours.3 Ongoing economic development has reduced China’s poverty rate from 84.02 percent in 1981 to 13.06 percent in 2008.4

There are signi+cant environmental con- sequences caused by this urban growth (for trend evidence see Vennemo et al. 2009). Two leading indicators of city pollution are local air pollution and greenhouse gas pro- duction (GHG). Today, many cities in China have extremely high air pollution levels. Based on an ambient particulate concentra- tion criterion of PM10 , twelve of the twenty most polluted cities in the world are located in China (World Bank 2007b). In 2003, 53 percent of the 341 monitored cities— accounting for 58 percent of the country’s urban population—reported annual average PM10 levels above 100 !g/m3, and 21 percent of China’s cities reported PM10 levels above 150 !g/m3.5 Only one percent of China’s urban population lives in cities that meet the European Union’s air quality standard of 40 !g/m3 (World Bank 2007a). Unless ambi- tious policy action is taken, GHG emissions from the BRIC nations, i.e., Brazil, Russia, India, and China, are expected to grow by

2 http://www.youtube.com/watch?v=E76uJi744Do. 3 http://www.bloomberg.com/news/2011-01-28/china-s-

power-demand-growth-may-slow-to-9-this-year-nea-says. html.

4 http://databank.worldbank.org/ddp/home.do (Poverty and Inequality Database). In their database, poverty line is set as the daily income below 1.25 U.S. dollars.

5 Particulate matter less than 10!m in diameter, i.e., +ner particles, are typically used in health damage assessments.

46 percent from 2005 to 2030, and in total could roughly equal emissions from the thirty OECD countries combined by 2030 (OECD Environmental Outlook to 2030, 2008).6

This article surveys the recent literature on the environmental consequences of urban growth with a focus on pollution dynamics in China’s cities. We will be careful to contrast urban pollution dynamics that affect local environmental criteria such as ambient air pollution, water pollution, and sanitation ver- sus the global challenge of climate change.

If urban environmental quality is sacri+ced as economic growth takes place, then per capita income growth over-states improve- ments in the standard of living (Nordhaus and Tobin 1971). Easterlin et al. (2012) document that self reported life satisfaction indicators have not increased in China as much as would be expected during a time of annual 8 percent economic growth.

A long-run urban environmental history for developed nations suggests that local environmental problems could improve in Chinese cities. Optimists can point to a vari- ety of examples ranging from air to water pollution. In each of these cases, the urban externality +rst became worse over time and then sharply improved. U.S. studies such as Cain and Rotella (2001), Clay, Troesken, and Haines (2006, 2010), Cutler and Miller (2005), Ferrie and Troesken (2005), and Haines (2001) have each highlighted the role that government investment and regula- tion played in taming urban environmental externalities. The transition from relying on dirty coal to cleaner fuels such as natural gas for cooking and heating has reduced urban particulate levels in U.S and European cit- ies (Clay and Troesken 2010). The U.S. experience also highlights the role that fed- eral legislation such as the Clean Air Act has played in reducing urban pollution (Chay and

6 http://www.oecd.org/document/26/0,3746,en_2649_3 4283_40243802_1_1_1_1,00.html.

http://www.youtube.com/watch?v=E76uJi744Do
http://www.bloomberg.com/news/2011-01-28/china-s-power-demand-growth-may-slow-to-9-this-year-nea-says.html
http://www.bloomberg.com/news/2011-01-28/china-s-power-demand-growth-may-slow-to-9-this-year-nea-says.html
http://www.bloomberg.com/news/2011-01-28/china-s-power-demand-growth-may-slow-to-9-this-year-nea-says.html
http://databank.worldbank.org/ddp/home.do
http://www.oecd.org/document/26/0,3746,en_2649_34283_40243802_1_1_1_1,00.html
733Zheng and Kahn: Understanding China’s Urban Pollution Dynamics

Greenstone 2005, Reyes 2008). These exam- ples demonstrate the key role that the central government plays in mitigating urban envi- ronmental externalities. In a one party nation such as China, whether the central and local governments pursue aggressive antipollution regulations will be a crucial factor determin- ing China’s environmental quality dynamics.

These long-run trends documenting the rise and fall of urban pollution levels in developed countries bear a similarity with the Environmental Kuznets Curve (EKC) lit- erature. Building on the in-uential Grossman and Krueger (1995) study, an entire sub+eld of environmental economics has emerged that focuses on the association between national per capita income and different indi- cators of pollution (see Hilton and Levinson 1998 and Harbaugh, Levinson, and Wilson 2002).

In this survey, we will examine several factors that determine a city’s environmen- tal quality at a point in time and over time. Per capita income will be one of these fac- tors but we will not embrace the EKC as a unique “law of physics” that offers a deter- ministic law of motion for pollution in every city at every point in time. Unlike the typical EKC study, we will focus on the incentives and choices of individuals, +rms, and gov- ernments and how the byproduct of these choices determines local pollution levels and population exposure.

We focus on cities instead of the whole country mainly for three reasons. First, cities are centers of production and consumption. Second, with the fast urbanization taking place in China, more and more people are moving to cities and thus are exposed to ele- vated pollution levels. Third, understanding the trade-off between the positive (agglom- eration) and negative (congestion and pol- lution) externalities of urban growth has long been the core issue in urban and envi- ronmental economics (Tolley 1974; Glaeser 1998). Our review contributes to this ongoing

literature by presenting a detailed analysis of China’s urban dynamics.

To organize our survey, in section 2 we present a conceptual framework that sketches how urban pollution and exposure to this pollution arises within a spatial equi- librium model featuring optimizing +rms and households. This approach allows us to preview the various supply side and demand side factors that we will discuss in detail in later sections of this survey. On the supply side (section 3), we provide a detailed over- view of the roles of the industrial sector, power generation sector, transportation, and urban form in determining pollution pro- duction. On the demand side (section 4), we predict that many of China’s cities featuring rising educational attainment and per capita income levels will aspire to become the high quality of life consumer cities we observe in the United States and Western Europe (Glaeser, Kolko, and Saiz 2001).

Growing cities face a fundamental tragedy of the commons problem. In the absence of a Coasian bargaining solution, government has a central role in mitigating pollution external- ities. In section 5, we will pay careful atten- tion to China’s unique political structure at the national and local levels with an emphasis on the incentives and constraints that politi- cians face in supplying “green cities.”

While this survey will focus on the environ- mental consequences of China’s recent urban growth, we will be careful to point out when we are con+dent that the results generalize to other LDC cities. At the end of this survey, we present topics that merit future research.

2. Conceptual Framework and Data Sources

In this section, we present a conceptual overview of urban pollution production, the demand for environmental protection, and the government’s role in mitigating pollution

Journal of Economic Literature, Vol. LI (September 2013)734

externalities. This section seeks to highlight how the various detailed subsections we present below +t together into a coherent framework. At the end of this section, we will discuss the various sources of pollution data in China and the quality of those data sets.

We start with two well-known theoreti- cal models that provide micro foundations for the EKC (Andreoni and Levinson 2001; Stokey 1998). In Stokey’s (1998) model, pol- lution is a function of the scale of aggregate economic activity but it can be reduced by investing in costly cleaner technology. For very poor nations, the social planner chooses to use the dirtiest technology and pollution increases as income rises. As income rises above a critical level, the planner chooses to sacri+ce some consumption in order to use the costly cleaner technology and an EKC can emerge. Andreoni and Levinson (2001) offer a simple parametric example. The representative consumer gains utility (U) from consumption (C), and loses utility from exposure to pollution (P). This repre- sentative agent spends resources (M) on con- sumption (C) and emissions control (E), with the latter reducing pollution (P). The algebra can be expressed as:

(1) U = U(C, P) = C " P (2) P = P(C, E) = C – C#E$

(3) M = C + E. Pollution has two components, C and

–C#E$. The former term is directly propor- tional to consumption, and the latter term represents “abatement,” which is related to consumption (C) and resources spent on environmental effort (E). The latter term has a standard Cobb–Douglas form (0 < #, $ < 1). Andreoni and Levinson (2001) solve for the closed form solution of optimal pollution production and study how

pollution evolves as a function of income (M). In this pareto planner’s problem, an Environmental Kuznets Curve (EKC) emerges if # + $ > 1.

This association between national per capita income and different indicators of air and water pollution was documented in a cross-national study by Grossman and Krueger (1995). A continuing research agenda seeks to study the empirical valid- ity of this hypothesis. In some cases such as lead emissions it emerges (Hilton and Levinson 1998), while subsequent cross- national research by Harbaugh, Levinson, and Wilson (2002) suggests the EKC is a fragile empirical result.

These models of the relationship between pollution and economic activity do not explicitly incorporate geography. A straight- forward way to introduce spatial consider- ations is to incorporate ideas from two classic urban models—the Rosen (1979)—Roback (1982) open system of cities model and the classic urban monocentric model developed by Alonso (1964), Muth (1969) and Mills (1967). Those spatial approaches allow us to analyze the possibility that households and +rms can migrate across cities and within a single city. We will +rst discuss the blending of the nonspatial and spatial models on the supply side (pollution production), and then move to the demand side (i.e., the demand for greenness). We then discuss how local and national government choices over regu- lations and public goods provision are likely to affect key outcomes.

2.1 Pollution Production in Cities

A long tradition in urban economics, start- ing with Alonso, Muth, and Mills (together called the AMM model), and Rosen and Roback, models the spatial compensating differentials equilibrium across and within cities. Factor prices for labor and land adjust so that the marginal household/worker is indifferent across locations (Rosen 2002). If

735Zheng and Kahn: Understanding China’s Urban Pollution Dynamics

a city suffers from low quality of life, then its wages and rents will adjust to compensate workers for such disamenities (Albouy 2009). Within cities, if a speci+c neighborhood suf- fers from low quality of life, then rents will be lower.

In the classic AMM model, all employ- ment is located downtown in the city center. This urban model does not introduce indus- trial pollution but an easy way to introduce pollution is to posit that industrial emissions are proportional to industrial employment and the pollution is more heavily concen- trated in geographical areas closer to the downtown factories. In this case, the resi- dential communities close to the city center will offer a trade-off of short commute times but higher pollution levels. If jobs and hous- ing are more spread out throughout the met- ropolitan area, then pollution “hot spots” in the city center are less likely to exist as the jobs and population will be more uniformly distributed.

Andreoni and Levinson’s (2001) and Stokey’s (1998) models adopt a single sector production function approach. While greatly simplifying the problem, they abstract away from introducing heterogeneity with respect to industrial structure, vintages of capital, and the spatial distribution of industrial activities and households. These speci+c details play a key role in our analysis below.

In section 3, we will blend the core ideas of the above spatial and nonspatial mod- els and set up an accounting framework to track the main sources of urban pollution. There are two key pieces to this framework. The +rst dimension is to examine pollution sources in two basic categories—produc- tion based (such as industrial production) and consumption based (such as residential electricity consumption and private vehicle driving) externalities. The second dimension categorizes pollution indicators that are local (such as air pollution) and global externali- ties (greenhouse gas emission).

2.2 Demand for Green Cities

Avoiding pollution represents an input in the production of health capital. Human capital and health capital are complements (Currie 2009), so urban environmental qual- ity is a key determinant of urban earnings and people’s happiness. The Grossman (1972) model of health production can be aug- mented to include location-speci+c attributes such as local air and water pollution levels. A household will recognize that everyone within the household is exposed to similar pollution levels once a residential location choice has been made. If households differ with respect to their susceptibility then the most vulner- able households will choose to self-protect by living in low pollution cities and cleaner parts of a speci+c city. In this sense, this location choice is a type of self-protection investment (Ehrlich and Becker 1972).

As the average household in Chinese cities becomes richer and better educated, there will be rising demand for living in “green cities.” Such households will seek informa- tion and products that help them to reduce their exposure to health risks. The core idea of the Rosen–Roback model is that urbanites are able to “vote with their feet” in an open system of cities. This migration leads to local wages and rents adjusting so that high qual- ity of life cities will feature relatively lower wages and higher rents. In section 4, we discuss recent empirical +ndings based on revealed preference techniques for studying the demand for “green cities” in China.

2.3 Introducing Government

To simplify the discussion, we have not introduced a national or local government in the above demand-side and supply-side discussion. But the urban air and water are common property and it is costly to monitor who is emitting what into the commons. At any point in time, there are millions of urban actors who are taking privately bene+cial

Journal of Economic Literature, Vol. LI (September 2013)736

actions such as driving or turning on the air conditioning that have social consequences. In the absence of pollution pricing, no one has an incentive to internalize such externali- ties and transaction costs preclude a Coasian bargaining solution.

Standard public +nance economics highlights the role that government can play in taming Pigouvian externalities and providing public goods. Given China’s sharp economic growth and rising educa- tional attainment, it is noteworthy that the J-Curve theory conjectures that the inten- sity of environmental regulation increases as per capita income rises (Selden and Song 1995). The J-Curve hypothesis provides a government induced micro foundation for an observed EKC relationship. Hilton and Levinson’s (1998) work on urban lead emis- sions highlights that as nations grow richer they reduce the allowable lead content per gallon of gasoline.

La Porta et al. (1999) and Botero, Ponce, and Shleifer (2012) argue that the quality of governance and hence its capacity to supply public goods increases when a nation’s per capita income and educational attainment rises. Such governments have the resources to collect information and hire competent people to enforce laws. In democracies, poli- ticians who seek reelection have an incen- tive to pursue policies that the median voter demands. Whether a one party nation’s lead- ers are also responsive to citizen preferences remains an open research question (see Mulligan, Gill, and Sala-i-Martin 2004 for some cross-national evidence).

In section 5, we will discuss China’s gov- ernance structure, its public +nance system and analyze the objectives and priorities of of+cials at different levels of government. The central and local governments have different incentives for pursuing the green agenda. The central government is keen to address domestic energy security con- cerns and seeks to be a global leader in the

nascent green technology market. A desire to achieve increased “legitimacy” with the Chinese people and to raise China’s image in the international community may further motivate the central government to pursue “environmentally friendly” policies. The central government has recently shifted the local of+cials’ promotion criteria from solely focusing on economic growth to “harmony development” and this latter criteria places increased importance on tangible environ- mental goals. Urban of+cials who seek pro- motion within the system now have stronger incentives to meet environmental targets. While this could create incentives to cheat and fudge reporting data, the rise of infor- mation technology and micro blogging and the reduction in the cost of collecting up to date environmental data reduces the con- cern about strategic misreporting.

2.4 Data Sources and Data Quality

Data on pollution in China is relatively scarce compared to developed countries, and questions about the data quality have been raised. Here we provide a brief over- view of pollution data sources and discuss their reliability. There are two broad sources of pollution data. One is the of+cial pollution data released by the Chinese government, and the other is data collected by research- ers and nongovernmental organizations. The former category includes pollution indicators published in a series of yearbooks, and on the websites (or reports) of China’s Ministry of the Environmental Protection (MEP) and local environmental authorities. Table 1 pro- vides a brief summary of the major sources of publicly available key environmental indi- cators in China. Two main data collection methodologies are employed for generating these data sets. One is bottom–up account- ing—summing up the self-reported numbers from individual +rms to cities, provinces and then to the national level. Examples include solid waste disposal, waste water discharge,

737Zheng and Kahn: Understanding China’s Urban Pollution Dynamics

TABLE 1 M!"#$ D!%! S#&$'() #* K(+ Q&!,-%+ #* L-*( I./-'!%#$) -. C0-.!

Name Data level Period available Data source

Public health Life expectancy By province From 2003, by year China Statistical Yearbook of Health

National-level From 1995, by year International Statistical Yearbook Infant mortality By province From 2003, by year China Statistical Yearbook of Health

National-level From 1994, by year International Statistical Yearbook U5MR National-level From 2005, by year China Statistical Yearbook of Health Mortality at birth National-level and

by province From 2003, by year China Statistical Yearbook of Health

Tuberculosis incidence rate National-level From 1997, by year China Statistical Yearbook Tuberculosis and heart disease death rate

National-level From 2001, by year Annual Statistical Report of Health in China

Pollution Ambient air pollution: PM10 86 major cities From 2006, by day Data Center, Ministry of Environmental

Protection, available online 31 major cities From 2003, by year China Statistical Yearbook of

Environment Water quality of main 1ows By water 1ow From 2004, by week Data Center, Ministry of Environmental

Protection, available online Water quality of major river systems

By river From 2002, by year Annual Statistical Report on Environment in China

Generation and discharge of industrial solid waste

By province and by city (all municipality- level cities)

From 2003, by year China Statistical Yearbook of Environment, and China Statistical Yearbook

Emission/discharge and treat- ment of industrial smoke, SO2, waste water

By province and by city (all prefecture-level cities)

From 2003, by year China Statistical Yearbook of Environment, and China Statistical Yearbook

Urban garbage disposal and treatment

By province From 2005, by year China Statistical Yearbook of Environment

Urban road traf2c noise 30 major cities From 2005, by year China Statistical Yearbook of Environment

Energy consumption Total electricity consumption By industry and

by province From 1990, by year China Energy Statistical Yearbook

Total coal consumption By industry and by province

From 1990, by year China Energy Statistical Yearbook

Total natural gas and LPG consumption

By industry and by province

From 1990, by year China Energy Statistical Yearbook

Residential coal, natural gas and LPG consumption per capita

All prefecture- level cities

From 1985, by year China City Statistical Yearbook

Residential electricity and water consumption per capita

All prefecture- level cities

From 1995, by year China City Statistical Yearbook

Number of civil vehicles By province From 1985, by year China Statistical Yearbook Total gasoline and diesel consumption

By industry and by province

From 1995, by year China Energy Statistical Yearbook

Gasoline price National-level From 2000, by year International Statistical Yearbook

Journal of Economic Literature, Vol. LI (September 2013)738

smoke emissions, sulfur dioxide emissions, and pollution control investment by city and by province. Such data are less reliable because +rms and local governments both have the incentive to underreport their emis- sions but overreport their economic outputs and pollution-mitigating efforts.

Of+cial pollution data are also collected from ambient pollution monitors. Examples include readings of ambient particulates con- centration and water quality by city. These data sets are posted on the MEP’s website. Such monitors are controlled by MEP, the province or the city’s environmental authori- ties, so it is hard for individual +rms to in-u- ence them. Therefore such monitoring data is more reliable (especially those controlled by MEP), though some scholars also point out that some city governments have the incentive to manipulate such data (Wang et al. 2009; Andrews 2008).

Skeptics have questioned the quality of data produced by China’s government. They argue that the statistical approach to data collection, reporting and validation is opaque. Similar to the questions raised about the statistics released by the Soviet Union, when a one party state controls information releases it may systematically choose to release information that helps it to achieve its political goals and may suppress negative information (Liu and Yang 2009; Guan et al. 2012). There have been some controversies on the disparities between publicly released and privately collected pollution data, such as the Beijing air qual- ity readings from the Beijing environmental authorities and the U.S. Embassy (see our detailed discussion in section 4). Due to the rise of information technology, the Chinese government has been losing its data provi- sion monopoly. Improvements in remote sensing and cheaper pollution monitors have allowed others to measure China’s pol- lution levels (Akimoto et al. 2006; Sinton 2001; Zhang et. al. 2007). We predict that

the quality of Chinese data will continue to improve attributed to technological innova- tions that foster data supply competition, and also due to the increased demand for pollution related information by educated Chinese urbanites (see section 4).

Figure 1 highlights some recent trends in ambient air pollution between 2003 and 2010 for thirty-+ve major cities and +fty- one medium-sized cities using data from the MEP’s website. Though the absolute PM10 concentration level is still quite high, a regression with city +xed effects yields a time trend decline by 2.19 percent per year. This declining trend is larger for large cit- ies (3.09 percent). Figure 2 presents some time trends between 2001 and 2010 in sur- face water quality. Similar to the air quality trends, the time trend suggests signi+cant progress but indicates that water pollution still remains a serious challenge.

Figure 3 displays the trends in per capita CO2 for the world, India and China, based on data from the World Development Indicators. China’s sharply sloping per capita line starting in 2001 highlights the serious global pollution consequences of China’s sharp recent economic growth.

3. Understanding the Causes of China’s Urban Pollution Growth

In this section, we employ an account- ing framework to discuss the major causes of China’s urban pollution, including city population growth, industrial activity, driv- ing and urban form, power generation and winter heating. At the end of this section, we will discuss the global externalities—natu- ral resource extraction and GHG emissions. In each category, we emphasize the role of scale, composition and technique effects (Copeland and Taylor 2004). Scale refers to the sheer count of people and jobs located in a city while composition refers to a city’s industrial composition and the vintage of its

739Zheng and Kahn: Understanding China’s Urban Pollution Dynamics

capital stock. Technique represents emis- sions per unit of economic activity.

Those sectors in the economy all contrib- ute to a city’s overall public health externality. Particulate matter (PM) is an important indi- cator of local air pollution that raises mortal- ity risk as highlighted in U.S studies (Chay and Greenstone 2003). One PM2.5 emissions

inventory study for China (Cao et al. 2011) indicates that on average, industrial activity is responsible for 68.6 percent of total PM2.5 emission, 4.5 percent is from transportation, and 20.4 percent is from the residential sec- tor. In the case of sulfur dioxide emission, 30.5 percent is from power generation, 63.2 percent from industrial sector, and 5 percent

Figure 1. Air Quality Indicators in Chinese Cities

Source: China Statistical Yearbook, Data Center of Ministry of Environmental Protection (MEP) of the People’s Republic of China.

Panel A. Annual percentage of days with air quality at or above grade II (de!ned by mep) in Chinese major cities (2000 to 2010)

Panel B. PM10 Concentration (mg/m3) in Chinese major cities (2003 to 2010)

70

75

80

85

90

95

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

(%)

0.08

0.09

0.10

0.11

0.12

2003 2004 2005 2006 2007 2008 2009 2010

(mg/m3)

All cities (with city !xed effects): ln(PM10) = – 0 .0219 ! year trend –2.382

Large cities (with city !xed effects): ln(PM10) = – 0.0309 ! year trend –2.271

35 major cities Other large and medium-sized cities

35 large and medium-sized cities Other 51 major cities

Journal of Economic Literature, Vol. LI (September 2013)740

from the residential sector. Individual cities have quite different emission inventory percentages.7

3.1 City Population Growth

There are hundreds of millions of rural Chinese households seeking the employ- ment bene+ts of urbanization as well as nonwage bene+ts (quality of life and learn- ing opportunities) (Harris and Todaro 1970). Geographers focusing on China have documented the migration trends (Fan 2005a, 2005b). The level of urbanization in China increased by about one percentage point a year from 26 percent in 1990 to 51 percent in 2010.8 The 2010 census reveals

7 In Beijing’s PM10 inventory, 21.3 percent of this pollut- ant is from soil dust, 18.2 percent is from coal burning, 16.5 percent is from vehicles, 6.6 percent is from construction sites, 14.7 percent is from other particles (such as S O 4

2" , N O 3

" and N H 4 + ), and 22.7 percent is from nonidenti+ed source (Chen et al. 2006). Hangzhou has a very different PM10 emission inventory with the respective shares being 17.0 percent, 13.9 percent, 16.9 percent, 8.0 percent, 29.1 percent, and 15.1 percent (Bao et al. 2010).

8 National Bureau of Statistics of China (2012a).

that 78 percent of migrants originated from rural areas. The three biggest winners in terms of population gains during 2005–10 as a percentage of their initial 2005 popula- tion were Zhejiang (a leading manufactur- ing belt in the coastal southeast), Beijing (the national capital) and Shanghai (China’s emerging commercial center). They are all located in the eastern region. It is expected that 70–80 percent of Chinese people will live in cities by the year 2030.9

Moving people to cities introduces two offsetting environmental effects. In cities, there are more people living in a small geo- graphic area (i.e., high population density) but they are richer. The former effect should mean that urbanization reduces energy con- sumption while the latter effect suggests that urbanization increases energy consump- tion. Zheng et al. (2011) documents both of these effects and show that the income effect

9 h t t p : / / w w w. b i s . g o v. u k / a s s e t s / f o r e s i g h t / d o c s / migration/11-1116-migration-and-global-environmental- change.pdf.

Figure 2. Surface Water Quality in China (2001–2010)

Source: Status of China Environment reports 2001–2010.

20

30

40

50

60

70

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

(%) The percentage of Chinese National Monitoring River Sections with water quality of grade iii or above

Grade I~III

http://www.bis.gov.uk/assets/foresight/docs/migration/11-1116-migration-and-global-environmental-change.pdf
http://www.bis.gov.uk/assets/foresight/docs/migration/11-1116-migration-and-global-environmental-change.pdf
http://www.bis.gov.uk/assets/foresight/docs/migration/11-1116-migration-and-global-environmental-change.pdf
741Zheng and Kahn: Understanding China’s Urban Pollution Dynamics

dominates in China. Given that the average rural person only consumes one third of the energy consumed by the average urban per- son in China, this vast migration -ow will result in signi+cant increases in energy con- sumption.10 Past U.S. research has found that urbanization offers environmental bene+ts for rural areas that depopulate. Pfaff (1999) documents the reforestation of the Northeast as people urbanized during the nineteenth century. In China, however, Ebenstein et al. (2011) document the negative environmen- tal impacts caused by China’s massive rural to urban migration. As rural workers have urbanized, farmers have faced labor shortages and have responded to this by substituting chemical fertilizer as an input in production.

10 In 2000, the annual total energy consumption of an average urban person was 210 kg standard coal, while that of an average rural person was 76 kg standard coal, about 36 percent of the former (China Energy Statistic Yearbook 2011).

Such chemicals raise local nitrogen levels and this increases water pollution.

Despite the huge migration -ow tak- ing place, the Chinese hukou system has restricted domestic migration. Au and Henderson (2006) estimate an econometric model to explain city GDP as a function of the city’s population, and generate empiri- cal estimates of the GDP-maximizing city size. They +nd that most Chinese cities are undersized due to the hukou constraint.11

11 The hukou system, put in place in the 1950s, was to register people by their hometown origin and by urban ver- sus rural status for the purpose of regulating migration. In the wake of transition to a market economy, the hukou’s regulation on population mobility was largely relaxed. Population mobility, especially rural to urban migration, was substantially elevated in the 1990s when urban hous- ing market and labor market were liberalized and private sector employment grew rapidly with the in-ow of foreign direct investment (FDI) to Chinese cities. Nevertheless, residents without local urban hukou can be denied access to public schools, public health care, public pensions, and unemployment bene+ts in the city.

Figure 3. Time Trends in Per Capita Carbon Dioxide Emissions

Source: World Development Indicators, World Bank 2012 (http://databank.worldbank.org).

0

1

2

3

4

5

6

1960 1970 1980 1990 2000 2010

(Metric tons per capita)

World China India

http://databank.worldbank.org
Journal of Economic Literature, Vol. LI (September 2013)742

Though hukou status is no longer a constraint for +nding a job in cities, it remains an impor- tant tool for rationing access to local public services such as school, healthcare and social security bene+ts. Based on a cross Chinese city real estate price regression, Zheng et al. (forthcoming) document that in those cities where the hukou permit is a binding entry constraint that local quality of life factors are not capitalized to the same extent into real estate prices relative to other cities featuring lower cross-city migration costs.12

3.2 Urban Industrial Production and Pollution Growth

China’s phenomenal industrial growth over the last twenty years is well docu- mented. For example, 40 percent of the world’s clothes are “Made in China.”13 Most trade studies focus on how China’s exports impact consumer prices, labor markets, and +rm structure in importing nations (Broda, Leibtag, and Weinstein 2009; Feenstra and Wei 2009; Bloom, Draca, and Van Reenen 2011). In contrast, our focus is on how such industrial production affects China’s pollution levels. China’s Energy Statistics Yearbooks show that in 2010 the industrial sector was responsible for 89.1 percent of total energy consumption (end use) while the residential sector only consumed the left 10.9 percent. Industrial production con- tributes emissions to the air and water. Two other major industrial emissions are lead and mercury, and public health researchers have documented the health consequences of

12 Zheng et al. (forthcoming) estimate a cross-city hedonic real estate regression using data for 86 Chinese cities and include interactions between a vector of city quality of life attributes such as average air pollution and a dummy variable that equals one if the city has a binding hukou permit system. Since migrants cannot easily enter cities with binding hukou, the spatial equilibrium is such that there can be cities with high quality of life without very high home prices because external migrants cannot easily move to such cities.

13 China National Textile and Apparel Council (2011).

exposure to these toxics (Ratcliffe, Swanson, and Fischer 1996).

The geographic concentration of manu- facturing and mining activities has signi+cant local environmental consequences. During the 1970s, black smoke from stacks became the characteristic of Chinese industrial cities; in later years, many southern cities began to suffer from extremely high levels of acid rain pollution (He, Huo, and Zhang 2002). Ebenstein (2012) +nds that industrial activ- ity has led to a severe deterioration in water quality in China’s lakes and rivers. He esti- mates that a deterioration of water quality by a single grade (on a six-grade scale) increases the digestive cancer death rate by 9.7 per- cent. His econometric strategy contrasts OLS estimates of site speci+c death rates on local water quality with IV estimates where he instruments for a site’s water quality using variation in precipitation across the sites and variation in the distance from the site to the nearest river’s headwaters.

Cai, Chen, and Qing (2012) document cross-boundary river pollution effects in China caused by seven main industries: agricultural products and byproducts, textile, garments manufacturing, pulp and paper, petroleum and nuclear fuel processing, chemical indus- try, and nonferrous metals smelting and press- ing. They +nd qualitatively similar results as Sigman (2002). Both water pollution studies document that polluting sources are located at political boundaries so that the social costs of industrial activity are borne by downstream adjacent political units. Hong Kong’s resi- dents suffer from cross-boundary spillovers of industrial smoke from nearby manufactur- ing regions such as Zhaoqing, Qingyuan, and Heyuan in Guangdong Province (Zheng et al. forthcoming).

Rare earths mining and processing is a widely publicized example of a pro+table industrial activity with signi+cant environ- mental consequences. Rare earth is a key component in “green products” such as

743Zheng and Kahn: Understanding China’s Urban Pollution Dynamics

hybrid vehicles, wind turbines, and solar pan- els but extracting them causes major environ- mental problems in the areas where such rare metals can be found such as Baotou city in Inner Mongolia and Liangzhou in Sichuan.14

The links between industrial production and local pollution are well known from U.S. experience. Consider the rise of the U.S. steel industry and its impact on ambient pollution levels in Pittsburgh, and Gary, Indiana, or oil re+ning activity in Richmond, California. Empirical research has documented how the scale of urban industrial activity contributed to signi+cant levels of ambient air pollution (Kahn 1999; Chay and Greenstone 2005). Major U.S. cities have been deindustrializing as improvements in transportation, cheaper land and differential regulatory standards have all pushed dirty manufacturing outside of these cities towards the areas that are not pro-union and have low electricity prices (Henderson 1997; Holmes 1998; Becker and Henderson 2000; Greenstone 2002; Kahn and Mansur 2013). Such deindustrialization has contributed signi+cantly to recent gains in local air and water quality.

In China, manufacturing activity is mov- ing to the second and third tier cities due to the increasing costs of labor, land, and stricter environmental regulations in large cities. Overall, the more developed cities in the coastal region have the most stringent environmental regulations (Van Rooij and Lo 2010). Compared to the United States, national and urban policies play an even larger role in shaping China’s economic geography. We will discuss this in section 4.

A literature has examined the role that foreign direct investment (FDI) plays in determining China’s industrial pollution levels (Dean, Lovely, and Wang 2009). In

14 Henderson (1988) stresses the importance of natu- ral resource endowments as driving a city’s industrial base. For example, in China, a city such as Lanzhou specializes in primary metals.

China in the 1980s, attracting FDI meant more factories and thus more pollution. Conversely, it is possible that FDI provides the +nancial capital and access to developed nations’ technologies that allows cleaner factories to be built. The existing litera- ture has generated mixed +ndings on the FDI -ows’ environmental consequences. Wang and Jin (2007) +nd that foreign +rms exhibit better environmental performance than state-owned and privately owned +rms because the foreign +rms use cleaner tech- nology and are more energy ef+cient. Zheng, Kahn, and Liu (2010) use data across thirty- +ve major Chinese cities over the years 2003 to 2006 and report a negative correlation between a city’s FDI in-ows and its ambient air pollution level. Using data from 1990s and early 2000s, He (2006, 2009) +nds evi- dence that pollution and FDI are positively correlated. One possible explanation for these facts is that, up until the early 2000s, FDI increased the scale of industrial activity in China’s cities and this increased pollution, while in recent years FDI has improved the techniques used in China’s industrial sec- tor such that pollution per unit of industrial production has declined.

Two other key determinants of a given manufacturing plant’s pollution production are its management quality and its owner- ship status. A promising trend in China is the rising quality of its universities and the professionalization of the managerial class as more MBAs graduate from elite universities. If energy prices rise in China, then there will be a premium for managers who +gure out strategies for economizing on energy con- sumption. Bloom, Draca, and Van Reenen (2011) document a negative correlation between manager quality and energy inten- sity in the United Kingdom. This sugges- tive result has powerful implications for the future emissions of Chinese industry.

A special feature of Chinese manufactur- ers is that many of the major traditionally

Journal of Economic Literature, Vol. LI (September 2013)744

dirty industries are state owned enterprises (SOEs).15 SOE managers are regarded as politicians in various layers in China’s gov- ernment system. This ownership structure implies that there are very close connections between SOEs and the governments they are owned by. They have access to cheap loans and land. Whether SOE managers have strong incentives for reducing pollution hinges on their promotion criteria. The old criteria usu- ally emphasized short-term targets such as output and pro+t (Lee 2009), thus they had little incentive to engage in precautionary investment that lowers future environmental risk. On the other hand, SOE managers are aware that the government now also judges them on a company’s “green performance,” and environmental disasters will destroy their political career.16 This provides an incentive for them to clean up pollution because it is relatively easy for governments to monitor SOEs’ environmental performance.

3.3 Driving and Urban Form

The growth of private car ownership in China has been an international news topic. The overall car ownership rate in China is still not high (18 percent of households own cars in 2010), but with vast road construction investments and rising per capita income, an unprecedented increase in private vehicle ownership has taken place in Chinese cities over the last decade. The number of private cars in Chinese cities increased from 6.25 mil- lion in 2000 to 73.27 million in 2011, with an average annual growth rate of 22.76 percent

15 The SOE’s sales share was 34.5 percent in 1998 and it fell to 9 percent in 2007. Many of the small and less ef+- cient plants have been closed or merged during the SOE reform in late 1990s and early 2000s.

16 According to the requirement by the Organization Department of China Communist Party (CCP), which is in charge of the appointments of government of+cials and SOE senior managers, environmental performance (measured by the number of pollution accidents, or pol- lutant indicators reported by environmental authorities) is included in the promotion criteria for SOE managers.

(see +gure 4). In Beijing in 2010, there were 900,000 civilian registered vehicles.17

Both cross-national studies such as Dargay, Gately, and Sommer (2007), Ingram and Liu (1999) and within-China studies such as Zheng, Wang, Glaeser and Kahn (2011) have documented that richer people are more likely to own vehicles. In the case of the United States, Goldberg (1998) jointly models the vehicle purchase and vehicle utilization decision. Similar estimates using micro data from China would be quite use- ful for predicting how gasoline consumption will evolve over time and for predicting what types of vehicles will be intensively driven. One example in this research line is Anas, Timilsina, and Zheng’s study (2009) that employs a nested multinomial logit model of car ownership and personal travel in Beijing circa 2005 to compare the effectiveness of different policy instruments aiming to reduce traf+c congestion and carbon dioxide emissions. They +nd that a congestion toll is more ef+cient than a fuel tax in reducing traf+c congestion, whereas a fuel tax is more effective as a policy instrument for reduc- ing gasoline consumption and emissions. Such +ndings are crucial for predicting how changes in Chinese public policies directed to increase the cost of driving will affect the pollution produced by the vehicle sector.

The full price of driving depends on the price of the vehicle, gasoline, insurance and parking and the time required for a trip. The price of domestic-produced vehicles in China is roughly $16,000–$24,000 U.S. dol- lars or about 1.8–2.5 times of annual house- hold income. There are large tariffs applied to imported new cars from other countries such as Japan and Europe. In China, many carmakers are joint ventures with interna- tional companies. Examples include Yiqi– Volkswagen and Guangzhou–Honda. Before

17 China Vehicle Technology Research Center and the China Association of Automobile Manufacturers (2012).

745Zheng and Kahn: Understanding China’s Urban Pollution Dynamics

2005, gas prices in China had been much lower than that in the West, but the price has been rising in recent years. Today, the gas price in China is roughly 30 percent higher than that in the United States, but it is still lower than that in Japan, South Korea, the United Kingdom, and France (International Energy Agency 2011). Parry and Small (2005) argue that the optimal gas tax (to min- imize Pigouvian externalities) is greater than the U.S. tax and less than Europe’s gasoline tax. The share of various taxes (added-value tax, consumption tax, urban maintenance and construction tax, education surcharge, etc.) in the total price of a gallon of gasoline is about 45 percent in China, which is about twice of that in the United States but is lower than that in the Europe.18

The local air pollution implications of more vehicles driving more miles hinges on

18 See: http://news.163.com/special/reviews/youjiagaige. html.

the emissions technology bundled into the vehicle. Regulation determines new vehicles’ emissions per mile. China has implemented four versions of its emissions standards, which are equivalent to the Euro I, II, III, and IV standards respectively (Song, He, and Lei 2012).19 Beijing introduced the new National Standard V (equivalent to Euro V) in 2012. All vehicles are required to have an inspec- tion every two years. Such inspections require investments in emissions control technology if the vehicle fails the smog test. Perhaps surprisingly, vehicle manufacturers in China have not lobbied against these regulations. Instead, they express concern that gasoline is not clean enough to comply with the new

19 They are National Standard I (initiated in 1999 for light vehicles), II (2004), III (2007), and IV (2008, but only applied in selected cities). For each standard, the imple- mentation date for heavy vehicles was always later than that for light vehicles. The implementation date of each standard version was always 2–3 years earlier in large cities such as Beijing and Shanghai.

Figure 4. The Number of Private Cars in Chinese Cities

Source: China Automotive Industry Yearbook 2010.

Trend in the number of private cars in Chinese cities

0

10

20

30

40

50

60

70

80

1985 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011

(Million)

Number of private cars

http://news.163.com/special/reviews/youjiagaige.html
http://news.163.com/special/reviews/youjiagaige.html
Journal of Economic Literature, Vol. LI (September 2013)746

standard, so that the buyers are not willing to pay a price premium for cars that have high technology emission controls installed. We know of no research estimating the bene+ts and costs of vehicle environmental regulation compliance in China. Greenstone and Hanna (2011) evaluate how India’s introduction of air pollution regulation has affected air qual- ity dynamics. In the late 1990s, the air quality in those Indian cities that required catalytic converters for new vehicles jumped sharply and death rates in these cities fell over the medium term. Studies using U.S. data show that regulation’s impact on reducing ambi- ent air pollution should rise with time as the share of pre-regulation cohort vehicles on the roads declines (Kahn 1996; Kahn and Schwartz 2008).

Technological advance and the diffusion of existing technologies could break the link between private motorization growth and increased pollution (Acemoglu et al. 2012). After all, if electric vehicles replaced gasoline vehicles and if the electricity is generated using renewable energy sources then miles driven would not lead to large amounts of local air pollution or green- house gases. Such technological progress would shift the classic EKC curve down so that less pollution is produced during the early phase of economic growth (Dasgupta et al. 2002).

Households simultaneously choose their residential location and commuting mode. In Beijing, 48 percent of all jobs were located within three miles from the city center in the year of 2004 (calculated by the authors using 2004 Economic Census data). In the United States, New York City is the only metropoli- tan area with such job concentration down- town (Glaeser and Kahn 2001, 2004). People are much more likely to commute using pub- lic transit when they work in the center city (Baum-Snow and Kahn 2005).

Zheng, Fu, and Liu (2006) +nd that the rich live downtown in major Chinese cities.

The rich’s willingness-to-pay for downtown locations is signi+cantly higher than that of the poor, due to the fact that most job oppor- tunities and high-quality public services (such as schools, hospitals) are concentrated in the central city. This +nding is similar to the work documenting that the rich tend to live close to the city center in leading European cities such as Paris (Brueckner, Thisse, and Zenou 1999). Population and employment are both decentralizing in Beijing. The population den- sity gradient with respective to the distance to CBD was –0.156 in 2000, and it shrank to –0.123 in 2010. The job density gradients were –0.204 in 2001 and –0.159 in 2004.20

Given that private vehicles are signi+cant contributors to local air pollution, the intro- duction of subways and other high quality public transit can sharply improve urban environmental performance. In an analysis of the impact of Taiwan’s recent investments in subways, Chen and Whalley (2012) con- clude that the introduction of rapid public transit can offer signi+cant public health bene+ts if people substitute from cars and buses to traveling by subway.

China is in the midst of a major subway construction period. In 1980, only Beijing had two subway lines. In 2000, three cities had subways. In 2010, twelve cities have subways and another sixteen cities have lines under construction. In the case of Beijing, the construction of new subways has attracted private developers and new res- taurants to locate in close proximity and this reinforces the desire to live close to public transit (Zheng and Kahn 2013).

20 We estimate the following regression to recover the job density gradient parameter a1: log(density) = a0 + a1 % distance to CBD + u. Density is measured in ten thousand people (or jobs) per square kilometer. Distance to CBD is measured in kilometers. The density gradient measures the decrease of population/job density (in percentage) as the distance to CBD increases by 1 kilometer. The data for population density comes from 2000 and 2010 Population Census. The data sources for the employment density are Basic Unit Census 2001 and 2004 Economic Census.

747Zheng and Kahn: Understanding China’s Urban Pollution Dynamics

Transportation investment in suburban highways encourages suburbanization.21 Research using U.S. data has documented the causal role that new highways have played in causing suburbanization in the United States (Baum-Snow 2007). Baum- Snow et al. (2012) study how China’s urban form has been affected by its highway invest- ment. They report evidence that urban compactness is reduced by radial and ring road construction, but enhanced by public transport.

Common sense suggests that suburbaniza- tion reduces public transit use and increases the demand for private vehicle driving. Given that households are not randomly assigned to live in suburban locations, whether the asso- ciation between suburbanization and miles driven is a selection effect or a treatment effect remains an open question. If subur- banization causes increased driving, then this trend contributes to local air emissions and greenhouse gas production (Glaeser and Kahn 2010).

A unique factor affecting Chinese cities’ ability to suburbanize is the price of land at the urban fringe. Given the absence of a property tax in China and the fact that many cities rely on land sales for a large share of their annual revenues (around 60 percent), city mayors have a strong incentive to acquire rural land. We will return to this point below in the government section.

3.4 Fossil Fuel Consumption for Power Generation and Winter Heating

There has been an incredible growth of demand for electricity in the entire developing world and in China (Wolfram, Shelef, and Gertler 2012). China’s total

21 Unlike in the United States, “-ight from blight” has not been mentioned as a cause of China’s suburbanization (Mieszkowski and Mills 1993). Neither center city crime nor race relations have arisen as causal factors encouraging Chinese suburbanization.

electricity consumption in 2010 was 3.11 times that in 2000. In 2010, the average urban household owned 1.22 air condition- ers compared to 0.31 in 2000.22

This sharp rise in electricity demand poses local environmental challenges when such power is supplied using nearby power plants that rely on fossil fuels such as coal. Coal plays a vital role in electricity genera- tion worldwide. Coal-+red power plants cur- rently fuel 41 percent of global electricity, and China generates 79 percent of its elec- tricity using coal. 23

Coal burning releases large quantities of polycyclic aromatic hydrocarbons (PAHs) and other pollutants. PAHs are reproduc- tive and developmental toxicants, mutagens, and carcinogens.24 Muller, Mendelsohn, and Nordhaus (2011) estimate that coal +red power plants are responsible for 25 percent of total U.S. industrial pollution damage or $53 billion dollars of environmental damage each year. Tang et al. (2008) exploit a natural experiment in which a coal +red power plant in Tongliang, Chongqing of China was shut down in May 2004 by examining health out- comes for two identical prospective cohorts of nonsmoking women and their newborns in 2002 (before shutdown) and 2005 (after shutdown). Their +ndings indicate that neurobehavioral development in Tongliang

22 National Bureau of Statistics of China (2012b). 2 3 h t t p : / / w w w. w o r l d c o a l . o r g / c o a l / u s e s - o f - c o a l /

coal-electricity/. 24 “Power plants also emit low levels of uranium, tho-

rium, and other radioactive elements as well as mercury, and other heavy metals. These toxic pollutants have been associated with serious health problems including cognitive impairment, mental retardation, autism and blindness. . . . Power plants also generate immense quantities of ash and other residues. When fossil fuels are burned the noncom- bustible portion of the fuel is left behind along with resi- dues from dust-collecting systems, sulfur dioxide scrubbers and other emissions abatement equipment. These residues consist mostly of silicon, aluminum, and iron, but also con- tain lead, cadmium, arsenic, selenium, and mercury. Many plants land+ll these residues on site. If managed improp- erly, particles can be picked up by wind and transported locally or enter drinking water supplies” (Davis 2011).

http://www.worldcoal.org/coal/uses-of-coal/coal-electricity/
http://www.worldcoal.org/coal/uses-of-coal/coal
Journal of Economic Literature, Vol. LI (September 2013)748

children bene+ted by the elimination of PAH exposure from the coal-burning plant. A public health literature has examined how the geography of China’s coal +red power plants affects public health (Zhou et al. 2006). The total damage is larger if more people live close to a large coal +red power plant with high emissions. Based on U.S. data, Davis (2011) +nds that there is a 5 per- cent real estate price discount for homes located within two miles of a new fossil fuel +red power plant.

Based on our calculations using geograph- ical data on the location of China’s power plants, 46 percent of coal-fueled power plants are built around the median- and large-size cities with +ve million people or above. The good news is that China has invested exten- sively in sulfur dioxide scrubbers since 2006, and new policies were enacted in 2007 to increase the likelihood that installed scrub- bers actually operate (Xu, Williams, and Socolow 2009).25 Electric utilities are point sources that export pollution to nearby cit- ies. In the cases of water pollution (Cai, Chen, and Qing 2012; Sigman 2002), ambi- ent air pollution in the United States (Bayer, Keohane, and Timmins 2009) and ambient air pollution China (Zheng et al. forthcom- ing), empirical studies have documented the importance of accounting for cross-boundary pollution -ows.

Millions of Chinese urbanites rely on coal for winter heating. The United States and

25 Xu, Williams, and Socolow (2009) shows that China installed over 100 GWe of SO2 scrubbers in coal power plants in 2006 and again in 2007. The share of coal power plants with SO2 scrubbers increased from 10 percent to 48 percent from 2005 to the end of 2007. Under current regu- latory policies, all coal power plants with SO2 scrubbers must install continuous monitoring systems and transfer real-time data to the government. Second, electric utilities are incentivized to install scrubbers. For power plants that operate at 100 percent capacity, they can sell electricity to the grid at a $2.0/MWh price premium compared to a plant without scrubbers, but the premium falls to zero at an 80 percent operation rate.

Europe followed this strategy in the mid- twentieth century (Clay and Troesken 2010). There were major environmental costs of relying on this cheap but dirty fuel. Today, China is facing similar issues. Winter heating is subsidized for homes and of+ces in north- ern China above the Huai River and Qinling Mountains where the average January tem- perature is roughly 0° Celsius. This sector creates high level of emissions because heat- ing’s main energy source is coal (Almond et al. 2009). Accordingly, during the winter heating season, the contributions from coal combustion and biomass aerosol to PM2.5 mass increased in northern cities (Zheng et al. 2005).

Roughly 15 percent of China’s elec- tricity is generated using hydropower. Environmentalists have been deeply con- cerned about the consequences of building the Three Gorges Dam along the Yangtze River. While hydropower is “green” in terms of greenhouse gas and local air pollution emissions, such dam construction has impli- cations for local communities and down- stream areas (Du-o and Pande 2007).

China’s central and local govern- ments are also investing in promoting the “green economy” because this is viewed as a promising new export market (Boyd 2012). Anticipating the rising international demand for electricity and the desire to reduce greenhouse gas emissions associ- ated with power generation, developing nations such as China and India are play- ing a leading role in increasing their pro- duction capacity for producing renewable power equipment such as wind turbines and solar panels. Sawhney and Kahn (2012) document the role of China and India as growing players in the nascent green econ- omy supply chains for renewable power equipment. Those types of equipment are both exported globally and are also sold in the domestic market. China is now the world’s largest exporter of solar panels, with

749Zheng and Kahn: Understanding China’s Urban Pollution Dynamics

over 40 percent of global market share.26 On the other hand, the growth of China’s wind turbine production has focused in the domestic market, and the export share is only 1.2 percent in terms of total installed capacity in 2011.27

3.5 Natural Resource Extraction and GHG Emissions

Throughout this section, we have focused on how China’s urban growth affects local environmental quality but many externalities feature degradation of regional and global public goods. For example, the creation of greenhouse gas emissions (see +gure 3) is a global public bad.

Rising demand for meat, electricity, wood, and +sh takes place as a direct consequence of economic growth. An active research +eld on sustainability seeks to measure “ecologi- cal footprint.” It is a construct that translates total annual resource consumption into a physical land required to create and offset those -ows. Wackernagel et al. (2002) docu- ment the sharp rise in the earth’s ecological footprint during a time of ongoing growth in China and other developing nations.

Consider the example of shark +n soup. If this delicacy is increasingly in demand in a richer China, then there can be an over- shoot of extracting sharks from the world’s oceans.28 Diamond (2006) argues that an unintended consequence of the develop- ing world achieving U.S. consumption lev- els could have devastating implications for our natural capital stocks especially when such capital is public property. In this case,

26 See http://yaleglobal.yale.edu/content/chinas-green- ambition-us-sees-red.

27 Li et al. (2012). 28 Taylor (2011) presents an economic model of bison

extraction and extinction. He emphasizes that extinction’s likelihood hinges on the price for buffalo products that was largely invariant to changes in supply; open access condi- tions with no regulation of the buffalo kill; and, +nally, a newly invented tanning process that made buffalo hides into valuable commercial leather.

economic growth fueled by urbanization exacerbates the classic tragedy of the com- mons problems. An optimist would posit that an awareness of the challenge caused by growth creates a push to change the prop- erty rights regime to encourage conservation (Smith et al. 2010).

In a growing economy, the only way to reduce overall emissions is to reduce emis- sions per dollar of GDP. Since greenhouse gas emissions are a global public bad, there has been great concern that no individual nation has an incentive to reduce its own emissions. National level panel studies of carbon emissions such as Schmalensee, Stoker, and Judson (1998) conclude that the marginal effect of income on increasing carbon dioxide emissions declines with eco- nomic development. China has emerged as the largest energy consumer and GHG emit- ter in the world (Netherlands Environmental Assessment Agency 2012). Auffhammer and Carson (2008) use panel data for China’s twenty-+ve provinces in twenty years and proxy for GHG using waste gas emissions. Their forecasting model yields a pessimistic prediction of considerable expected growth in China’s GHG emissions. The good news is that the Chinese government’s policy response to climate change has also shifted markedly. At the Copenhagen Climate Summit in 2009, China pledged to achieve a carbon intensity reduction of 40–45 percent by 2020 (Department of Climate Change, NDRC, 2010). In the current 12th Five Years Plan (FYP), China is targeting a 17 percent reduction in CO2 intensity (tons of carbon dioxide emitted per unit of GDP output) between 2011 and 2015 (Government of the People’s Republic of China 2011, National Development and Reform Commission 2007). This means that, on average, energy intensity (energy consumption per unit of GDP output) should decline by an annual rate of about 3 percent–3.5 percent between 2011 and 2015. In the year 2011, there was a

http://yaleglobal.yale.edu/content/chinas-green-ambition-us-sees-red
Journal of Economic Literature, Vol. LI (September 2013)750

2.01 percent reduction29 and this is expected to decline further by 5 percent in 2012.30 Whether these targets can be achieved remains an open question.

3.6 Summary

In the previous subsections, we have sur- veyed recent work discussing the environ- mental impact of population growth, the transportation, industrial, household and power generation sectors in China. For each of these sectors, pollution production is an unintended byproduct of economic growth. The extent of the urban environ- mental impact hinges on each sector’s scale, composition and technique effects, as well as the geography of where the activity is concentrated. The common factor linking these spatial externalities is the tragedy of the commons. While hundreds of millions of Chinese urbanites suffer from the pollution’s consequences, the powerful central and local governments have been slow to address these issues. Until recently, such of+cials have focused on economic growth maximi- zation and the scale effects associated with growth only exacerbate the externalities we have discussed in this section. As we will dis- cuss below, there are reasons to believe that a transition is taking place in China’s environ- mental governance such that more effort will be invested in protecting the public’s health.

4. Rising Demand for “Green Cities” in Chinese Cities

In Chinese cities, the new cohorts of urbanites are richer and better educated. This section explores the association between household income and education in increas- ing the demand to live in “green cities.”

2 9 h t t p : / / w w w. s t a t s . g o v. c n / e n g l i s h / p r e s s r e l e a s e / t20120817_402828531.htm.

30 http://www.wantchinatimes.com/news-sublcass-cnt. aspx?id=20121205000083&cid=1105.

Such cities offer direct aesthetic bene+ts and “greenness” also represents an input in the production of health capital. We +rst discuss the driving forces behind the rising demand for environmental quality in China, and then survey recent evidence of this rising demand. We then discuss the implications of this trend for home owners and the poor.

4.1 Rising Demand for Risk Reduction

Death rates in China’s cities are declin- ing due to improvements in medical care, and better diets. The infant mortality rate had decreased sharply from 1.73 percent in 1991 to 0.58 percent in 2010 in Chinese cit- ies. Life expectancy at birth (in years) had improved from 66 to 73.3 during this thirty- year period. As death risk from other diseases declines, the bene+ts of pollution control increases (Dow, Philipson, and Sala-i-Martin 1999; Murphy and Topel 2006).

Educational attainment in major Chinese cities is rising sharply over time. In year 2000, 11.0 percent of people above 25 years old had college or above degrees in Chinese cities, while in 2010 this share increased to 21.2 percent.31 Education’s role in produc- ing health capital remains an active research topic (Conti, Heckman, and Urzua 2010). Urban researchers such as Moretti (2004) and Glaeser (1998, 2011) have documented the social bene+ts of highly educated cities. More educated people are more likely to sup- port environmental protection (Kahn 2002, Kahn and Matsusaka 1997). One explanation for this fact is that education may make us more patient and future focused (Becker and Mulligan 1997).

There are clear complementarities between human capital and health capital (Currie 2011). In a Becker household pro- duction function approach, altruistic parents who seek to develop the human capital of

31 Statistical analysis based on the data of Population Census of China in 2000 and 2010.

http://www.stats.gov.cn/english/pressrelease/t20120817_402828531.htm
http://www.stats.gov.cn/english/pressrelease/t20120817_402828531.htm
http://www.wantchinatimes.com/news-sublcass-cnt.aspx?id=20121205000083&cid=1105.
http://www.wantchinatimes.com/news-sublcass-cnt.aspx?id=20121205000083&cid=1105
751Zheng and Kahn: Understanding China’s Urban Pollution Dynamics

their children are likely to seek out safe and clean communities in order to reduce risks to health capital. Skill is a key determinant of urban earnings in both developed and devel- oping nations (Glaeser 2011; Moretti 2012). A major research effort in human capital research today is to improve our understand- ing of skill formation over the life-cycle. Heckman’s (2007) work has emphasized the importance of early life investments in children and has posited a dynamic comple- mentarity model such that learning begets learning and skill begets skill. Pollution expo- sure at early ages lowers the likelihood of later life development (Currie 2011; Currie and Almond 2011; Currie and Schmieder 2009; Currie and Walker 2011). Recent cohort studies such as Almond (2006), and Almond, Edlund, and Palme (2009) have documented the role that early life events such as expo- sure to in-uenza or exposure to Chernobyl nuclear radiation play in long-term human development. Reyes (2008) documents the impact of early lead exposure on the propen- sity of young adults to have attention de+cit disorder and lower IQs.

Chinese urbanites are aware of the risks they face due to pollution exposure and risks from day to day products such as smoking, milk, and the food supply. Richer people are willing to pay more to avoid risk. U.S. research on measuring the value of a statis- tical life has documented that it rises faster over time than per capita GNP (Costa and Kahn 2004).32 If this +nding also applies in Chinese cities, the population’s demand for risk reduction will rise sharply over time. Kenkel, Lillard, and Liu (2009) documents that from the 1950s to the 1990s, smoking rates in China generally increased, while

32 Costa and Kahn (2004) combine IPUMS Census data with data from the BLS with respect to industrial fatality risk. Based on repeat hedonic wage OLS regressions from 1940, 1960, 1970, 1980, 1990, and 2000, they +nd that the value of a statistical life has increased by a larger percent- age than national GNP growth.

starting in the mid-1990s smoking rates for all education groups began to drop. Chinese urban households are increasingly investing in self-protection as revealed by paying a higher price premium for imports from the United States and Hong Kong such as imported baby milk that is perceived to be safer.

Given the importance of early childhood development, pollution is a crucial public policy issue (Chay and Greenstone 2003, Currie and Neidell 2005). China’s unique one child policy creates the incentive for house- holds to invest more time and resources in their one child (Becker and Lewis 1974).33 After China’s tainted baby milk scandal (Yang et al. 2009), more than 60 percent of the baby milk Chinese parents bought were from over- sea places, and they paid a price premium of roughly 33 percent.34 Such a price premium means that urbanites are sacri+cing other consumption in order to produce healthy chil- dren. Urban parents also put political pressure on the government to reduce risk. While the concept of the “median voter” may not apply in China, below we will argue that politicians are responsive to households’ desires.

4.2 Rising Demand for Information and Accountability

A recent U.S. literature has documented that access to high quality, up to date infor- mation on local pollution challenges leads to self protection investments that break the link between ambient pollution and exposure. Jin and Leslie (2003) +nd that the provision of restaurant public hygiene report cards improved California public health by causing

33 This “one child policy” is more binding in the pub- lic sector (governments, public institutions and SOEs), but less binding in the private sector (private and foreign invested companies). Households that violate this rule must pay a considerably high +ne. In addition, the parents will be +red if they are employed in the public sector.

34 For more background information, see http://top- ics.nytimes.com/top/reference/timestopics/subjects/m/ melamine/index.html.

http://topics.nytimes.com/top/reference/timestopics/subjects/m/melamine/index.html
http://topics.nytimes.com/top/reference/timestopics/subjects/m/melamine/index.html
http://topics.nytimes.com/top/reference/timestopics/subjects/m/melamine/index.html
Journal of Economic Literature, Vol. LI (September 2013)752

changes in behaviors on both the demand and supply sides. Neidell (2009) exploits a regression discontinuity design to study how Smog Alerts affect household behavior and documents statistically signi+cant evidence that attendance drops at the Los Angeles Zoo and Dodger Stadium (both are located in polluted parts of Los Angeles) when a Smog Alert is announced.35 For evidence on the persistence of these behavioral responses, see Graff-Zivin and Neidell (2009).

Recent research set in India, Brazil, Indonesia, and China highlights the power of the media and information disclosure to mitigate classic principal–agent problems and to nudge government of+cials to sup- ply public goods. Besley and Burgess (2002) report evidence from India that in those states where there is greater literacy, media coverage disciplines politicians to act in the public’s interest and minimizes agency con- cerns as politicians are more likely to be held accountable if they do not deliver public goods. Ferraz and Finan (2008) document that Brazilian voters responded to informa- tion about their municipalities’ expenditures of federally transferred funds. They show that the release of the audit outcomes had a signi+cant impact on incumbents’ electoral performance, and that these effects were more pronounced in municipalities where

35 Moretti and Neidell (2011) recognize that the popu- lation is likely to invest in self protection when they antici- pate high levels of ambient pollution. This suggests that the correlation of sickness and unanticipated pollution will be higher than the correlation between sickness and antici- pated pollution. In the former case, the population is less able to self protect against a threat they do not anticipate. Moretti and Neidell exploit cargo boat arrivals into the Port of Los Angeles as a plausibly exogenous instrumental variable that affects local ambient air pollution. They docu- ment that in a regression of health on pollution exposure that the IV results are larger than the OLS results, which provides strong evidence that the affected population is taking self protective steps to minimize the impact of expected high pollution levels. Intuitively, their IV strategy captures surprise shocks to pollution for which the public is less able to protect against. In this case, pollution causes more health damage.

local radio was present to divulge the infor- mation. Pargal and Wheeler (1996) pres- ent evidence from Indonesia highlighting the role that information disclosure plays in nudging polluting +rms to change their behavior. Jin, Wang, and Wheeler (2010) report that China’s Green Watch ratings pro- gram implemented in four cities of Jiangsu province (Huaian, Wuxi, Yangzhou, and Zhenjiang) was effective in reducing +rms’ emissions.36 The percentage of +rms with positive ratings increased from 75 percent in 1999 to 85 percent in 2000 while the per- centage of extremely noncompliant +rms declined from 11 percent in 1999 to 2 per- cent in 2000. This program has been scaled up over time.

In the past, the Chinese state monopolized news coverage and this meant that environ- mental disasters caused by industrial negli- gence would not generate the same attention as if a similar event took place in the United States or Europe. In the absence of media attention and public outrage, of+cials would have less of an incentive to nudge polluting industries to invest in costly precautions.37 As the new urban cohorts in China become richer and more educated, they have a greater willingness to pay to avoid risk, thus their demand for information and political accountability is likely to rise.

The rise of demand for environmental information creates incentives for the media to cover such stories (Gentzkow and Shapiro

36 The Green Watch program seeks to reduce indus- trial pollution by rating the environmental performance (air, water, and toxic pollutants) of approximately 2,500 polluting enterprises in thirteen municipalities of Jiangsu Province. The ratings are publicly released. Those “green” enterprises with better ratings are able to access lower- interest loans from banks. “Brown” enterprises will receive local governments’ strict supervision and face signi+cant pressure from the public.

37 If individuals (the victims) cannot identify the pollu- tion source or be aware of the pollution then they cannot take ex ante precautions or seek restitution through ex post liability suits. In such a case, the polluter faces no account- ability for its actions and this encourages malfeasance.

753Zheng and Kahn: Understanding China’s Urban Pollution Dynamics

2010). In recent years, the Chinese media has devoted much greater attention to envi- ronmental issues. For instance, the number of Google news on this topic in year 2011 is 240 times than ten years ago. With the modern media and IT technologies such as blogs, micro blogs (weibo, the Chinese ver- sion of Twitter), and mobile phone messages, the government does not have a monopoly allowing for a suppression of information any longer. 38 This should incentivize politicians to invest more effort in supplying environ- mental regulation.

Martinez-Bravo et al. (2012) !nd that the introduction of competitive election in China’s rural villages signi!cantly increases the village leaders’ accountability, and thus increase public goods expenditure and provision. Competitive elections have not been introduced into China’s urban sector. However, the power of the media and infor- mation disclosure encourages city mayors to supply environmental regulations.

One salient example is the recent Internet criticism that pushed China’s government to report on PM2.5 air pollution.

39 In fall 2011, some skepticism was raised in the media and micro blogs on the authenticity of the of!- cial Air Pollution Index (API) reported in

38 In Western countries, the media devotes ample attention to environmental disasters. Examples include the Exxon Valdez Oil Spill or the BP Oil Spill in 2010. This media attention affects interest group competition and gives pro-regulation forces the upper hand to potentially overcome the asymmetric interest group problem noted by Olson (1965).

39 Total suspended particles (TSP) measure the mass concentration of particulate matter in the air. Within TSP, PM10 stands for particles with a diameter of 10 microm- eters or less, and PM2.5 stands for those with a diameter of 2.5 micrometers or less. Particulates that are ten microm- eters or greater are !ltered and generally do not enter the lungs. Particulates smaller than ten micrometers are likely to enter the lungs. Particulate matter that is smaller than 2.5 micrometers (PM2.5) can enter into the Alveoli where gas exchange occurs. Throughout the world, ambient mon- itoring now focuses on PM10 and PM2.5. See http://www. epa.gov/ttnamti1/contmont.html and http://www.aidic.it/ aaas08/webpapers/44Lanzani.pdf.

Beijing. This Index mainly measures PM10, which stands for particles with a diameter of 10 micrometers or less.40 The most recent public concern has focused on the diver- gence between the of!cially released PM10 readings in Beijing versus the daily PM2.5 readings reported by the U.S. Embassy in Beijing. On a foggy day of Oct. 9, 2011, the U.S. Embassy’s PM2.5 reading was so high compared with the standards set by the U.S. Environmental Protection Agency that it was listed as “beyond index.” But China’s own assessment based on PM10 was merely “slightly polluted.”41 This large divergence took place several times in that month and this triggered a debate in the media and micro blogs. At !rst, the Beijing local of!- cials argued that these two readings mea- sured different sized particulates so that a direct comparison was not valid, but the public was not convinced. Later the central government stepped in and expressed the opinion that “air quality monitoring reports should be consistent with people’s real feel- ing.” It also required that all municipalities and provincial capital cities should start to monitor and report PM2.5 starting in 2012, and all prefecture-level cities should start this in 2015.

High-pro!le industrial accidents in China offer a second example of the role of mobi- lizing public opinion. In August 2011, mes- sages widely spread through micro blogs,

40 In academic circles, serious questions have been raised about the quality of this API. For instance, Wang et al. (2009) !nd that their self-measured PM level in Beijing during the 2008 Olympics Game period is corre- lated with of!cial API, but their readings are 30 percent higher. Andrews (2008) pointed out a likely systematic downward-bias around the “Blue Sky” standard (API less than or equal to 100), and also highlighted a sampling downward bias due to the dropping of monitoring stations in the more pollution concentrated traf!c areas in Beijing. These studies triggered scholars’ concerns about the mea- surement error embedded in Chinese of!cial API data.

41 For more background information, see http: / / a r t i c l e s . l a t i m e s . c o m / 2 0 1 1 / o c t / 2 9 / w o r l d / la-fg-china-air-quality-20111030.

http://www.epa.gov/ttnamti1/contmont.html
http://www.epa.gov/ttnamti1/contmont.html
http://www.aidic.it/aaas08/webpapers/44Lanzani.pdf
http://www.aidic.it/aaas08/webpapers/44Lanzani.pdf
http://articles.latimes.com/2011/oct/29/world/la-fg-china-air-quality-20111030
articles.latimes.com/2011/oct/29/world/la
Journal of Economic Literature, Vol. LI (September 2013)754

Twitter, and other Internet forums that a Dalian PX (paraxylene) chemical factory (a joint venture between the city and a private company) was at high risk to !ood the town with the highly toxic chemical. Twelve thou- sand Dalian residents organized a peaceful public protest in Dalian’s People’s Square on August 14th, demanding that the factory be immediately shut down and relocated, and that the details about the investigation into the factory should be made public. The Dalian government forbade the factory from opening.

An optimistic hypothesis is that envi- ronmental of"cials who anticipate ex post accountability for disasters will be more pro active to take regulatory actions that reduce the probability that industrial disasters occur in the "rst place. An alternative view is that media competition does not necessarily lead to greater accuracy of coverage. Gentzkow and Shapiro (2010) "nd evidence support- ing a Hotelling differentiation theory that newspapers cater their stories to appeal to their base readers. Such a theory of ideologi- cal sorting could explain why the U.S. media differs (contrast Fox News with MSNBC) in its portrayal of the challenge of climate change. More rigorous studies are needed to test these hypotheses.

4.3 Rising Demand for Green Cities: Evidence from Home Price Compensating Differentials

The urban quality of life literature emphasizes that spatial variation in wages and rents represents a compensating dif- ferential for place based local public goods (Rosen 2002). This revealed preference methodology allows scholars to identify urban households’ demand for nonmar- ket goods, including urban environmental amenities and allows researchers to rank areas’ quality of life (Albouy 2008, Albouy and Lue 2011, Gyourko, Kahn, and Tracy 1999).

Using standard revealed preference meth- ods, new research is documenting local public goods demand in China’s cities. An intraurban study by Zheng and Kahn (2008) "nds that proximity to fast public transit, clean air, high-quality schools, major uni- versities, and environmental amenities are capitalized into home prices in Beijing. An intercity study of thirty-"ve large Chinese cities by Zheng, Kahn, and Liu (2010) "nds that home prices are lower in cities with higher ambient pollution levels, and the marginal valuation for green amenities is ris- ing over time. They report evidence of an Environmental Kuznets Curve such that the richer cities are beyond the turning point at a much lower level of per capita income than has been estimated in cross-country work (Grossman and Krueger 1995).

All hedonic studies that rely on OLS potentially suffer from omitted variable bias (Gyourko and Tracy 1991, Chay and Greenstone 2005). For example, a city’s air pollution could be high in cities featuring booming industrial industries. In this case, a hedonic researcher conducting a cross-city study may estimate a positive association between air pollution and local real estate prices because air pollution is positively correlated with the positive local indus- trial demand shock. In a U.S. study, Bayer, Keohane, and Timmins (2009) introduce a potentially credible instrument. They docu- ment in the "rst stage of an IV regression that cross-boundary air pollution externality spillovers are correlated with a city’s ambient air pollution. Zheng et al. (forthcoming) use this same IV approach in China. They exploit the fact that the quantity of emissions that are imported into a city depending on the dominant wind direction and emissions from nearby cities and from the sandstorms from Inner Mongolia. Both of these exporting sources raise the “destination” city’s ambient pollution level. They "nd that on average, a 10 percent decrease in imported neighbor

755Zheng and Kahn: Understanding China’s Urban Pollution Dynamics

pollution is associated with a 0.76 percent increase in local home prices.