Regional Science and Urban Economics - Fudan University · and quantitative targets, individual interview, and qualitative assessment of capacity and potential. Therefore, the promotion

Regional Science and Urban Economics xxx (2013) xxx–xxx

REGEC-02998; No of Pages 11

Contents lists available at ScienceDirect

Regional Science and Urban Economics

j ourna l homepage: www.e lsev ie r .com/ locate / regec

Incentives for China's urban mayors to mitigate pollution externalities:The role of the central government and public environmentalism☆

Siqi Zheng a, Matthew E. Kahn b,⁎, Weizeng Sun a, Danglun Luo c

a Department of Construction Management and Hang Lung Center for Real Estate, Tsinghua University, Chinab UCLA and NBER, United Statesc Lingnan College, Sun Yat-Sen University, China

☆ We thank the editor (Dan McMillen) and two reviethank the UCLA Ziman Center for Real Estate for fundingOctober 2012 Lincoln Institute Conference honoring JohnSivitanidou Annual Research Symposium at USC. We tGreen and Armando Carbonell for useful comments. SiqiNational Science Foundation of China (No. 71322307, NoProgram for New Century Excellent Talents in UniversityUniversity Initiative Scientific Research Program for researNational Science Foundation of China (No. 70902024) for⁎ Corresponding author.E-mail addresses: [email protected] (S. Zheng), mk

[email protected] (W. Sun), [email protected] See http://www.chinadaily.com.cn/china/2013-01/14/c

background information.2 See http://www.chinafile.com/airpocalypse-now-china

or less in diameter (PM2.5) are referred to as “fine” particleshealth risks than larger particles because they can embed d

0166-0462/$ – see front matter © 2013 Elsevier B.V. All rihttp://dx.doi.org/10.1016/j.regsciurbeco.2013.09.003

Please cite this article as: Zheng, S., et al., Inceand public environmentalism, Reg. Sci. Urban

a b s t r a c t

Article history:Received 26 August 2013Accepted 2 September 2013Available online xxxx

Keywords:ChinaUrban mayorsPolitical economyPollution

China's extremely high levels of urban air, water and greenhouse gas emission levels pose local and global environ-mental challenges. China's urban leaders have substantial influence and discretion over the evolution of economicactivity that generates such externalities. This paper examines the political economy of urban leaders' incentivesto tackle pollution issues. We present evidence consistent with the hypothesis that both the central governmentand the public are placing pressure on China's urban leaders to mitigate externalities. Such “pro-green” incentivessuggest that many of China's cities could enjoy significant environmental progress in the near future.

© 2013 Elsevier B.V. All rights reserved.

1. Introduction

China's pollution challenges are well documented. Many cities inChina have extremely high air pollution levels. In early 2013, the terriblesmog haze pollution in North China caught the world's attention.1 ThePM2.5 concentration in those cities has been two, three, or even fourtimes the emergency threshold of 250 μg/m3 (and up to 40 times levelstheWHO considers healthy).2 Based on an ambient particulate concen-tration criterion of PM10, twelve of the twenty most polluted citiesin the world are located in China (World Bank, 2007). This pollutionhas mainly been caused by emissions from the heating and electricitysector (based on coal), and the industrial and transportation sectors.

As China surpassed Japan as the second largest economy in theworld at the end of 2009, China's energy consumption and electricity

wers for useful comments. We. We thank participants at theM. Quigley and the 2013 Renahank Yongheng Deng, RichardZheng and Weizeng Sun thank. 70973065 and No. 71273154),(NCET-12-0313), and Tsinghuach support. Danglun Luo thanksresearch support.

[email protected] (M.E. Kahn),(D. Luo).ontent_16115953.htm for more

-tipping-point. Particles 2.5 μmand are believed to pose greatereep in people's lungs.

ghts reserved.

ntives for China's urbanmayoEcon. (2013), http://dx.doi.or

demand have also been soaring. The nation's electricity consumptionreached roughly 4.5 trillion kilowatt hours in 2011.3 Given that 80%of China's electricity is produced by coal fired power plants this hasled to a huge increase in greenhouse gas emissions.

If China's central and local governments stepped in and mandatedcredible regulations, then pollution externalities across China's citiescould be mitigated. Environmental economists have argued based oncross-national evidence that there is a “J” curve for regulation suchthat poor nations implement none and middle income nations start toimplement such regulation which grows more intense as these nationsdevelop from being middle income to being rich (Selden and Song,1995). As China becomes one of the world's leading economies, it ispossible that a similar dynamic could play out there.

Such an optimistic, and deterministic, vision of regulatory adoptionas a function of only national per-capita income abstracts away frominstitutions and incentives as important determinants of whethergovernment officials are “up to the job” of combatting pollution. Yet,leading studies in growth economics have emphasized the fundamentalrole that institutions play in economic development (Acemoglu andRobinson, 2012).

Until recently, neither China's national government officials norlocal urban officials prioritized environmental protection. The Chinesecentral government creates a “tournament competition” among localmayors by promoting or demoting them on the basis of relative per-formance (Bo, 1996; Wu, 2010). The central government had beenfocusing on economic growthwith an emphasis on GDP as the key eval-uation criterion for local officials' performance (Chen et al., 2005; Li and

3 See: http://www.bloomberg.com/news/2011-01-28/china-s-power-demand-growth-may-slow-to-9-this-year-nea-says.html.

rs tomitigate pollution externalities: The role of the central governmentg/10.1016/j.regsciurbeco.2013.09.003

2 S. Zheng et al. / Regional Science and Urban Economics xxx (2013) xxx–xxx

Zhou, 2005) for a long time. Local officials thus sought to boosttheir local economy through attracting dirty industries, but had littleincentive to reduce energy consumption or protect the environmentin their own jurisdictions since such actions did not help their politicalcareer (Wu et al., 2013).

Such a tournament raises the possibility that the central governmentcan incentivize urban officials to devotemore attention to environmen-tal challenges. In recent years the central government has been chang-ing the performance evaluation criteria for local officials from purelyoutput-based to including more “greenness” in the performance vector(Landry, 2008). Belowwe discusswhy the central government changedits focus on GDP growth to an objective function that also includesenvironmental goals. The driving forces were both a desire to improvethe people's quality of life and a desire to establish legitimacy in thepublic's mind to help retain political power (Wang, forthcoming).

Local residents provide a second source of pressure on urbanmayors. In democracies, voters have the ability to hold elected officialsaccountable for their policy choices (Hårsman and Quigley, 2010;List and Sturm, 2006). While China's urbanites do not directly vote,they have alternative strategies for expressing their views. As the newurban cohorts become richer and more educated, they are likely tovalue safety and greenness. An educated publicwill seek outmore infor-mation about environmental threats. Recent trends that reduce the costof information acquisition, such as the rise of the Internet media, microblogs (weibo, the Chinese version of Twitter),4 instant phone messages,andmore liberated local newspapers have increased the public's aware-ness of pollution challenges. The salience of this news allows themto overcome potential free rider issues and to unite to express theirconcerns and displeasure with current urban quality of life. Since socialstability is an important target when the State evaluates local officials,local officials are keen to address their people's demand for a cleanerenvironment.

This paper uses unique city level panel data to test several predic-tions related to how a city's environmental performance influences amayor's career prospects. We also study how quality of life conditionsis associated with the public's interest in environmental issues. Ourstudy exploits cross-regional and within city variation in economicand environmental conditions to generate new facts about the causesand consequences of pollution on city leader's priorities. We hypothe-size that relative to the past, urban mayors in China now face politicalpressure from the central government and the local public who areeach demanding environmental progress. In a metaphorical sense, themayors are “sandwiched” by these two different pressure groups andthus have less discretion than they had in the recent past.

We create several new data sets including information on the pro-motion propensities and demographics of prefecture-level city mayors,and their city's industrial energy intensity and ambient particulatemat-ter (PM10) levels of 86 Chinese cities during the years 2004 to 2009.We use these data to test whether there is an association betweenenvironmental performance and an urban leader's probability of beingpromoted. We also test whether objective measures of urban residents'environmentalism are associated with environmental progress. Wepresent evidence consistent with the hypothesis that both the centralgovernment's regime shift and urban households' rising demand forgreenness are contributing to local politicians' accountability for theircity's energy and environmental performance.

This paper contributes to a nascent empirical literature on the rolethat political leadership plays in determining government prioritiesover public good provision. Jones and Olken (2005) document the rolethat national leaders play in affecting macroeconomic growth. List andSturm (2006) find that U.S. governors' environmental policy prioritieschangewhen they are restricted by term limits from remaining in office.Ferreira and Gyourko (2009) document differentials in U.S. mayor

4 The micro blog, as a nascent web application emerged in 2009, had 331 million usersby June, 2013.

Please cite this article as: Zheng, S., et al., Incentives for China's urbanmayoand public environmentalism, Reg. Sci. Urban Econ. (2013), http://dx.doi.or

policies over taxes, spending and public sector employment. Jia(2012) develops a model of politicians with career concerns makingchoices over the use of clean and dirty technologies. She exploits aunique data set identifying social networks between Chinese localgovernors and key central government officials, and concludes thatpoliticians are motivated by strong promotion incentives which pro-mote growth, regardless of its social costs.

The rest of the paper is organized in five sections. Section 2 describesthe political economy of environmental regulation in China, especiallythe role of promotion criteria, and also the green nudge from thepublic. Section 3 discusses our empirical hypotheses and data creationas we construct several unique data sets, including the energy–environmental quality and mayor promotion data by city/year, andtwo indices reflecting Chinese urbanites' concern intensity over pollu-tion. Section 4 presents the empirical equations and results. Section 5concludes.

2. Background on the central government's promotion rules and therecent emphasis of environmental goals

2.1. The evolvement of the promotion rules

China has a strong one-party central government, but hundredsof local governments act as competing enterprises. The State Councilappoints the governors of provinces, municipalities, and some majorcities (so-called “provincial-level” and “vice-provincial” cities) directly.Provincial governments appoint the governors of prefecture-level cities.How to select and reward subordinate officials is central to the effectivegovernance of every large organization. The selection and promotionprocess is performed by the upper-level CCP (China's CommunistParty) Committee's personnel department, which is a key sector in theupper-level government.5

In the past, local GDP growth was the main criterion used by upper-level governments in evaluating the performance of lower-level offi-cials' performance and deciding whether to promote them to higherpositions. Recently, sustainability and social stability are included inthe promotion criteria.

The Chinese State has established a number of notable targets for en-ergy efficiency and pollution reduction. Specific energy efficiency andpollution reduction targets were clearly set and included in the tenth,eleventh and twelfth “Five-Year Plan” (2001–2005, 2006–2010, 2011–2015 FYP, respectively). In the tenth FYP, the target was set that majorwater and air pollutants should decrease by 10% over the five-year peri-od. In the eleventh FYP, the target was that major pollutants such asCOD (Chemical Oxygen Demand) and SO2 to decrease by 10% eachyear from the 2005 level; energy consumption per unit ofGDP to declineby about 20% from the 2005 level. At the Copenhagen Climate Summitin 2009, China pledged to achieve a carbon intensity reduction of 40–45% by 2020 (Department of Climate Change, NDRC, 2010).

There are several motivations behind the Chinese central govern-ment's ambitious shift to emphasize pollution reduction and climatechange mitigation goals. First, domestic energy security concerns haverisen on the central government's agenda as a result of electricity short-ages and rapidly rising energy consumption. Second, the central govern-ment believes that the rest of the world is embracing the low-carbonenergy agenda which has created a market imperative for China to be-come a technological and economic leader in this nascent field (Boyd,2012). Third, the central governmentmay be concerned about thedirectproductivity loss and the disamenity effects caused by pollution expo-sure. Another possible explanation is that the central governmentseeks “legitimacy”with the Chinese people and also in the internationalarena, and making a commitment to pursuing environmental goals

5 This process is quite complicated, including performance evaluation with objectiveand quantitative targets, individual interview, and qualitative assessment of capacityand potential. Therefore, the promotion rule cannot be written out as a simple function.

rs tomitigate pollution externalities: The role of the central governmentg/10.1016/j.regsciurbeco.2013.09.003

7 http://news.bbc.co.uk/2/hi/asia-pacific/7195434.stm.8 In August 2011,messages werewidely spread throughmicro blogs, blogs, Twitter and

other internet forums that a PX (paraxylene) chemical factory (a joint venture betweenthe city and a private company) built in Dalian city was at high risk to flood the townwiththe highly toxic chemical. Twelve thousand Dalian residents organized a peaceful publicprotest inDalian's People's Square onAugust 14, demanding the factory to be immediatelyshut down and relocated, and that the details about the investigation into the factoryshould be made public. The Dalian government forbade the factory from opening. Seehttp://www.bbc.co.uk/news/world-asia-pacific-14520438.

3S. Zheng et al. / Regional Science and Urban Economics xxx (2013) xxx–xxx

is one way to credibly signal to both domestic constituents andinternational actors that China is an international leader and thatthe Communist Party leadership cares about its own people (Wang,forthcoming). This suggests that environmental protection is part of abroader public relations campaign to boost popular domestic supportand to reduce the risk of social instability.

Including greenness targets in the performance evaluation and pro-motion criteria of local government leaders is the State's key approachto encourage local leaders to address sustainability challenges. Sincethe 11th FYP, the State has started to mobilize local government viathe application of the target responsibility system (TRS) of energy con-servation and pollution reduction, which is a top–down policy imple-mentation mechanism based on China's prevalent top–down pressuretransfer political hierarchy (Qi, 2013). TRS involves four major steps:disaggregating targets, signing target responsibility contracts, accountingand monitoring energy consumption, and assessing target performance.For the first step, the central government disaggregated the total energyconservation and pollution reduction targets to provincial governments,and then provincial governments disaggregate their targets to municipalgovernments. The target responsibility contract is normally signedbetween the top officials of the upper and lower level governments.

Assessment and evaluation of target responsibility are a quantitativeexercise. The energy conservation target, with the decline of energy in-tensity (EI, energy consumption per unit of GDP) as themajor indicator,accounts for 40 points out of the total 100 points. The accounting systemof this EI decline indicator was set up by the State Council. The other60 points include many items, ranging from regularly reporting energyconsumption numbers to upper level governments; investing in energyconservation and pollution reduction infrastructures, to effectivelyimplementing environmental regulations. The assessment result is usedin the performance evaluation of local leaders.6

The main TRS targets are not closely linked to environmental out-comes that have significant impacts on the public's health and qualityof life. Instead, they are linked to the accounting indicators such asenergy intensity and environmental infrastructure investment. Theseaccounting indicators are more easily to be measured and collected.Credit for pollution reduction might be granted, for example, for theconstruction of a waste gas treatment plant or installation of pollutioncontrol technology in a power plant. Therefore local officials are incen-tivized to invest in environmental infrastructure and pollution controltechnology. With insufficient monitoring, there was much less focuson whether these investments are operated properly such that theyactually reduce pollution. It was reported that factories adjusted pollu-tion control equipment to report false data, treatment plants were leftidle, local governments forced emergency shutdowns of electricityto local public services to meet energy efficiency targets, and so on(Wang, forthcoming). We will test this in Section 4 below by includingdifferent indicators in the promotion equation – objective quality of lifeindicators (such as PM10) and yearbook statistics (energy intensity andannual expenditure on waste gas treatment facilities) – and comparetheir effects.

2.2. Pressure applied by the urban public for environmental progress

While the central government has set performance standards basedon criteria such as the number of pollution control facilities, the urbanpublic has different priorities. They care about clean air and water.In the past, the public faced greater information costs concerning theenvironmental challenges their city faced. The State and local govern-ments monopolized the media — newspaper, radio and television.When a one party state controls information releases it may systemati-cally choose to release information that helps it to achieve its political

6 See the requirement in “Interim Procedures for Comprehensive Assessment and Eval-uation of Local Leading Groups and Leading Cadre of CPC and Governments Embodyingthe Scientific Outlook on Development”, published in September 2006.

Please cite this article as: Zheng, S., et al., Incentives for China's urbanmayoand public environmentalism, Reg. Sci. Urban Econ. (2013), http://dx.doi.or

goals and may suppress negative information (Liu and Yang, 2009;Guan et al., 2012).

With the rise of themodernmedia and IT technologies such as blogs,micro blogs, instant phone messages, China's government has beenlosing its information disclosure monopoly. Local newspapers are alsomore liberalized. To attract readers, they report negative news such aspollution, corruption and land seize disputes. Improvement in remotesensing and cheaper pollution monitors has allowed others to measureand release China's pollution levels (Zhang et al., 2007).

Recent research set in the US, India, Brazil, and Indonesia highlightsthe power of the media and information disclosure to mitigate classicprincipal–agent problems and to nudge government officials to supplypublic goods (Gentzkow et al., 2010; Besley and Burgess, 2002; Ferrazand Finan, 2008; Pargal and Wheeler, 1996). But those studies are allconducted in democracies. We are interested in whether the risingof information transparency in Chinese cities plays a similar role. Therecent upsurge of environmental mass incidents (i.e., large-scale dem-onstration, protect or march triggered by environmental degradationor serious pollution events) provides some clues of this. In those massincidents, the modern media helps to trigger a snowball effect, andthis allows the public to co-ordinate and overcome transaction coststo unite together to pretest against pollution accidents. Examplesinclude the Xiamen PX protest in 2007,7 Dalian PX protest in 2011,8

Shifang MoCu project protest in 2012,9 and Qidong protest on thepaper mill's pollution discharge into the sea in 2012.10 The number ofmass incidents caused by pollution increased at an annual rate of 29%(Tong, 2013). Those events significantly threatened social stability,which is now another key target when evaluating local officials' per-formance. Therefore mayors are becoming more concerned about localpeople's concern about environmental quality and local quality of life.

3. Empirical hypotheses and data

3.1. Hypotheses

Based on the above discussion, we focus on testing four hypothesesrelated to the correlates of urban leaders pursuing policies that bringabout environmental progress in China:

H1. Local officials are more likely to be promoted if their city experi-ences environmental progress.

H2. Public concern over urban pollution varies across space. Thoseregions (province/city) featuring stronger demand for environmentalquality and with greater media openness have higher public concernintensity.

H3. City leaders facing recent public concern over environmental issuesput more effort in pollution mitigation. There is a positive correlationbetween mayoral attributes such as his educational attainment andenvironmental progress.

H4. City leaders facing more pressure from the public will engage ingreater energy conservation and environmental protection and this

9 http://www.nytimes.com/2012/07/05/world/asia/chinese-officials-cancel-plant-project-amid-protests.html?_r=0.10 http://www.nytimes.com/2012/07/29/world/asia/after-protests-in-qidong-china-plans-for-water-discharge-plant-are-abandoned.html.

rs tomitigate pollution externalities: The role of the central governmentg/10.1016/j.regsciurbeco.2013.09.003

Energy Intensity in 86 Cities (2009)

Fig. 1. Energy intensity in 86 Chinese cities in 2009.

11 Total suspended particles (TSP) measures the mass concentration of particulate mat-ter in the air. Within TSP, PM10 stands for particles with a diameter of 10 μm or less. Par-ticulates that are ten micrometers or greater are filtered and generally do not enter thelungs. Particulates smaller than ten micrometers are likely to enter the lungs.12 http://datacenter.mep.gov.cn/.13 The quality of China's API data has been debated. For instance, Wang et al. (2009)found that his self-measured PM level in Beijing during Olympic period is correlated withofficial API, but 30% higher. Andrews (2008) pointed out a likely systematic downward-bias around the “Blue Sky” standard (API less or equal to 100), and also highlighted a sam-pling downward bias for dropping monitoring stations in more pollution concentratedtraffic areas in Beijing. These studies triggered some concerns on themeasurement errorsusing Chinese official API data. Later studies suggest that Wang's measurement gap be-tween the self-measured data and official API data ismainly due to sampling andmethod-ological differences (Tang et al., 2009; Yao et al., 2009; Simonich, 2009). A recent paper byChen et al. (2011) uses both API and AOD data to analyze the changes before and after Bei-jing Olympic. Their study suggests that the two different data sources provide similar re-sults. In our study, we convert API index back to PM concentration data using the SEPAAPI formula. Andrews (2008) shows that this approach is reliable, especially when themain purpose is to study the cross-city variation for a large number of cities.

4 S. Zheng et al. / Regional Science and Urban Economics xxx (2013) xxx–xxx

affects the shape of the pollution versus income relationship (i.e. theEnvironmental Kuznets Curve).

3.2. Data

3.2.1. Energy/environment indicatorsWe select three energy/environment indicators. One is the

energy intensity measure, which is a key component of the TRS. Energyintensity (EI) of urban productivity is measured as “energy consumptionper GDP dollar” (ton standard coal per 10,000 RMB) by city/year (Eq. (1)).

EIit ¼EnergyitGDPit

¼

XJ

j

GDPijt � EIjt� �

GDPit

¼

XJ

j

GDPijt �EnergyjtGDPjt

!

GDPit

ð1Þ

where subscript i represents city; t represents year; and j representsindustry. This energy intensity variable reflects a city's industrial compo-sition in a given year. If a city's industries are very energy intensive thenits energy intensity will be high.

The energy consumption and GDP data are collected from the “ChinaCity Statistical Yearbook”. Fig. 1 shows the energy intensity values forthe 86 cities in our sample in 2009. The energy intensity had decreasedfor many cities during 2004 and 2009. As we do not have city levelenergy intensity data, our EI variable is calculated using the nationallevel energy intensity of each industry and creating the city's indexbased on its industrial mix in a given year. Given this definition,a city's energy intensity would decline if a mayor actively sought toreplace dirty industries (i.e. steel) with clean industries such as services.But this index does not reflect within industry technique effects such asnew clean factories opening and old dirty factories closing.

Please cite this article as: Zheng, S., et al., Incentives for China's urbanmayoand public environmentalism, Reg. Sci. Urban Econ. (2013), http://dx.doi.or

We have collected two indicators on air pollution. One is an“accounting” indicator – annual expenditure on waste gas treatmentfacilities per GDP dollar (FACILITY_EXP); and the other one is the “localquality of life” indicator – the ambient particulate concentration in theair (PM10).11 The first variable measures local governments' effort inproviding air pollution mitigation facilities, which help them to gainsome credit in the TRS. For the second measure, we first collect theAPI (air pollution index) of each city by week from the website of theMinistry of Environmental Protection, People's Republic of China,12

and then calculate the average PM10 concentration (mg/m3) by cityby year (PM10).13 Since people are more sensitive to severely polluteddays, we also construct the variable of PM10p75 which stands for the75 percentile value of PM10 concentration by city/year. Fig. 2 shows

rs tomitigate pollution externalities: The role of the central governmentg/10.1016/j.regsciurbeco.2013.09.003

Air Quality (PM10) in 86 cities (2009)

Fig. 2. PM10 in 86 Chinese cities in 2009.

5S. Zheng et al. / Regional Science and Urban Economics xxx (2013) xxx–xxx

the PM10 values for the 86 cities in 2009. Again, many cities enjoyedair quality improvement during 2004 and 2009.

3.2.2. City mayor and CCP secretary data setA city has two leaders— the mayor and the CCP secretary. Our local

official data set is by city/year and it contains the mayor's and the CCPsecretary's name, information of age, gender, educational attainment,starting year and finishing year of his/her term on this position, theprevious position and the next position. This unique data set is not pub-licly available and it is collected from internal sources in China's politicalsystem.

By law, the mayor is the executive officer of the municipal (city)government. At the same time, the law also says that the mayor isunder the guidance of the city communist party committee of whichthe party secretary is the head. In practice, the division of labor is thatthe party secretary is in charge of the personnel and other politicalduties, while themayor is in charge of the daily operation of the govern-ment for which economic growth is a top priority and now energy con-servation and environmental protection are also addressed.14 Since thedeterminants of promotion may differ between party secretaries andmayors, we run the probit model for mayors and secretaries separately.Here we mainly discuss the regression results of the mayor's promotionequation, and place those of the city's CCP secretary in theWeb Appendix(Table A3).

14 In most cities, the party secretary is clearly the no. 1 leader because key decisions aremade in theparty committee. However, his power is checkedby themayor because in the-ory the executive orders should be delivered through themayor. In the end, the party sec-retary and themayor share power in a city. To the extent that themayor has to rely on thebureaucracy to manage the economy, his contribution to local economic growth is tied tothe party secretary's efforts to select more capable subordinates. In the real world, the in-teraction between the party secretary and themayor takes many forms and the pattern oftheir contributions to local economic growth cannot be readily parameterized (Yao andZhang, 2012).

Please cite this article as: Zheng, S., et al., Incentives for China's urbanmayoand public environmentalism, Reg. Sci. Urban Econ. (2013), http://dx.doi.or

The time length and spatial coverage of the local official and airquality data sets constrain the sample of our empirical study to be 86cities from 2004 to 2009. Among these 86 cities, there are 35 majorcities (4municipalities plus 31 provincial capital cities) and 51mediumand small-sized cities. We also collect the data of city attributesfrom China Statistics Yearbooks and China City Yearbooks, includingGDP per capita (GDPPC), city population (POP), annual rainfall in year1999 (RAIN, in mm), and temperature discomfort index in year 1999(See Zheng et al. (2009,2010) for the construction of this index)(TEMP). The variable definitions and descriptive statistics are listed inTable 1.

3.2.3. Public environmentalism measuresWe construct two indices to reflect urban residents' environmental-

ism, or concern intensity, over environmental issues. The first is theGoogle Insight index based on the internet search intensity of the keyword — “environmental pollution (huan jing wu ran)”, as a measure ofpublic concern intensity on Internet. Google Insights15 is a publicallyavailable online tool for tracking aggregate Google search intensityover time for specific geographic areas.16 Recent research (Kahn andKotchen, 2011) shows that Google search terms are a powerful tool topredict public health epidemics and economic activity. Google Insightscan report a search intensity index of a specific key word by geographicarea (in the Chinese version of Google Insights, the geographic unit isprovince) during a time period specified. Here, we construct this publicconcern index (PCI_1) by province/year.

15 Google Insights is available at http://www.google.com/insights/search/#.16 Baidu is a local search engine that iswidelyused inChina. It started to provide a similarsearch intensity index from June 2006.We do not use the Baidu index due to two reasons.First, the availability of this Baidu index cannot cover our study period; second, it isclaimed that the Baidu search engine manipulates the relative sorting order of somesearch outcomes.

rs tomitigate pollution externalities: The role of the central governmentg/10.1016/j.regsciurbeco.2013.09.003

Table 1Variable definitions and summary statistics.

Variable Definition Obs. Mean Std. dev.

City attributesGDPPC GDP per capita (RMB yuan), by city by year. 498 2.58 1.67GDP_GROWTH_M Relative GDP growth rate comparing to the previous mayor, by city by year. 461 0.03 0.04GDP_GROWTH_S Relative GDP growth rate comparing to the previous secretary, by city by year. 490 0.02 0.04POP City POP, by city by year. 498 535.16 397.17EDU Average years to schooling, by city by year. 498 6.17 4.32DIS_HK The distance to Hong Kong (km), by city by year. 498 1292.65 687.49INTERNET The number of internet users in region (10 thousand), by city by year. 498 69.48 126.58FDI Accumulatively foreign direct investment in the city (10 thousand RMB yuan), by city by year. 498 78,7706.8 1,363,649FACILITY_EXP Per unit GDP annual expenditure on waste gas treatment facilities, by city by year. 498 0.408% 0.904%RAIN Annual Rainfall in year 1999 (mm), by province. 31 922.99 568.28TEMP Temperature discomfort index in year 1999, defined in Eq:

TEMi = Sqrt{[Winter_temperaturei – max(Winter_temperature)]2

+ [Summer_temperaturei – min(Summer_temperature)]2}.See Zheng, Kahn and Liu (2010).

31 19.61 6.11

Mayor attributesPROMOTION_M Whether the mayor is promoted: 1 = yes, 0 = no. By city by year. 484 0.21 0.41AGE_MAYOR Mayor's age, by city by year. 484 51.13 4.21MASTER_MAYOR Whether the mayor has a master degree: 1 = yes, 0 = no. By city by year. 484 0.40 0.49TERM_MAYOR Whether the mayor is on his/her second term: 1 = yes, 0 = no. 484 0.07 0.25

Secretary attributesPROMOTION_S Whether the secretary is promoted: 1 = yes, 0 = no. By city by year. 491 0.20 0.40AGE_SECRETARY Secretary's age, by city by year. 491 52.37 4.19MASTER_SECRETARY Whether the secretary has a master degree: 1 = yes, 0 = no. By city by year. 491 0.40 0.49TERM_SECRETARY Whether the secretary is on his/her second term: 1 = yes, 0 = no. 491 0.10 0.30

Environment indicatorsPCI_1 Google Insight index of “environment pollution”, by province by year. 186 16.76 16.28PCI_2 Google search index of “environmental pollution”, by city by year. 498 0.43 2.59PM10 Average PM10 concentration (mg/m3), by city by year. 498 0.09 0.03PM10p75 75th percentile of PM10 concentration (mg/m3), by city by year. 498 0.12 0.03PM10_DECLINE Decline rate of PM10 concentration. 461 −0.02 0.11PM10p75_DECLINE Decline rate of PM10p75 concentration. 461 −0.02 0.12

Energy indicatorsEI Energy intensity: energy consumption per GDP added value (t standard coal per 10,000 RMB added value). 498 1.05 0.26EI_DECLINE Decline of energy intensity in the year. 498 0.09 0.09

6 S. Zheng et al. / Regional Science and Urban Economics xxx (2013) xxx–xxx

The above index relying on Google Insights cannot report the publicintensity by city, therefore we construct the second index by city/year.This second index aims to measure the frequency of pollution-relatedarticles reported in a city's local newspapers. We search for the samekey word “environmental pollution (huan jing wu ran)” in Google searchand set the search criteria to be the articles published in a city's majorlocal newspapers in a year. We count the number of entries and dividethis number by the total circulation of those local newspapers in eachyear to obtain a standardized index by city/year. Once published, newspa-per articles will be cited by webmedia sites and appear on those sites, sothis index captures the total concern intensity on the original articlespublished in local newspapers, and the webpages citing those articles.

Fig. 3. Energy intensity gradient with respect to GDP per capita (2004–2009).

Please cite this article as: Zheng, S., et al., Incentives for China's urbanmayoand public environmentalism, Reg. Sci. Urban Econ. (2013), http://dx.doi.or

4. Results

4.1. Local official's promotion propensities

In this section we test hypotheses H1 and H2. First, we test whetherenergy conservation and pollution reduction indicators are reflected incity mayors' promotion criteria. As economic growth is well known tobe a prime determinant for promotion in China (Li and Zhou, 2005),we test if the upper level government has begun to include proxies forgreenness into the promotion evaluation system.

Promotionit ¼ β0 þ β1 � GDP�GROWTHit þ β2 � GREENit þ β3 � Zitþ Year fixed effectsþ δit:

ð2Þ

In Eq. (2) the unit of analysis is city/year. The dependent variable is adummy indicating whether the mayor of city i gets promoted or not inyear t, which equals 1 if the officer moves to a higher level (including amayor promoted to be a CCP secretary in the same or another city),and equals 0 if he or she remains on the current position, or movesto another position in the same or lower level, or retires. “Abnormal”changes, e.g. death, arrest due to corruption, are excluded from thesample. We include year fixed effects. Standard errors are clustered bycity. This equation is estimated using a probit model.

We include GDP growth which is measured as the differencebetween the average annual GDP growth during this mayor/secretary'sterm (until that year) and that during his or her predecessor's tenure(Wu et al., 2013). Our focus is on the green indicators (GREEN) — EI(the key indicator in the TRS), FACILITY_EXP (the “accounting” indicatorof pollution mitigation effort), PM10 and PM10p75 measures. EI, PM10and PM10p75 are all measured in annual percentage decreases. The

rs tomitigate pollution externalities: The role of the central governmentg/10.1016/j.regsciurbeco.2013.09.003

7S. Zheng et al. / Regional Science and Urban Economics xxx (2013) xxx–xxx

annual percentage decrease in EI is a commonly-used quantitativetarget when evaluating local officials' performance. FACILITY_EXP is anannual flow variable so we use its lag term to mitigate inverse causalityproblem. Personal attributes (age, educational attainment and the termlength till that year) of the officer are also included in the model.17

Table 2 reports the results of mayors' promotion equation regres-sions. Our results indicate that the relative GDP growth rate (comparingto the previous mayor) is the most important determinant of a mayor'spromotion. This variable (GDP_GROWTH_M) is statistically significant at1% level in all regressions. Column (1) is the baseline model. In column(2) we augment the regression by including the PM10 decline measure(PM10_DECLINE). It contributes to the promotion probability and theeffect is marginally significant. In column (3) we replace this vari-able with the improvement in air quality in severely polluted days(PM10p75_DECLINE). This variable has a larger positive effect (sig. at10% level) on the promotion probability, indicating that air quality im-provement in themost polluted days helps the mayor in his/her perfor-mance evaluation. In column (4) we replace the air quality measure tothe energy intensity (EI) decline measure. It is positive and statisticallysignificant at 10% level. In column (5) we change to the lagged wastegas treatment facility expenditure (FACILITY_EXP), which is significantlypositive at 1% level. Further calculations based on columns (1) to (5)show that a 1% increase in PM10 declining rate and energy intensitydeclining rate, and 1% increase inwaste gas treatment facility expenditurewill increase the mayors' promotion odds by 0.26%, 0.37% and 0.9%, re-spectively. In our sample, the annual declining rate of PM10 varies from−42% to 41% for all city/year observations. This means that, comparingto a city that experiences no air quality improvement in a given year,the mayors in the best-performing city and the worst-performing city(in terms of air quality improvement) will have a 10.9 percentage pointshigher or 10.7 percentage points lower promotion odds, respectively.

In column (6), we include the three green indicators (PM10, EIand FACILITY_EXP) together. The facility variable is positive and stillstatistically significant at the 1% level and the energy intensity variableis significant at 10% level. The joint F-test shows that the three variablesare jointly significant at 1% level. In column (7)we replace PM10_DECLINEwith PM10p75_DECLINE. The above results support the hypothesisthat energy/environmental improvements are positively associated withmayor's promotion odds,18,19

It is possible that the annual changes in air quality and energy intensi-ty are fluctuating and contain lots of noise, and this year's annual changedoes not have an “instant” impact on the mayor's promotion. In Table A1in the Web Appendix, we replace this year's PM10 and EI decline rateswith the average annual declining rates in the last two years. The resultsare quite similar with those in Table 2, except that the EI decline ratehas the right sign but is not statistically significant.

17 In estimating Eq. (2), we are implicitly assuming that the central government does notfollowa systematic rule assigning certainmayors based on quality dimensionsunobservedby the econometrician to specific cities. For example, our estimates would be biased if thecentral government sent the best mayors (based on unobserved attributes) to the highpollution cities with the goal of cleaning them up. We only have a limited number ofmayors' attributes (age, years on position and educational attainment). Therefore omittedmayor attributesmay also correlatedwith both themayor's promotion probability and hiseffort in reducing pollution (as well as booming the economy). We acknowledge that weare unable to control for time-variant city attributes, and omitted mayor characteristics.For example politicians may pursue environmental progress due to self-interest. The cur-rent model specification neglects what presumably might be another important motiva-tion: officials might care about pollution because of their own family's health and safety.It would be interesting to test whether there is a difference in the policy outcomes for of-ficials with children relative to those without children. We thank a reviewer for this point.18 We acknowledge the possibility that environmental progress is positively correlatedwith other unobserved improvements in a city's quality of life. In this case,wewould over-state the role that environmental progress has played in causing the mayor's promotionwhen instead the true mechanism is improved overall quality of life.19 Table A3 in the Web Appendix reports the same regression results for cities' CCP secre-taries. In Table A3, the relative GDP growth rate is also the most important determinant ofparty secretaries' promotion odds. However, neither of the green indicators shows significantcontribution to party secretaries' promotion probability. They are also not jointly significant.

Please cite this article as: Zheng, S., et al., Incentives for China's urbanmayoand public environmentalism, Reg. Sci. Urban Econ. (2013), http://dx.doi.or

To control for possible time invariant omitted variable at city level(such as natural endowments and city political status) in Eq. (2), wehave also estimated linear probability model with city fixed effects.These are reported in the Web Appendix (see Table A2). The resultsare quite similar to the probit results reported in Table 2.

The highly significant facility expenditure variable but less statisticallysignificant air quality measures supports the claim that the promotioncriteria emphasize “accounting” measures over the more relevantpublic health variables such as PM10. In fact, annual PM10 decline rate(PM10_DECLINE) and annual facility expenditure per GDP (FACILITY_EXP)have a very weak correlation of 0.02 during our study period, and thecorrelation between PM10_DECLINE and last year's FACILITY_EXP is alsoweak with a correlation of 0.09.20,21

4.2. Measuring spatial variation in the public concern overenvironmental issues

We estimate Eq. (3) to explore the spatial and temporal variationsin the two public concern indices. Our hypothesis is that the publicconcern intensity over pollution will be higher if urban households'demand for environmental quality is stronger, the city (province) hasa higher level media openness level, and the city (province) has poorerenvironmental condition. We estimate:

log PCIitð Þ ¼ α0 þ α1 � log DIS�HKið Þþα2 � log INTERNETitð Þþα3 � log PM10itð Þþα4 � log GDPPCitð Þþα5 � log POPð Þþα6 � EDUitþ region fixed effectsþ year fixed effectsþ εit:

ð3Þ

Due to data availability, the unit of analysis is province/year forthe first index (PCI_1) and city/year for the second index (PCI_2). InEq. (2), DIS_HK is city/province i's distance to Hong Kong. Those placesclose toHongKong are exposed to a relatively freermedia environment.People there can watch Hong Kong TV, have a better access to HongKong newspapers and publications, and also have some contacts withHong Kong people. Therefore they have a better understanding of civilsociety, how the media works and what they can do. This effect dimin-ishes fast as the distance to Hong Kong increases so we measure thedistance in logarithm term. INTERNET is the number of internet usersin city/province i in year t. This measure should be positively correlatedwith media openness.22 Cities/provinces with higher GDPPC (per capitaGDP) and EDU (average years to schooling) are expected to face a strongerdemand for environmental amenities. People in cities/provinceswith badair quality (PM10)may complainmore.23We control for a city/province'spopulation, city/province fixed effects and year fixed effects.

20 The coefficients here show the average effects of energy efficiency and environmentalimprovement onmayors' promotion odds. In results that are available on request, we havere-estimated Eq. (2) for a subset of 35 major cities and estimated this equation for a laterperiod (2006–2009). We sought to test for heterogeneity with respect to the correlationbetween urban environmental performance and promotion probabilities. We supportthe hypothesis that in larger cities that there is a stronger correlation between environ-mental performance and promotion. To our surprise, we reject the hypothesis that this as-sociationhas grown larger in recent years (we compare the estimate ofβ2 in 2004–2005 toan estimate of this coefficient between 2006 and 2009).21 We have interacted PCI with EI decline or PM10 decline to see if energy efficiency orenvironmental improvements have larger effects on promotion odds in the cities withhigher public concern intensity, but no significant heterogeneity effect is found.22 We acknowledge the limitation of using the number of internet users (INTERNET) as aproxy for media openness. It may proxy for many other unmeasured things. For instance,those who are savvy Internet users likely have more unobservable human capital thanothers. Also, cities/provinces with more Internet users are more likely to have better ITand other infrastructure condition. Therefore we regard the effect of this variable on PCIas correlation rather than causality. We thank one reviewer for pointing out this.23 One reviewer points out that public concern index (PCI) may not bear a log–log rela-tionshipwith PM10 concentration. Itmay be possible that people living in the cities whereair quality is better than some threshold are far less likely to complain than thosewhere airquality is worse. We have tried this but did not find significant threshold in our sample.Therefore we stick to the continuous measure of air quality.

rs tomitigate pollution externalities: The role of the central governmentg/10.1016/j.regsciurbeco.2013.09.003

Table 2Probit estimates of a mayor's promotion probability.(Dependent variable: PROMOTION, whether the mayor was promoted in that year).

(1) (2) (3) (4) (5) (6) (7)

GDP_GROWTH_M 1.716*** 1.668*** 1.661*** 1.789*** 1.636*** 1.668*** 1.662***(3.12) (3.04) (3.01) (3.23) (2.96) (3.00) (2.98)

PM10_DECLINE 0.261 0.246(1.47) (1.38)

PM10p75_DECLINE 0.272* 0.253(1.72) (1.57)

EI_DECLINE 0.373* 0.380* 0.382*(1.73) (1.78) (1.80)

FACILITY_EXP(lag1) 0.878*** 0.898*** 0.876***(3.00) (2.93) (2.91)

AGE_MAYOR 0.00234 0.00206 0.00210 0.00198 0.00220 0.00155 0.00159(0.54) (0.48) (0.49) (0.45) (0.49) (0.35) (0.36)

MASTER_MAYOR 0.00701 0.00473 0.00389 0.00241 0.00510 −0.00151 −0.00226(0.19) (0.13) (0.11) (0.07) (0.14) (−0.04) (−0.06)

TERM_MAYOR 0.236*** 0.242*** 0.245*** 0.228*** 0.248*** 0.245*** 0.247***(2.70) (2.72) (2.76) (2.62) (2.81) (2.77) (2.79)

Year fixed effects Yes Yes Yes Yes Yes Yes YesObservations 461 461 461 461 461 461 461Pseudo R2 0.097 0.102 0.103 0.102 0.106 0.117 0.118chi2 47.67 48.76 50.50 48.68 55.85 57.53 58.67Joint F test for PM10_DECLINE (PM10p75_DECLINE), EI_DECLINE and FACILITY_EXP(lag1) 12.82***

(0.0050)13.37***(0.0039)

Notes: (1)Marginal effects are reported. (2) z-statistics are reported in parentheses. (3) ***: significant at the 1% level; **: significant at the 5% level;*: significant at the 10% level. (4) Stan-dard errors are clustered by the city level. (5) See Table 1 for variable definitions.

8 S. Zheng et al. / Regional Science and Urban Economics xxx (2013) xxx–xxx

Table 3 reports the regression results for Eq. (3). The dependent var-iable in column (1) is log(PCI_1). Provinces close toHongKong andwithmore internet users have higher public concern intensity over pollution.These two variables are both statistically significant. Richer provinceshave higher index values. People in the provinces with worse air quality(higher PM10) have significantly higher public concern intensity on “en-vironmental pollution”. In column (2) with the dependent variable oflog(PCI_2), distance to Hong Kong and the number of Internet users alsosignificantly contribute to higher public concern intensity over pollutionin local newspapers. All else equal, one percentage increase in the dis-tance from Hong Kong is associated with 1.4 and 0.5 percentage increasein PCI_1 and PCI_2, respectively. The elasticities of PCI_1 and PCI_2with re-spect to the number of IT users are 1.1 and 0.2, respectively. Cities withhigher human capital (EDU) and worse air quality have significantlyhigher public concern over pollution. Onemore year in local residents' av-erage years to schooling contributes to a 26.5% higher value of PCI_2.

Table 3The cross-city determinants of public concern about environmental issues.

Dependent variables log(PCI_1) log (PCI_2)

(1) (2)

log(D_HK) −1.417⁎⁎⁎ −0.500⁎⁎⁎

(−3.46) (−5.05)log(INTERNET) 1.118⁎ 0.241⁎

(1.90) (1.84)log(PM10) 1.632⁎ 0.524⁎⁎

(1.74) (2.07)log(GDPPC) 3.254⁎⁎⁎ 0.0516

(4.91) (0.35)log(POP) −2.804⁎⁎⁎ 0.538⁎⁎⁎

(−4.47) (4.85)EDU 0.961 0.265⁎⁎⁎

(1.33) (3.13)Constant 3.292 6.056⁎⁎⁎

(0.40) (2.80)East/west/central region dummies Yes YesYear fixed effects Yes YesStandard errors clustered By province By cityObservations 180 498R2 0.505 0.492

Notes: (1) t-statistics are reported in parentheses. (2) ***: significant at the 1% level;**: significant at the 5% level;*: significant at the 10% level. (3) See Table 1 for variabledefinitions.

Please cite this article as: Zheng, S., et al., Incentives for China's urbanmayoand public environmentalism, Reg. Sci. Urban Econ. (2013), http://dx.doi.or

4.3. Local officials' effort and environmental outcomes

In this sectionwe investigatewhether pressure from the public affectsa city's energy intensity and other environmental outcomes. In Table 4,we examine if there is an association between our public concern intensi-ty index and the city's performance on environmental criteria. The depen-dent variable in columns (1) and (2) is log(FACILITY_EXP). On the right-hand side,we include the one-year lagged public concern indices. Popula-tion and GDP per capita are controlled for. Province fixed effects (or cityfixed effects) are also included. For those provinces with lagged higherPCI_1 (or cities with lagged higher PCI_2), they experience significantlylarger increase in waste gas treatment facility expenditure. In columns(3) and (4), the dependent variable is log(EI). The public concern indiceshave weaker effects on energy intensity though the signs are intuitive.

We examine the role of leadership in determining energy/environ-mental actions and outcomes. Here we focus on city leaders' humancapital level, measured by years to schooling (EDU_MAYOR). Highly-educated leadersmay devotemore efforts in protecting the environment,and thus their cities may benefit from their leadership and enjoy anaggressive environmental/energy progress.

We exploit leadership transitions across China's cities.24 For eachcity i, if there is a leadership transition for the mayor position duringour study period, we calculate the change of air pollution, energy inten-sity and waste gas treatment facility expenditure between the last yearof the new leader (mayor II) and the last year of the previous leader(mayor I), and see if this change is correlated with the human capitaldifferential between the former and the new leader (Eq. (4)).25

Δi;mayorII−mayorIPM10 ¼ ϕ1 � Δi;mayorII−mayorIEDU þ ϕ2 � Δi;mayorII−mayorIPOPþ ϕ3 � Time�lengthi;mayorII−mayorI þ ϕ4 � PM10i;mayorI þ εit:

ð4Þ

Table 5 reports the regression results for the mayor transitions.We include the time length between the two time points to controlfor the time trend in the dependent variable. City population change is

24 Our approach builds on Jones andOlken (2005)who use the deaths of leaderswhile inoffice as a source of exogenous variation in leadership, and askwhether these plausibly ex-ogenous leadership transitions are associated with shifts in country growth rates.25 We acknowledge that the change of mayor in a city in our sample is not exogenous. Itis possible that a capable mayor is assigned to a city under-performing on environmentalcriteria to help that city. Therefore the effect ofmayor's education on environmental prog-ress is likely to be over-estimated in our model, and it can only be regarded as suggestiveevidence.

rs tomitigate pollution externalities: The role of the central governmentg/10.1016/j.regsciurbeco.2013.09.003

Table 5City leaders' attributes and environmental outcomes.

△PM10 △PM10p75 △EI △FACILITY_EXP

(1) (2) (3) (4)

△EDU_MAYOR −0.00212⁎⁎ −0.00237⁎⁎ −0.00870 −0.297(−2.59) (−2.17) (−1.30) (−0.90)

△POP 0.0101⁎ 0.0131⁎ 0.0700⁎⁎ −2.53⁎⁎

(1.80) (1.77) (2.06) (−2.55)△YEAR 0.00248⁎⁎ 0.00427⁎⁎ −0.0471⁎⁎⁎ 3.09⁎⁎⁎

(2.03) (2.43) (−4.28) (3.28)PM10_lag −0.186⁎⁎⁎

(−3.07)PM10p75_lag −0.231⁎⁎⁎

(−3.30)EI_lag −0.117⁎⁎⁎

(−2.99)FACILITY_EXP_lag 0.646

(1.64)Observations 70 70 88 88R2 0.296 0.299 0.760 0.440

Notes: (1) t-statistics are reported in parentheses. (2) ***: significant at the 1% level;**: significant at the 5% level;*: significant at the 10% level. (3) See Table 1 for variabledefinitions.

Table 4Environmental outcomes as a function of public concern.

Dependent variables log(FACILITY_EXP) log(EI)

(1) (2) (3) (4)

log(PCL_1) (lag1) 0.0793⁎ −0.00279(1.88) (−1.38)

log(PCL_2) (lag1) 0.0950⁎⁎⁎ −0.000758(3.15) (−0.13)

log(GDPPC) 0.716⁎⁎⁎ 1.372⁎⁎⁎ −0.291 −0.669⁎⁎⁎

(3.08) (7.68) (−1.35) (−19.27)log(POP) 0.191 2.210⁎⁎ −0.773 −1.410⁎⁎⁎

(1.23) (2.20) (−1.41) (−5.63)Constant −4.229⁎⁎⁎ −18.00⁎⁎ 4.786 11.52⁎⁎⁎

(−2.99) (−2.22) (1.32) (5.73)Province fixed effects Yes – Yes –

City fixed effects – Yes – YesYear fixed effects Yes Yes Yes YesObservations 153 415 153 415R2 0.939 0.901 0.951 0.946

Notes: (1) t-statistics are reported in parentheses. (2) ***: significant at the 1% level;**: significant at the 5% level;*: significant at the 10% level. (3) See Table 1 for variabledefinitions.

9S. Zheng et al. / Regional Science and Urban Economics xxx (2013) xxx–xxx

also controlled for. In columns (1) and (2), we find that if a new mayorhas a higher educational attainment than the former mayor, the cityenjoys a significant air quality improvement.26 A one year increase inthemayor's years to schooling is associatedwith a 2%decrease of the av-erage value of the PM10 concentration.27

China's unique political structure provides a plausible explanationfor the correlationswe have documented in Tables 4 and 5. A distinctivefeature in Chinese cities is that local governments have a “visible” handin influencing economic activities. That is why the TRS places localgovernments at the center of policy implementation. City leaders facetrade-offs between economic growth and environmental quality. Theycan use cheap land and favorable tax deduction policies to attract thefirms that can generate high GDP output, high tax revenues and morejob opportunities, but those firms may be energy-intensive ones.28

If city leaders want to achieve pollution control requirement, they canalso shut down heavily-polluted factories, and force those factories toleave (Witte et al., 2009). In this way the city will lose tax revenueand certain types of jobs. In addition, city leaders have a powerfulcontrol over the big SOE (State owned enterprise) energy-intensivefirms within their jurisdictions. Those SOE firms are also included inthe TRS. City leaders will sign target responsibility contracts with thosefirms' managers, and the evaluation of the managers' performance onthe energy and environmental dimensions will affect those managers'career (Qi, 2013).

4.4. Do public concern and the local leadership's characteristics influencethe relationship between pollution production and localeconomic development?

Building on the influential Grossman and Krueger (1995)'s studyof the “Environmental Kuznets Curve” (EKC), an entire subfield of

26 In columns (3) and (4), it seems that a citywith a newhighly educatedmayordoes notnecessarily achieve lower energy intensity and higher waste gas treatment expenditurerecords. Our guess is that highly-educated mayors may be more focused on real environ-mental outcomes.27 The educational background of CCP secretaries does not show any significant effect onthe city's energy and environmental progress (Table A2 in the Appendix).28 For example, in Zhejiang Province's “new technology zones”, the government spent100 thousand Yuan per mu (96 thousand US dollars per acre) on average to provide basicinfrastructure to the industrial land, but the average sale price of such industrial land tofirmswas only 86 thousand Yuan permu (83 thousandUSdollars per acre), even less thanthe infrastructure cost. Half of the industrial land parcels were sold at the price less than50% of the infrastructure construction cost. In some inland provinces that are keep to at-tract FDI and high-tax-revenue industries, some “new technology zones” sold their indus-trial land at zero price. See http://www.snzg.cn/article/2011/0318/article_22780.html

Please cite this article as: Zheng, S., et al., Incentives for China's urbanmayoand public environmentalism, Reg. Sci. Urban Econ. (2013), http://dx.doi.or

environmental economics has emerged that focuses on this “inverse-U” association between national per-capita income and pollution(Andreoni and Levinson, 2001; Stokey, 1998). In past research, wehave estimated Environmental Kuznets Curve across China's cities(see Zheng, Kahn and Liu 2010).

We estimate Eq. (5) below in order to examine how the shape of theEKC and in particular the GDP “turning point” varies as a function of cityattributes and the city's political leader's attribute.Where, Y represents ei-ther EI or PM10.We let the data itself tell the best order of the polynomialexpression of GDPPC and the form of this variable (in logarithm or not) toproduce the highest R2.29 City population, temperature index, and rain-fall are included as controls. Year fixed effects are also included.

log Yitð Þ ¼ η0 þXJ

j¼0

η1 j � GDPPC jit þ η2 � Xit þ Year fixed effectsþ ξit : ð5Þ

City leaders in urban China have greater influence than their coun-terparts in the US in influencing their cities' industrial composition.They use cheap land and big tax reduction to attract those firms theyfavor, and also shut down ormove those firms they dislike. City leaders'“visible” handwill reinforce this inverse U relationship between energyintensity and per capita GDP. We expect that those cities with higherpublic concern intensity over pollution or higher human capital, orthose cities with highly-educated mayor can reach the turning pointat a relatively low income level.

Our data tells us that for EI–GDPPC relationship, including thelevel and quadratic terms of GDPPC as explanatory variables generatesthe best fit of the regression. The PM10 bears the best-fit functionwith GDPPC when including the level, quadratic and cubic terms of thelatter (in log form). This indicates that the relationship is an “S” shape.A possible explanation is that, at the beginning of a city's growthwhen population is small, environmental input has a relatively largereffect on reducing pollution. As the city becomes larger, this marginaleffect diminishes, and technique and composition effects becomedominate. Thus the inverse-U shape will emerge in the latter period.Harbaugh et al. (2002) also find such an “S” shape relationship for SO2

and GDP per capita in his cross-country study.Table 6 reports the baseline regression results of estimating Eq. (5).

Standard errors are clustered by city. Columns (1) and (3) are trend

29 Harbaugh et al. (2002) suggest that the EKC is a fragile empirical result, and thepollution–income relationship is quite sensitive to functional forms (the order of the in-come polynomial function), the variable form, additional covariates, and the samplecomposition.

rs tomitigate pollution externalities: The role of the central governmentg/10.1016/j.regsciurbeco.2013.09.003

Fig. 4. PM10 gradient with respect to GDP per capita (2004–2009).

Table 6The energy intensity and PM10 gradients with respect to GDP per capita.

log(EI) log(PM10)

Trend regression Baseline EKC regression Trend regression Baseline EKC regression

(1) (2) (3) (4)

log(POP) 0.135⁎⁎⁎ 0.00126 log(POP) −0.00106 0.163⁎⁎⁎

(6.89) (0.04) (−0.07) (4.60)GDPPC 0.152*** log(GDPPC) −0.189⁎⁎

(4.29) (−2.16)GDPPC2 −0.0110** log(GDPPC)2 0.383⁎⁎⁎

(−2.44) (3.06)log(GDPPC)3 −0.150⁎⁎⁎

(−2.91)log(RAIN) −0.0381 log(RAIN) −0.126⁎

(−0.71) (−1.84)log(TEMP) −0.149 log(TEMP) 0.275

(−1.42) (1.61)Year dummies Yes Yes Year dummies Yes YesConstant −3.132⁎⁎⁎ 0.633 Constant 0.199⁎⁎ −3.285⁎⁎⁎

(−24.18) (1.13) (2.19) (−3.68)Observations 498 498 Observations 498 498R2 0.112 0.558 R2 0.354 0.372Peak turning point: Peak turning point:(2003 RMB 1000) 69.1 (2003 RMB 1000) 40.7(2003 US dollars) 8324 (2003 US dollars) 4904Joint F test for GDPPCj 18.73⁎⁎⁎

(0.000)Joint F test for log(GDPPC)j 3.37⁎⁎

(0.0223)

Notes: (1) t-statistics are reported in parentheses. (2) ***: significant at the 1% level; **: significant at the 5% level;*: significant at the 10% level. (3) See Table 1 for variabledefinitions.

10 S. Zheng et al. / Regional Science and Urban Economics xxx (2013) xxx–xxx

regressions for PM10 and EI. We only control for city population and ex-amine the coefficients of year dummies. For the average city, air pollu-tion (PM10) has been sharply declining since 2006 (the beginning ofthe 11th FYP), and energy intensity's decline followed since 2007. In col-umn (2), after controlling for city population, rainfall and temperatureindex, there is a clear inverted-U relationship between EI and GDPPC.The turning point is about 69.1 thousand Yuan (8324 US dollars,in 2003 constant price30). When estimating the PM10 gradient withrespect to GDP per capita (column (4)), we include the level, quadraticand cubic terms of log(GDPPC) to achieve a better goodness of fit. Theresult shows an S-shape of the PM10-GDP path. Those cities that alreadypass the peak point and enter the declining part on the right of thatpoint face a bright future in air quality improvement. The peak pointis about 40.7 thousand Yuan (4900 US dollars, in 2003 constant price)per capita. Figs. 3 and 4 plot all the cities on the predicted inverted-Ushape of the EI–GDP curve, and the S-shape of the PM10–GDP curve.31

30 $1 = RMB8.3, in 2003.31 All other independent variables are set at their mean values.

Please cite this article as: Zheng, S., et al., Incentives for China's urbanmayoand public environmentalism, Reg. Sci. Urban Econ. (2013), http://dx.doi.or

To test if the nudge from the general public pushes the city to reachthe turning point on the EKC curves at a lower GDP per capita level, wecompare the turning points of the curves for different city sub-groups(Table 7). Those sub-groups are classified using the median values ofthe corresponding indicators. In columns (1)–(4), we use the medianvalue of each public concern index to divide the city sample into twosubsamples. Cities with higher public concern intensity over pollution(PCI_1 or PCI_2) have earlier turning points for both EI–GDP andPM10–GDP curves. If we employ a stricter classification method anddefine those cities with both index values exceeding the correspondingmedian index values, cities with this “hybrid pollution concern index”higher than its median value have much earlier turning points forboth greenness indicators (columns (5) and (6)).

Cities with higher human capital have a stronger demand forenvironmental quality, andwe do observe those cities to have an earlierturning point for both EI and PM10's gradients with respect to GDP(columns (7) and (8)).32 Furthermore, cities with highly-educatedmayors enjoy earlier environmental progress as they reach the incometurning point at a lower level of income. Future research should testwhether the rise of civil society in China will further lower the EKCincome turning point.

5. Conclusions

The rise of “green cities” in China would directly improve quality oflife for hundreds of millions of people as reductions in pollution wouldraise the standard of living. But, over the last 30 years China's localleaders have had strong incentives to pursue local economic growtheven if the environment was sacrificed. A recent qualitative politicalscience literature has argued that a regime shift has taken place so thatreducing pollution in urban China is now a priority (Landry, 2008;Wang, forthcoming).

This paper has used several data sets to test this optimistic hypothesisby studying regional differences in the propensity for local officials to be

32 The joint F-tests for GDPPCj (or log(GDPPC)j) are reported. Most of these estimates arestatistically significant. This indicates that the reported pollution–income relationships al-so hold for the sub-groups.

rs tomitigate pollution externalities: The role of the central governmentg/10.1016/j.regsciurbeco.2013.09.003

Table 7Peak income turning points for city energy intensity and PM10 by city type.

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Pollution concernindex #1 (PCI_1)

Pollution concernindex #2 (PCI_2)

Hybrid pollution concern index City human capital Mayor education

Higher concern Lower concern Higher concern Lower concern Higher concern Lower concern Higher Lower Lower Higher

EI (RMB 10,000) 5.24 7.61 5.97 6.82 5.55 7.19 6.45 8.17 7.41 8.29(US dollar) 6319 9166 7198 8217 6686 8659 7773 9838 8922 9990

Joint F test for GDPPCj 15.02⁎⁎⁎

(0.000)16.69⁎⁎⁎

(0.0005)24.38⁎⁎⁎

(0.000)2.13(0.132)

18.16⁎⁎⁎

(0.000)12.28⁎⁎⁎

(0.000)4.22⁎⁎

(0.0217)38.85⁎⁎⁎

(0.000)16.97⁎⁎⁎

(0.000)12.36⁎⁎⁎

(0.000)PM10 (RMB 10,000) 3.48 4.27 3.66 4.48 2.88 4.32 3.93 6.27 3.69 4.06

(US dollar) 4193 5145 4412 5400 3471 5209 4735 7554 4446 4892Joint F test forlog(GDPPC)j

3.82⁎⁎

(0.0134)1.46(0.231)

2.03(0.125)

12.02⁎⁎⁎

(0.000)3.10⁎⁎

(0.037)3.75⁎⁎

(0.0141)8.90⁎⁎⁎

(0.000)1.86(0.151)

5.71⁎⁎⁎

(0.0019)1.21(0.313)

Notes: (1) The variables are the same as columns (2) and (4) in Table 6. (2) See Table 1 for variable definitions. (3) “Human capital” ismeasured as the ratio of peoplewho have the highesteducational attainment as senior high school or above.

11S. Zheng et al. / Regional Science and Urban Economics xxx (2013) xxx–xxx

promoted and for the public to reveal concern about environmentalchallenges. We have documented a number of associations that togethersuggest that a regime shift has takenplace in China such that local officialsare increasingly incentivized to consider the pollution consequences oftheir actions. Both the central government and many of the increasinglyeducated, informed and sophisticated urbanites demand environmentalprogress. Based on new estimates of the cross-city pollution versusincome relationship, we find that cities with higher human capital andhigher public concern about environmental issues have earlier EKC per-capita income turning points for energy intensity and particulate matter.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.regsciurbeco.2013.09.003.

References

Acemoglu, Daron, Robinson, James, 2012. Why Nations Fail: The Origins of Power,Prosperity, and Poverty, Crown Business, New York.

Andreoni, James, Levinson, Arik, 2001. The simple analytics of the environmental Kuznetscurve. J. Public Econ. 80, 269–286.

Andrews, S.Q., 2008. Inconsistencies in air quality metrics: ‘Blue Sky’ days and PM10concentrations in Beijing. Environ. Res. Lett. 3 (3), 034009.

Besley, Timothy, Burgess, Robin, 2002. The political economy of government responsiveness:theory and evidence from India. Q. J. Econ. 117 (4), 1415–1451.

Bo, Z., 1996. Economic performance and political mobility: Chinese provincial leaders.J. Contemp. China 5 (12), 135–154.

Boyd, Olivia T., 2012. China's energy reform and climate policy: the ideas motivatingchange. CCEP Working Paper 1205. Crawford School of Public Policy. AustraliaNational University.

Chen, Ye, Li, Hongbin, Zhou, Li-An, 2005. Relative performance evaluation and the turn-over of provincial leaders in China. Econ. Lett. 88 (3), 421–425.

Chen, Y., Jin, G.Z., Kumar, N., Shi, G., 2011. The Promise of Beijing: Evaluating the Impact ofthe 2008 Olympic Games on Air Quality (No. w16907). National Bureau of EconomicResearch.

Department of Climate Change, NDRC, 2010. Appendix II — nationally appropriatemitigation actions of developing country parties: China, letter including autonomousdomesticmitigation actions, UnitedNations framework convention on climate change.Available at http://unfccc.int/meetings/cop_15/copenhagen_accord/items/5265.php.

Ferraz, Claudio, Finan, Frederico, 2008. Exposing corrupt politicians: the effect of Brazil'spublicly released audits on electoral outcomes. Q. J. Econ. 123 (2), 703–745.

Ferreira, Fernando, Gyourko, Joseph, 2009. Do political parties matter? Evidence from U.S.cities. Q. J. Econ. 124 (1), 399–422.

Gentzkow, Matthew, Shapiro, Jesse, 2010. What drives media slant? Evidence from U.S.daily newspapers. Econometrica 78 (1), 35–71.

Grossman, Gene M., Krueger, Alan B., 1995. Economic growth and the environment.Q. J. Econ. 110 (2), 353–377.

Guan, Dabo, Liu, Zh.u., Geng, Yong, Lindner, Sören, Hubacek, Klaus, 2012. The gigatonnegap in China's carbon dioxide inventories. Nat. Clim. Change 2, 672–675.

Harbaugh, William T., Levinson, Arik, Wilson, David, 2002. Reexamining theempirical evidence for an environmental Kuznets curve. Rev. Econ. Stat. 84(3), 541–551.

Hårsman, Björn, Quigley, John M., 2010. Political and public acceptability of congestionpricing: ideology and self‐interest. J. Policy Anal. Manag. 29 (4), 854–874.

Please cite this article as: Zheng, S., et al., Incentives for China's urbanmayoand public environmentalism, Reg. Sci. Urban Econ. (2013), http://dx.doi.or

Jia, Ruixue, 2012. Pollution for Promotion. Stockholm University Working Paper.Jones, Benjamin F., Olken, Benjamin A., 2005. Do leaders matter? National leadership and

growth since World War II. Q. J. Econ. 120 (3), 835–864.Kahn, Matthew E., Kotchen, Matthew J., 2011. Business cycle effects on concern

about climate change: the chilling effect of recession. Clim. Change Econ. 2(03), 257–273.

Landry, Pierre Francois, 2008. Decentralized Authoritarianism in China: TheCommunist Party's Control of Local Elites in the Post-Mao Era. CambridgePress, Cambridge.

Li, H., Zhou, L., 2005. Political turnover and economic performance: the disciplinary role ofpersonnel control in China. J. Public Econ. 89, 1743–1762.

List, John A., Sturm, Daniel M., 2006. How elections matter: theory and evidence fromenvironmental policy. Q. J. Econ. 121 (4), 1249–1281.

Liu, Junguo, Yang, Hong, 2009. China fights against statistical corruption. Science 325,675–676.

Massey Simonich, S.L., 2009. Response to comments on “atmospheric particulatematter pollution during the 2008 Beijing Olympics”. Environ. Sci. Technol. 43 (19),7590–7591.

Pargal, Sheoli, Wheeler, David, 1996. Informal regulation of industrial pollution indeveloping countries: evidence from Indonesia. J. Polit. Econ. 104 (6),1314–1327.

Qi, Ye, 2013. Annual Review of Low-Carbon Development in China (2013) — Policy Impli-cation and Institutional Innovation. Tsinghua University and Climate Policy Initiative.

Selden, T.M., Song, D., 1995. Neoclassical growth, the J curve for abatement and theinverted U curve for pollution. J. Environ. Econ. Manag. 29, 162–168.

Stokey, N., 1998. Are there limits to growth? Int. Econ. Rev. 39, 1–31.Tang, X., Shao, M., Hu, M., Wang, Z., Zhang, Junfeng (Jim), 2009. Comment on “atmospheric

particulate matter pollution during the 2008 Beijing Olympics”. Environ. Sci. Technol.43, 7588.

Tong, Zhifeng, 2013. Reflections on environmental mass incidents in China. Chinese re-search perspectives online. Brill online. See: http://www.paulyonline.brill.nl/entries/chinese-research-perspectives-online/reflections-on-environmental-mass-incidents-in-china-C9789004173491_021.

Wang, Alexander, 2013. The search for sustainable legitimacy: environmental lawand bureaucracy in China. Harvard Environmental Law Review. (forthcoming,Available at http://ssrn.com/abstract=2128167 or http://dx.doi.org/10.2139/ssrn.2128167).

Wang, W., Primbs, T., Tao, S., Simonich, S.L.M., 2009. Atmospheric particulate matterpollution during the 2008 Beijing Olympics. Environ. Sci. Technol. 43 (14),5314–5320.

Witte, J.C., Schoeberl, M.R., Douglas, A.R., Gleeason, J.F., Krotkov, N.A., Gille, J.C., Pickering,K.E., Livesey, N., 2009. Satellite observations of changes in air quality during the 2008Beijing Olympics and Paralympics. Geophys. Res. Lett. 36, L17803.

World Bank, 2007. World Development Indicators. World Bank, Washington, DC.Wu, M., 2010. How does central authority govern the provincial leaders? Evidence from

China. 1–27.Wu, J., Deng, Y., Huang, J., Morck, R., Yeung, B., 2013. Incentives and outcomes: China's

environmental policy. NBER Working Paper.Yao, Y., Zhang, M., 2012. Subnational Leaders and Economic Growth.Yao, X., Xu, X., Sabaliauskas, K., Fang, M., 2009. Comment on “atmospheric particulate

matter pollution during the 2008 Beijing Olympics”. Environ. Sci. Technol. 43 (19),7589-7589.

Zhang, M., Wang, S., Wu, F., Yuan, X., Zhang, Y., 2007. Chemical compositions of wetprecipitation and anthropogenic influences at a developing urban site in southeasternChina. Atmos. Res. 84, 311–322.

Zheng, Siqi, Fu, Yuming, Liu, Hongyu, 2009. Demand for urban quality of living in China:evidence from cross-city land rent growth. J. Real Estate Finance Econ. 38, 194–213.

Zheng, Siqi, Kahn, M.E., Liu, Hongyu, 2010. Towards a system of open cities in china: homeprices, FDI flows and Air quality in 35major cities. Regional Science and Urban Econom-ics. 40 (1), 1–10.

rs tomitigate pollution externalities: The role of the central governmentg/10.1016/j.regsciurbeco.2013.09.003