Economic evaluation of environmental externalities in China’s coal-fired power generation

Energy Policy 102 (2017) 307–317

Contents lists available at ScienceDirect

Energy Policy journal homepage: www.elsevier.com/locate/enpol

Economic evaluation of environmental externalities in China’s coal-fired power generation

MARK

⁎

Xiaoli Zhaoa, , Qiong Caib, Chunbo Mac, Yanan Hud, Kaiyan Luob, William Lie a

School of Business Administration, China University of Petroleum-Beijing, Beijing, China School of Economics and Management, North China Electric Power University, Beijing, China c School of Agricultural and Resource Economics, Centre for Environmental Economics and Policy, University of Western Australia, Crawley, WA 6009, Australia d Kunshan China Resources City Gas Co., Ltd., Jiangsu, China e Swarthmore College, 500 College Avenue, Swarthmore, PA 19081, USA b

A R T I C L E I N F O

A BS T RAC T

Keywords: Coal-fired power industry Environmental externalities Willingness to pay China

Serious environmental externalities exist in China’s power industry. Environmental economics theory suggests that the evaluation of environmental externality is the basis of designing an efficient regulation. The purposes of this study are: (1) to identify Chinese respondents’ preferences for green development of electric power industry and the socio-economic characteristics behind them; (2) to investigate the different attitudes of the respondents towards pollution and CO2 reduction; (3) to quantitatively evaluate the environmental cost of China’s coal-fired power generation. Based on the method of choice experiments (CE) and the 411 questionnaires with 2466 data points, we found that Chinese respondents prefer PM2.5, SO2 and NOx reduction to CO2 reduction and that the environment cost of coal-fired power plants in China is 0.30 yuan per kWh. In addition, we found that the socioeconomic characteristics of income, education, gender, and environmental awareness have significant impacts on respondents’ choices. These findings indicate that the environmental cost of coal-fired power generation is a significant factor that requires great consideration in the formulation of electric power development policies. In addition, importance should also be attached to the implementation of green power price policy and enhancement of environmental protection awareness.

1. Introduction Along with its dramatic economic development over the past decades, China has become the largest country energy consumption and CO2 emission in the world, and is facing increasingly challenges of energy and environment in its path towards sustainable development. Numerous cities have witnessed frequent weather events of heavy fog and haze that linger over large areas of land, resulting in damages to people’s mental and physical health. The appearance of China’s “cancer villages”, one of the serious consequences of the environment deterioration, is the result of countrywide pollution of air, land, and water (Zhao et al., 2014). The mortality of lung cancer patients in China has increased by 465% over the past 30 years,1 and China’s annual newly added cancer cases account for more than 20% of the total new cases in the world.2 Balancing environmental protection and economic development is

a key step to realize sustainable development in China. Xepapadeas (1992) noted that the dischargers will always choose higher than socially desirable emission levels if by doing so they can increase their profits. As such, in the absence of environmental regulation, the external environmental cost is not the decision-making factor that producers consider, i.e., the cost of emission is zero (Tang, 2011). The objective of environmental regulation then is to motivate dischargers to internalize their external environmental cost. However, a serious challenge that policy makers are faced with is the lack of quantitative information on external environmental cost; there is no conventional market or price for pollutants to provide quantifiable measures of the environmental externality. Therefore, quantitative assessment of external costs to the environment is an important foundation for building stringent environmental regulations. Along with China’s dramatic economic growth over the past decades, the external environmental cost resulted from the coal-

⁎

Corresponding author. E-mail address: [email protected] (X. Zhao). Data source: the 3rd national investigation on death causes conducted by National Health and Family Planning Commission of PRC. 2 Data source: the 22nd Asia Pacific Cancer Conference. 1

http://dx.doi.org/10.1016/j.enpol.2016.12.030 Received 4 July 2016; Received in revised form 31 October 2016; Accepted 15 December 2016 0301-4215/ © 2016 Published by Elsevier Ltd.

Energy Policy 102 (2017) 307–317

X. Zhao et al.

Section 4. Although the current studies have found numerous significant conclusions in environmental cost evaluation of power sector, the following issues remain to be addressed: (1) There is a lack of evaluation of cleaner production preference in coal-fired power industry, especially in China’s coal-fired power industry based on CE. (2) The impact of environmental conscious on respondent preference has attracted limited concerned. Ellis et al. (2007) pointed out that respondents value and feeling have significant influence on their preference to renewable energy. However, the empirical studies of integrating environmental consciousness into the analysis of environmental cost of power sector are lack. (3) The different preference to GHG (such as CO2), PM2.5, SO2 and NOx is not considered. The contribution of this study is to investigate the respondent preferences to green development of China’s coal-fired power industry considering the differences in environmental consciousness and various emissions based on the CE model. To reduce hypothetical bias, we incorporate a “cheap talk script”3 method in the design of questionnaires following Carlsson et al. (2005) and Cummings and Taylor (1999).

dominated energy mix in the country is significant. In China, the coalbased power generation industry accounts for approximately 50% of the total coal consumption, and produces 40% of the CO2 emission, 60% of the SO2 emission and 60% of the NOx emission of the whole country (SO2 and NOx emissions are the primary drivers in acid rain creation). To control the pollutant and greenhouse gas (GHG) emissions in electric sector, China’s government has published a series of environmental regulations (Zhao et al., 2015). However, the implementation effects of these regulations are limited. One important reason is that the formulation of these regulations lacks strict reasoning and is subjective and arbitrary. According to the environmental economics regulation theory, an efficient environmental policy should be designed based on the cost caused by pollutant emissions. Hence, the quantitative analysis of the environmental cost of China’s coalbased industry will provide a scientific basis for the environmental regulation in China’s power industry. Based on the method of choice experiments (CE) (also called choice model), we found that respondents have a preference to an electricity premium for promoting cleaner production of coal-fired power plants, which is consistent with our expectation that greater environmental awareness translates into a higher premium paid. It is also concluded that Chinese correspondents have higher preferences to the reduction of PM2.5, SO2 and NOx than CO2. This result implies that China’s government should pay more attention to educating the public about the importance of emission reduction of CO2 and other kinds of GHG. Moreover, it is concluded that each household has the willingness to pay 40 yuan per month or 0.30 yuan per kWh for the best situation of environmental improvement. The rest of this paper is organized as follows: Section 2 goes through main studies and results applying the methods of CE. Section 3 is about the background of coal-fired power in China and its environmental externalities. Sections 4 and 5 describe our choice experiment design and methodology respectively; and Section 6 discusses the results. Conclusion and policy implications are provided in Section 7.

3. Current status and environmental externalities of China’s coal-fired power industry Currently, China’s power supply mix is dominated by coal-based power generation.4 Fig. 1 shows that prior to 2011, the proportion of thermal power (dominated by coal-based power) generation in the national total is approximately 85%, and since 2011, this proportion has decreased due to China’s renewable energy development and the restraint in coal consumption growth. However, in 2014 the proportion was still approximately 75%. The high proportion of thermal power generation results in the excessive environmental externality in China’s electricity production. According to the Annual Statistic Report on Environment in China (2006–2014), the emission of NOx, SO2 and smoke dust from thermal power industry accounted for 55.74%, 39.30% and 16.17% of China’s total emission of industrial sectors respectively (Fig. 2). Fig. 2 shows further that, SO2 and smoke dust emission has taken on a decreased trend in China’s thermal power industry since 2006. This is caused by stricter regulation standards. China published Standards for Air Pollutant Discharge from Thermal Power Plants (GB 13223-1991) (SAPD) for the first time in 1991. The standard was improved in 1996, 2003, and 2011. Along with increasingly stricter standards, China’s thermal power sector was making increasingly cleaner production. However, unlike the continuously decreasing emission trend of SO2 and smoke dust, NOx emission saw an increase at first and then began to decrease in 2012 (Fig. 2). This phenomenon is probably due to the fact that China’s government paid more attention to the regulation of SO2 and smoke dust than of NOx prior to 2011. The fourth amendment of the SAPD (2011) was much stricter than the SAPD (2003) in the emission limit of NOx from coal-fired power plants, significantly reducing the upper limit for NOx from 45 to 1500 mg per cubic meter to 100 mg per cubic meter for coal-based power plants. Differing from the localized impact on the environment of emissions from SO2, smoke dust, NOx, CO2 emission has a global impact.

2. Literature review Study of the environmental externalities of the power industry has recently become a hot research topic. Most studies have focused on the value of renewable energy by evaluating respondents’ preference to it. While Susaeta et al. (2011) assessed preferences for woody biomassbased electricity in the United States, which amounted to US $40.5 per capita annually, Longo et al. (2008) assessed the preferences of respondents for a policy to promote renewable energy in England and found that the willingness to pay (WTP) every year for reducing carbon emissions by one ton of CO2 is US $967 in UK. Ek and Persson (2014) explored public preferences for characteristics of wind farm establishments in Sweden. Their results indicated that respondents are willing to pay a higher electricity fee corresponding to approximately 0.6 Euro per kWh to avoid wind farms located in mountainous areas and private ownership regions. Kosenius and Ollikainen (2013) elicited people’s collective monetary preferences for four renewable energy sources: wind, hydro, crops, and wood, and considered the impacts of biodiversity, local jobs, carbon emissions, and the household’s electricity bill. They concluded that the national aggregate WTP, for a combination of renewable energy technologies, is over 500 million Euros in Finland. A few studies compared the environmental value of different renewable energy. For example, Borchers et al. (2007) estimated consumer preferences and the WTP for voluntary participation in green energy electricity programs, including wind, biomass, solar, and farm methane. Their results showed that individuals have a preference for solar over a generic green and wind; biomass and farm methane were found to be the least preferred sources. All of the above studies are based on the method of CE. The merits of CE comparing with other methods used for evaluating environmental cost will be discussed in

3 Estimation results of willingness to pay from experiments often demonstrate significant differences between responses to the real and hypothetical valuation questions. Such differences are usually called “hypothetical bias”. In order to avoid the bias, Cummings and Taylor (1999) put forward the “cheap talk script” method. By involving actual payments or designing a context that will happen, the “cheap talk script” method can elicit responses to hypothetical valuation questions that were indistinguishable from responses to valuation questions (Cummings and Taylor, 1999). 4 Coal used by power generation generally includes hard coal, bituminous coal, and poor lean coal. Around 90% of coal used in China’s power generation sector is bituminous coal (Zhang, 2007).

308

Energy Policy 102 (2017) 307–317

X. Zhao et al.

Fig. 1. China’s power generation mix. Data source: China Statistical Yearbooks (1978–2014).

Fig. 2. Pollutant emissions of China’s thermal power generation and their proportion to the total industries. Data source: Annual Statistic Report on Environment in China (2006– 2014).

Fig. 3. CO2 emissions in thermal power sector of China, in China and in the whole world. Data source: BP Statistical Review of World Energy 2015；CO2 emission of China thermal power plants is calculated based on the Intergovernmental Panel on Climate Change (IPCC) recommended method with the data from China Statistical Yearbooks (2000–2014).

309

Energy Policy 102 (2017) 307–317

X. Zhao et al.

Moreover, its environmental impact is far more dangerous and has been occurring for a longer period of time. China’s CO2 emission accounted for approximately 30% of the world total in 2014 (Fig. 3), among which approximately 40% were from the thermal power industry in China. CO2 emission in China’s thermal power industry has increased since 2000, mainly due to China’s government’s attention to CO2 emission is less than to other pollutants in the thermal power industry. For example, there is no limitation on CO2 emission in SAPD. However, the situation is changing – China has aimed for a CO2 emission peak in 2030 and to regulate more stringently the CO2 emission in the thermal power industry given this industry’s significant contribution to the national total. The quantitative evaluation of environmental cost in China’s coal-fired power industry supports setting suitable regulation policies for reducing pollutant and GHG emissions in the thermal power industry, as well as in other industries.

Table 1 Description of the attributes and levels. Attribute

Description

Level

CO2

Percentage reduction of CO2 emissions

1–5% (low) 6–10% (medium) 11–20% (high)

PM2.5

Air quality level, corresponding to percentage reduction of dust emissions

Excellent air quality Good air quality Light pollution Moderate pollution Heavy pollution(status)

Acid rain

Distribution of acid rain, corresponding to percentage reduction of SO2 and NOX

Bill

Increase in the monthly electricity bill

Non-Acid Rain Light Acid Rain Moderate Acid Rain Relatively severe Acid Rain Heavy Acid rain ¥0, ¥5, ¥10, ¥15, ¥25

4. Choice experimental design The basic idea of choice experimental design (CE) is to provide the choice set and the state combination (using a statistical model to conduct the profit-loss comparison of different attribute states) of different attributes to the participants who are required to select the scenarios they prefer. Due to its advantages in ease of data collection, this method has been widely used by researchers in environmental valuation estimation and environmental management fields (Hanley et al., 1998; Lee and Yoo, 2009; GarcíaLlorente et al., 2012). Other methods, such as a cost-benefit analysis (Faaij et al., 1998; Li et al., 2014; Bachmann and van der Kamp, 2014) and ExternE project (Sáez et al., 1998; Georgakellos, 2010) have also been used for evaluating the external environmental cost in electricity sectors. However, the cost-benefit analysis cannot accurately reflect the WTP of the public. The data inputs in this method are limited to an outline of the impact of environmental changes on productivity, or the policy cost change of governance (Turner et al., 1994). The advantage of the ExternE method is the simple calculation by which the economic value of environmental damage can be analyzed automatically with the input of parameters. However, this method cannot evaluate the external environmental costs across different countries or regions, given the differences between public value assessment across different regions and cultures. In contrast, the method of CE is able to not only make overall value assessments to the changes in environmental quality but also assess each attribute value for environmental quality by observing the trade-offs of respondents to different situations (corresponding combinations of different properties) (Hanley et al., 1998). In particular, a primary benefit of CE is the ability to elicit marginal attribute value of goods or services (Hanley et al., 2002). Moreover, the CE method is capable of analyzing large amounts of information and estimating the changing range of environmental attributes. These advantages prove useful in a context where many policies are concerned with changing the attribute levels rather than improving (or not) the overall environment. Therefore, the CE is used in this study to quantify respondents’ preferences for clean production in the coalfired power generation sector of China. The CE approach was initially developed by Louviere and Hensher (1982) and Louviere and Woodworth (1983). Random Utility Theory (Manski,1977; Mcfadden, 1986) and Characteristic Utility Theory are the two principle theories for modeling of choice experiments (Lancaster,1966). According to Characteristic Utility Theory, to evaluate environmental cost, a key issue is to construct a utility function, and transfer a choice issue into utility comparison (or value comparison).

A CE design includes three stages. The first stage identifies the attributes of evaluated objects. For example, the attributes for evaluating the value of renewable energy investment can include landscape, wildlife, air pollution, and employment (Ku and Yoo, 2010). In this study, we only concern the environmental cost of coal-fired power generation, hence, the attributes focus on environmental quality. We conducted an extensive literature review (Susaeta et al., 2011; Lee and Yoo, 2009; Georgakellos, 2010), and reviewed environmental quality standards, such as the Ambient Air Quality Standards (GB 3095— 2012), to identify the relevant attributes of environmental quality (e.g., PM2.5, CO2 emission, acid rain) and the proportional contribution of each attribute. Generally, dust (TSP) is used for measuring air pollution and includes total suspended particles with diameters of less than 100 micrometer. The suspended particles, known as PM 10 and PM 2.5, are particular matters with diameters of less than 10 micrometer, and 2.5 micrometer, respectively. Although the formal definition for dust includes both PM 10 and PM 2.5 particles, in view of the serious impact of PM2.5 on health, PM2.5 is used as one of the attributes in this study instead of dust. Moreover, to refine the attributes and proportions, we used a focus group to discuss respondents’ understanding of and reaction to them. The second stage tests the rationality of the CE design. To verify the appropriateness of our design, we fielded an exploratory survey to address the following questions: (1) Are we missing any important factors (attributes) related to the environmental externality of coal-fired power? (2) Are the levels of attributes identified properly? (3) Is the questionnaire easily and correctly understood? Based on the data compiled from the aforementioned two stages, four attributes and the relative levels of each attribute were selected for the CE (Table 1). Three environmental attributes are: the reduction of CO2 emission and PM2.5 (dust emission), amount of acid rain (SO2 emission and NOX emission), and one price attribute, which is defined as an increase in the monthly electricity bill per household (Table 1). The levels of the price attribute were decided through a pre-test and by consulting an expert at the University of Western Australia. The lower bound is 0 and the upper bound is 50% percent of the electricity bill, which was 25 Yuan monthly in 2013. The third stage of the experimental design involved data collection effort. In this study, the data were collected through 600 face-to-face questionnaires containing 6 choice sets per respondent (we designed three types of questionnaires resulting in 18 sets in total). To solve the key problem encountered in the CE – information overload i.e., too many alternatives with too many complex attributes – we applied the

310

Energy Policy 102 (2017) 307–317

X. Zhao et al.

Table 2 Statistical results of the samples. Variables

Definition and level

Frequency

Percent (%)

Gender

Gender 1=male 0=female

252 167

59.37 40.63

Age 1=50 years old and above 2=between 39 and 49 years old 3=between 29 and 39 years old 4=else

34 106 91 188

8.27 25.79 22.14 43.80

Age

Education

Income

No children

Greenhouse

Education level 1=college or higher 2=high school 3=else

287 75 57

67.88 18.25 13.87

Annual per capita income 1=above 50,000Yuan 2=25,000–50,000 Yuan 3=10,000–25,000 Yuan 4=else

112 125 104 78

27.25 29.68 25.30 17.76

Have no children 1=Yes 0=No

173 246

40.15 59.85

Awareness of greenhouse effect 1=Yes 0=No

343 76

81.51 18.49

Fig. 4. PM2.5 emission and its effect on person.

Table 3 An example of choice sets. Attribute

Plan 1

Plan 2

Plan 3

Status quo

PM2.5 Acid rain CO2 Bill Please choose

good non 6–10% ¥25 Plan 1

Excellent Non 1–5% ¥25 Plan 2

excellent light 6–10% ¥25 Plan 3

Moderate pollution relatively severe 0 0 Status quo

The second section recorded a set of statements regarding the respondent’s socioeconomic information. Generally, gender, age, education and income are usually used as co-variables to explore heterogeneity in preferences. In this study, besides the above variables, we also consider environmental awareness, such as “do you know the earth is becoming warmer and warmer”. This is because it is thought that people’s values affect his/her preferences and behavior choice (Ellis et al., 2007), and environmental awareness will impact respondent’s values in environmental fields. In the third section, we explained what the environment would look like and what people would suffer under various levels of attributes of environmental quality. For example, we reminded respondents to pay attention to the following information before they filled in the questionnaire (Fig. 4). The fourth section consists of the various alternatives. Alternatives were designed by combining the four attributes given in Table 1 based on the different levels of attributes. An example of the choice sets is presented in Table 3.

N=411

orthogonal main effects design (Margolin, 1968) to reduce the number of possible combinations of attributes.5 The orthogonal main effects design was implemented by using the SPSS 19.0 package and 18 choice sets were finally chosen. Given that previous studies (Susaeta et al., 2011; Lee and Yoo, 2009) have found that each respondent can fill four to six choice sets at most, we divided 18 choice sets into 3 questionnaires (titled Questionnaire1, Questionnaire2, and Questionnaire 3); each questionnaire had 6 choice sets. After removing the outliers and unfinished questionnaires, we were left with 411 questionnaires yielding 2466 observations (411 effective respondents×6 choice sets) available for analysis. Sampling was conducted across the east (35.28% of the total), west (24.09% of the total), and central (40.63% of the total) regions of China, including urban and rural areas (41.36% and 58.64% of the total, respectively). The population sampled was randomly selected with the aim of covering a wide range of demographic factors, including education levels, age, and income levels (Table 2). The questionnaire had four parts. In the first section, we illustrated the investigation purpose and applied the “cheap talk script” method by stressing that “We will select a special place to carry out a pilot project where the power price will be identified by your answer, and in the future such price policy will be implemented in the region where you live”. Since the environmental cost evaluation of China’s coal fired power generation using the CE is based on a hypothetical market, rather than a real market trading behavior. Hence, respondents would have higher WTP for getting clean environment in a questionnaire than in a real world, meaning that hypothesis bias will occur. “Cheap talk script” is first put forward by Cummings and Taylor (1999), and it has been proved to be an effective way to reduce hypothesis bias (Aadland and Caplan, 2006; Brown et al., 2008; Bulte et al., 2005; List et al., 2001).

5. Econometric model As we mentioned in previous section that a CE model includes two parts: random utility function and characteristic utility function. The random utility function is decomposed into the observable element of utility (deterministic component) and the unobservable element of utility (stochastic component) (Eq. (1)). According to Train (2009), the random utility for respondent n to choose alternative j in choice set t can be expressed mathematically as follows:

Unjt = Vnjt + εnjt ∀ j , t ,

(1)

where Unjt is random utility; Vnjt is observable element of utility, and it is the utility of respondent n when he/she choose alternative j in choice set t ; εnjt is unobservable element of utility associated with respondent n and option j . If Unit > Unjt for all i ≠ j in the choice set t , the respondent will choose alternative i over j . Moreover, Vnjt can be expressed as:

Vnjt = V (Xnjt , Sn ) ∀ j , t

(2)

where, Xnjt is the attributes associated with environmental quality for respondent n to choose alternative j in choice set t , and Xnjt represents the characteristic utility function; Sn is the socioeconomic attributes of respondent n . The random utility model can be transformed into a different class of choice models by varying assumptions about the distribution of the error term (Van der Kroon et al., 2014). If the distribution of the error term εnjt typically is assumed to be independently and identically

5 Orthogonal main effects design is one type of fractional factorial design aimed at not letting a single respondent face all choice situations. Based on the prerequisite that there are no interactions between factors, this type of design is able to permit orthogonal and unbiased estimation of all the main effects of the factors studies (Margolin, 1968).

311

Energy Policy 102 (2017) 307–317

X. Zhao et al.

Table 4 Estimated results of MNL, MNL with interaction and RPL. Variable

MNL Coeff (s.e.)

MNL with interactions Coeff (s.e.)

RPL Coeff (s.e.)

ASC1 ASC2 ASC3

−1.583***(0.321) −1.566***(0.330) −1.507***(0.318)

−3.528***(0.401) −3.508***(0.408) −3.448***(0.399)

−3.633***（0.508） −3.701***（0.523） −3.631***(0.509)

Random parameters (normal distribution): only applicable to RPL model PM2.5(Excellent) 1.119***(0.202) PM2.5(Good) 1.086***(0.156) Non-Acid rain 1.212***(0.270)

1.154***(0.205) 1.112***(0.158) 1.262***(0.275)

0.973***(0.349) 0.740***(0.278) 1.351***(0.370)

Non-random parameters: only applicable to RPL model PM2.5(Light) Light Acid rain Moderate Acid rain CO2 reduction (11–20%) CO2 reduction (6–10%) CO2 reduction (1–5%) BILL Male×ASC Age( > =50) ×ASC Age(39–49) ×ASC Age(29–39) ×ASC College or higher×ASC High school×ASC Income( > 50000) ×ASC Income(25000–50000) ×ASC Income(10000–25000) ×ASC No children×ASC Greenhouse×ASC

0.422*** (0.148) 0.910***(0.258) 0.725***(0.280) 0.717*** (0.155) 0.650*** (0.133) 0.378*** (0.128) −0.072*** (0.012) −0.468*** (0.105) 0.704*** (0.225) 0.511*** (0.180) −0.181 (0.175) 1.029***(0.137) 0.426*** (0.159) 0.505*** (0.156) 0.760*** (0.156) 0.679*** (0.160) 1.413***(0.178) 0.338*** (0.125)

0.438***(0.163) 0.998***(0.291) 0.892***(0.321) 0.788***(0.196) 0.720***(0.168) 0.529***(0.167) −0.073***(0.015) −0.480***(0.116) 0.728***(0.249) 0.544***(0.199) −0.210（0.189） 1.094***(0.158) 0.419**(0.176) 0.546***(0.171) 0.817***(0.172) 0.737***(0.178) 0.493**(0.238) 0.457**(0.219)

0.412***(0.147) 0.866***(0.254) 0.706***(0.275) 0.669***(0.148) 0.620***(0.126) 0.353***(0.123) −0.069***(0.011)

Heterogeneity in mean PM2.5(Excellent): No children PM2.5(Good): No children PM2.5(Good): Greenhouse Non-Acid rain: No children

1.041***(0.236) 1.038***(0.206) 0.091(0.239) 0.428**(0.174)

Standard deviations of parameter distributions sdPM2.5(Excellent) sdPM2.5(Good) sdNon-Acid rain

1.434***(0.416) 0.984***(0.316) 1.377***(0.341)

Model summary statistics Log-likelihood Restricted log-likelihood McFadden Pseudo R2 AIC/N* FIC/N* BIC/N* HIC/N* Number of respondents Number of observations

−3353.92

−3214.23

2.735 2.735 2.766 2.746 411 2466

2.631 2.631 2.687 2.651 411 2466

−3176.38.32 −3413.06 0.069 2.610 2.610 2.695 2.640 411 2466

Note: The value in brackets is the standard value. *“N” refers to the number of respondents, **represent significant at 5% and 1%, respectively, ***represent significant at 5% and 1%, respectively.

affecting the WTP by revealing the differences compared to an average respondent (Gordon et al., 2001; Han et al., 2008; Lim et al., 2014; Louviere et al., 2000; Train, 2009). Thus, the second method in this study is the MNL model with interactions, which is applied to represent heterogeneity in choice modeling. Although the MNL model and MNL model with interactions are commonly used to estimate choice probabilities, the two models require the restrictive assumption that choices are independent of irrelevant alternatives (IIA) (Borchers et al., 2007). Furthermore, the representation of preference heterogeneity is relatively crude based on the MNL model with interactions (Colombo et al., 2008). As a response to the weakness of MNL and MNL with interactions, the Random Parameter Logit (RPL) model has been developed (Colombo et al., 2008). The RPL model is a highly flexible model that can approximate any random utility model and obviate the three main limitations of MNL model by allowing for random taste variation, unrestricted

distributed (IID) extreme value distribution for all i , the function of choice probability can be expressed as:

Pnit =

exp(Vnit ) ∑ j exp(Vnjt )

(3)

Eq. (3) describes the multinomial logit (MNL) model, which is simple and most widely used choice model (Train, 2009); meanwhile, it is the basic of advanced models established (Hensher et al., 2005). Therefore, we employed the MNL model as the first method in the analysis. To explain preference heterogeneity and WTP variations among individuals, it is necessary to take some individual specific variables – socioeconomic, attitude, and past experience – into account (Lim et al., 2014). The interactions of ASC (alternative specific constant) with socioeconomic variables provide more information about the factors 312

Energy Policy 102 (2017) 307–317

X. Zhao et al.

calculated as follows (Hanemann, 1984; Bennett and Russell, 2001):

substitution pattern, and correlation in unobserved factors over time (Train, 2009). Therefore, we apply the RPL approach as the third method to make a further analysis. The typical formulation of the RPL model decomposed (1) into an unobserved preference heterogeneity component and a deterministic component (Colombo et al., 2008; Yoo and Ready, 2014). The utility for respondent n choosing alternative j in choice set t is specified as:

Unjt = βXnjt + ηnXnjt + εnjt ∀ j , t

TWTP = − (1/ βcos t )[ ln

⎛ exp(β X ) ⎞ n nit ⎟⎟ f (β )dβ ⎝ j exp(βnXnjt ) ⎠

∫ ⎜⎜ ∑

(4)

∑ exp(V0 )]

(7)

6. Results and discussions 6.1. The results based on MNL and MNL with interaction models We used Nlogit 5.0 to estimate the results. The MNL and MNL with interaction model results can be viewed in Table 4. In the MNL model with interactions, we added interaction variables of respondents’ characteristics with an alternative-specific constant (ASC), which represents a dummy variable for the respondent choosing the environmental improvement alternatives. The outcome of the AIC/N and BIC/ N value (“N” is the number of respondents) indicates that the MNL model with interactions results is a better model fit than the MNL model. The results of the MNL model and the MNL model with interactions show that all of the coefficients of attributes are statistically significant positive as expected, which indicates that respondents have a WTP for a greener environment. This result is consistent with Sun et al. (2016a). The premium (“Bill”) is statistically significant at a 1% level and negative, which means that the probability of paying extra for environmental improvement decreased as the premium increased and the environment improved, and such result supports the reasonability and credibility of our conclusions. As previously discussed, we have two principle concerns in this study. First, we concern the impact of environmental awareness on respondent’s preferences for green development of coal-fired power industry. We found respondents who demonstrate a strong environmental awareness, through indicating they are aware of the greenhouse effect, are more likely to pay a higher premium. This result is consistent with the argument of Ellis et al. (2007), which implies that it is important to make environmental deterioration widely recognized. Second, we concern the different preferences of Chinese respondents for various kinds of emissions. Table 4 shows that in the MNL with interactions model, the coefficients of PM2.5 emission reduction to a level of excellent and good air quality are 1.154 and 1.112 respectively, the coefficient of acid rain reduction to no, Light, and moderate acid rain are 1.262, 0.910 and 0.725 respectively; while the coefficients of CO2 reduction by 11–20%, 6–10%, and 1–5% are 0.737, 0.650 and 0.378 respectively. These results indicate that the respondents prefer to reduction in PM2.5, acid rain (reduction in SO2 and NOx) than CO2 reduction. It seems that Chinese respondents pay more attention on recent damages/loss caused by acid rain and air pollution, and they pay relatively less attention on the future damages/loss caused by GHG. However, from a long run perspective, the damages caused by climate change stemmed from GHG emission increase to a large extent will be much more serious than acid rain and air pollution. Hence, China’s government should educate the public to concern CO2 and other kinds of GHG emission reduction as much as the acid rain and PM2.5. Furthermore, the other five important findings are: (1) The coefficient for men is significant at a 1% level and negative, which indicates that men show more reluctance to support environmental improvement than women. There are mixed results in previous studies related to gender variable; while Wiser (2007) and Susaeta et al. (2011) showed the same finding of female has the willingness to pay higher amounts for green development in power industry, Aravena, et al. (2012), and Diaz-Rainey and Ashton (2007) found no significant effect of female or male on the price premium for cleaner production in power industry; Zarnikau (2003) even found male has the willingness

(5)

In sum, we will apply three different econometric models (MNL, MNL with interaction, and RPL) to analyze respondents’ preferences facing environmental improvement. However, we still faced the challenge of how to compare the three different models. A comparison between MNL, MNL with interaction, and RPL cannot be carried out using conventional log likelihood ratio tests because the models are non-nested (Colombo et al., 2008). Hence, according to literature reviews (Cerwick et al., 2014; Jaeger and Rose, 2008; Valck et al., 2014; Yoo and Ready, 2014), we used the Minimum Akaike Information Criterion (AIC), the finite sample version of AIC (FIC), the Minimum Bayesian Information Criterion (BIC) criteria and the Hannan and Quinn Information Criterion (HIC) to compare models applied in this study to determine which one fits better. Generally, AIC and BIC values are used to judge a model fitting effect. The lower their values, the better of a model’s fit. If the comparison results of AIC and BIC conflict against each other, the magnitude of the four values of AIC, FIC, BIC, and HIC will be considered together instead. The model with the greatest number of low magnitude values for the four factors above is taken as to be fit best. Moreover, we will calculate the WTP for green development of China’s power industry. Such evaluation can also be worked by the CE model (Hanley et al., 1998). WTP is calculated as ratios of two parameters: the coefficient of environmental attributes βattribute , and the coefficient of bill attribute βcos t (Hensher et al., 2005). Both parameters should be statistically significant; if not, no meaningful WTP measure can be obtained (García-Llorente et al., 2012). From Hanemann (1984, 1983), the marginal WTP is specified as:

MWTPattribute = − (βattribute / βcos t )

ln

where, βcos t is the estimated coefficient of cost, V1 represents the utility of any alternative option, and V0 represents the utility of the reference alternative.

where, Xnjt is a vector of observed variables that include choice attributes and socioeconomic characteristics of respondents; β is the vector of coefficients associated with these variables representing the average of taste in the population; ηn represents the deviation of individual taste from average taste in the population, and εnjt is IID extreme value distribution. Under this structure, each respondent has his/her own vector of parameter βn , which deviates from the population mean β by the vector ηn . The usual form of random parameter logit probability is describe in Eq. (4) (Train, 2009). Here, f (β ) is the density function of βn :

Pnit =

∑ exp(V1) −

(6)

where, βattribute is the coefficient of environmental attributes (CO2 emission reduction, PM2.5 emission reduction, and acid rain reduction——SO2 and NOx reduction). βcos t is the coefficient of bill/cost attribute. We estimate the marginal WTP of three attributes: PM2.5, Acid rain, and CO2, along with the 95% confidence intervals estimated using the procedure proposed by Krinksy and Robb (1986). Then, the marginal WTP can be estimated by taking the average over the sample distribution of WTPattribute coefficients. Beyond the marginal WTP estimations for single environmental attributes, it is required to estimate the compensation surplus or welfare change in three future scenarios (Table 4) in comparison to the status quo. We calculated the amount of money required to reach a higher environmental quality, comparing the utility of any alternative option to the reference alternative. It is called total WTP, which was 313

Energy Policy 102 (2017) 307–317

X. Zhao et al.

around mean are tested using the covariates GHG and No Children. The overall model is statistically significant and the estimated result is presented in Table 4. Since three types of IC/N values (AIC/N, FIC/N, and HIC/N) for RPL model are higher than that for MNL with interaction model. Hence, it indicates that RPL generates a better model fit than MNL with interaction. The results from RPL model are consistent with the results from the MNL model with interactions, which makes our conclusions stronger and more credible. Moreover, the spread or dispersion of all random parameters in Table 4 (PM2.5-Excellent, PM2.5-Good, Non-Acid rain) is statistically significant, suggesting that there exists heterogeneity around the mean of the three random parameters. Especially, it is showed that the heterogeneity in the mean parameter estimate for the PM2.5 (Excellent)×No children of 1.041 suggests that across the sampled population, respondents’ sensitivity to improve PM2.5 to excellent air quality increased if the respondents have no children. That is, respondents without children are more sensitive to PM2.5 reduction to an air quality excellent level. The statistically significant PM2.5 (Good)×No children parameter of 1.038, suggests that if respondents have no children, the respondents tend to improve PM2.5 to good air quality. The heterogeneity in the mean parameter estimate for the Non-Acid rain×No children of 0.428 suggests that respondents without children demonstrate willingness to pay higher premium for improving an acid rain to a non-acid rain.

to pay higher amounts than female for green development in power industry. (2) It is expected a positive income effect for green development in China’s coal-fired power industry with higher income respondents would be willing to pay higher for emission reduction. The expectation is true for the respondents whose income level is between 10,000 Yuan to 50,000 Yuan. However, we found respondents with more than 50,000 Yuan of per capita annual income will pay less for environmental improvement than those with lower household income (from 10,000 Yuan to 50,000 Yuan). We postulate that one reason may be that rich respondents are better able to avoid the damage of environmental pollution than low-income respondents. For example, they can move to clean regions and neighborhoods even move out of the country. Mendelsohn et al. (2006) also stated that if low income respondents expect to suffer disproportional damages from climate change, they may be more likely to pay more for green development in power industry. This reflects the findings of Susaeta et al. (2011), which stated that “a positive WTP were more likely to be middle income…”(P1114). (3) The higher the respondents’ education level is, the higher the premium they are willing to pay. This result is consistent with the studies of Susaeta et al. (2011), Aravena et al. (2012), and Sun et al. (2016b). Susaeta et al. (2011) argued that the respondents holding a bachelor degree (“Bachelor”) or higher as compared to respondents with only a high school education or less will pay higher premium for green electricity increased the Southern United States. Aravena et al. (2012) stated that the WTP for environmental price premiums increases with higher levels of education in Chile. Sun et al. (2016b) concluded that the higher the educational level is, the more the respondents are willing to pay for the smog mitigation in China. (4) In general, older respondents have a higher probability of contributing to environmental improvement than younger respondents. The results of the MNL model with interactions show that all of the interaction terms are statistically significant except the interaction between Age 29 and 39. The reasons that the interaction between Age 29 and 39 is not statistically significant are two-fold: one is that respondents aged 29 and 39 are facing high costs of living, and they have a limited ability to bear an environmental improvement cost. The other is that these respondents have comparatively stronger physiques and therefore have more physical ability to endure pollution than other respondents. Our result is different from Aravena et al. (2012) and Susaeta et al. (2011). Aravena et al. (2012) argued that younger people are willing to pay more for green development in power industry than older people. This is because “younger people are more likely to visit Chilean Patagonia than older people are; thus, they would be willing to pay more in order to keep it in its pristine state” (P1221). If the reason is true, it means that it is environment awareness rather than age differences that played its role on respondent’s preferences. Susaeta et al. (2011) concluded that respondents’ ages were not statistically significant to pay a premium for green development in power industry. (5) Finally, an interesting finding is that people without children are more inclined to pay for environmental improvement than those with children, and this result is opposite to our original expectation. However, this can possibly be explained by the fact that people without children have less financial burden than those with children and thus also have higher WTP.

Table 5 Marginal WTP values for MNL, MNL with interactions and RPL. MNL

MNL with interactions

RPL

PM2.5(Excellent)

16.258*** (12.064, 20.453)

16.105*** (11.850, 20.359)

13.282*** (5.356, 21.207)

PM2.5(Good)

15.7793*** (11.188, 20.371)

15.516*** (11.148, 19.885)

10.110*** (3.193, 17.027)

PM2.5(Light)

5.983*** (1.618, 10.348)

5.894*** (1.788, 10.000)

5.982*** (1.608, 10.356)

Non-Acid rain

17.612*** (10.310, 24.914)

17.614*** (10.625, 24.602)

18.452*** (9.120, 27.785)

Light Acid rain

12.580*** (4.986, 20.173)

12.694*** (5.435, 19.952)

13.628*** (6.098, 21.157)

Moderate Acid rain

10.254** (2.279, 18.228)

10.124** (2.232, 18.016)

12.178*** (3.993, 20.363)

CO2 reduction (11–20%)

9.727*** (4.604, 14.849)

10.011*** (4.935, 15.088)

10.757*** (5.037, 16.477)

CO2 reduction (6–10%)

9.010*** (4.082, 13.939)

9.078*** (4.345, 13.810)

9.836*** (4.565, 15.108)

CO2 reduction (1–5%)

5.136** (0.729, 9.544)

5.269** (1.036, 9.501)

7.223*** (2.084, 12.362)

6.2. RPL estimation results An important ability of random parameter logit (RPL) model is to determine the possible sources of any preference heterogeneity that may exist around the mean population parameter by analyzing the interaction of each random parameter with questionable variables (Hensher et al., 2005). In the RPL model, PM2.5 (Excellent), PM2.5 (Good) and Non-Acid rain are specified as random parameters with a normal distribution. The possible sources of heterogeneity of RPL

**represent significant at 5% and 1%, respectively, ***represent significant at 5% and 1%, respectively.

314

Energy Policy 102 (2017) 307–317

X. Zhao et al.

6.3. WTP analysis

Table 6 Comparison of household electricity price between China and some other countries in 2014 (US$/kw h).

The estimated marginal WTP values for the attributes in the CE are presented in Table 5. All of the marginal WTPs in Table 5 are significantly positive, indicating that respondents assigned positive values to environmental improvements. For example, in the RPL model results, the marginal WTP per household for PM2.5 improving from moderate pollution (status quo) to excellent air quality, good quality and light pollution is 13.282 Yuan, 10.110 Yuan, and 5.982 Yuan, respectively, per month; for acid rain, from severe acid rain to non-acid rain, light acid rain, and moderate acid rain is 18.452 Yuan, 13.628Yuan, and 12.178Yuan, respectively, per month; and for CO2 from no emission reduction to high emission reduction (11–20% emission reduction), middle emission reduction (6–10%), and light emission reduction (1–5%) is 10.757Yuan, 9.836Yuan, and 7.223 Yuan, respectively, per month. The above results indicate that, compared with the WTP for CO2 emission reduction, respondents intend to pay more for the emission reduction of PM2.5 and SO2, with the highest intent to pay for emission reduction of SO2. This reflects our sample make-up that skews toward rural areas: 241 samples, or 58.64% of the total, were collected from rural areas; while 170 samples, or 41.36% of the total, were collected from urban areas. Habitants in rural areas suffer more from acid rain those in urban areas, as acid will damage soil and crops. The total statistical WTP for improving the status quo to the best environmental situation, estimated by Eq. (7), is approximately 40 Yuan per month, or 480 RMB Yuan for year for average household. Referring to the data collected from China National Energy Administration, the electricity consumption for resident living is 6928 × 108 kW h per year. Meanwhile, according to the China Household Development Report (2014), there were 430 million households in China in 2014. Hence, the annual electricity consumption for one household is 1611 kW h per year or 134 kW h per month. Therefore, the total WTP for improving status quo to the best environmental situation is 0.30 Yuan per kW h, which indicates that the external cost of coal-fired power generation is 0.30 Yuan/kW h in China.

Country

Household electricity price

Country

Household Electricity price

Denmark Germany Italy Ireland Portugal Austria UK Japan Netherlands Belgium

0.403 0.395 0.307 0.305 0.292 0.267 0.256 0.253 0.252 0.244

New Zealand Greece Luxembourg Sweden Slovak Slovenia Switzerland France …… China

0.236 0.236 0.218 0.214 0.214 0.213 0.209 0.207 …… 0.091

Data source: IEA

plants in EU countries, arriving at 0.23–0.34Yuan/kW h (Söderholm and Sundqvist, 2003). IEA/NEA analyzed the environmental cost of various types of power generation in 19 countries and found the environmental cost of coal-fired power plants to be 0.20–0.45Yuan/ kW h (Sundqvist, 2004). Our result is within the scope of the above two studies. Based on these results, we have two arguments. First, our result is reasonable because it is close or within the results. Second, our result is slightly low compared to the higher-limit results in the 19 countries and European Union. This is caused by the three reasons: (1) China’s economic level is lower than that of some developed countries e.g., EU countries, hence, China’s respondents pay a lower premium for environmental improvement than those in some countries in the EU or other developed countries. (2) China’s household electricity fee is less than the fee found in other countries (Table 6) and given the premium for environmental improvement is no more than the 50% of the total electricity fee, a low electricity fee will translate into a low premium. (3) the environmental consciousness of China’s respondents is relatively low in comparison with that found in developed countries. Although fog and haze have covered many areas in China for a large period of time each year, the awareness of PM2.5 is low for many respondents. For example, the questionnaires shows that 54.8％ of respondents knew little, or nothing, about PM2.5; and 37.4％ of respondents thought that their daily action had little impact on the environment. Hence, the relatively low environmental consciousness led to a lower WTP for environmental improvement than what is found in developed countries.

6.4. Validity check 6.4.1. Theoretical validity check Theoretical validity check refers to testing the consistency between the results in this study and the original hypotheses. According to electricity consumption data for resident living in 2014 (6928 × 108 kW h) (the data is collected from China National Energy Administration), and the household number in 2014 (430 million) (Note: The data is collected from China Household Development Report), the electricity consumption for the average household in China can be calculated (134 kW h in 2014). Meanwhile, we know that the residential electricity fee in China is approximately 0.6 Yuan/kW h. Hence, every household electricity fee in 2014 was 80.4 Yuan. Our finding that respondents will pay an extra 40 Yuan for the best situation of environmental improvement is reasonable because the additional premium accounts for approximately 50% of the total electricity fee per month. Therefore, our findings are consistent with our original hypotheses (no more than 50%).

7. Conclusion and policy implications The main objective of this study was to assess the preferences of China’s households for green development (emission reduction) in coal-fired power industry and evaluate quantitatively the external environmental cost of China’s coal-fired power plants based on a choice experiment designed. We presented the results from applying three types of models (MNL, MNL with interaction, and RPL). All of the models’ results indicate that Chinese respondents had willingness for paying electricity premium to improve environmental externality of coal-fired power plants. It is concluded that environmental awareness has significant impact on respondents’ preferences for green development in China’s power industry, which supports the argument of Ellis et al. (2007) that people’s value will affect his/her attitude and behavior towards green development. Meanwhile, we found that China’s respondents paid the most attention to the reduction of acid rain (SO2 and NOx), followed by PM2.5, and CO2. These results are consistent with China’s current environmental reality, which is facing very serious acid rain damage. China’s acid rain area has the biggest scope among the three large acid

6.4.2. Standard validity check Mahapatra et al. (2012) calculated the environmental cost of coalfired power plants in India based on the dose-response model. The result obtained using that model is 0.26 Yuan/kW h, which is lower than but still close to our findings. Georgakellos (2010) evaluated the environmental cost of coal-fired power plants in Greece based on Ecosense LE method and obtained the result of 0.264 Yuan/kWh, which is also close to our result. European Commission used the ExternE method to calculate the environmental cost of coal-fired power 315

Energy Policy 102 (2017) 307–317

X. Zhao et al.

is not only for collecting fund for green development of power industry, but also for arousing people’s awareness to protect environment. Such awareness is important to formulating and strengthening the preferences for green development. Lastly, government awareness campaigns pertaining to environmental protection should be enhanced, with special attention on awareness campaigns related to CO2 emission control. This is not only because CO2 emissions cause more extensive damages than pollutant emissions in the long run, but also because Chinese respondents have paid insufficient attention to CO2 as PM2.5, SO2 and NOx. One concern with our results could lie with potential bias that comes with the use of a CM. The data used in the CM were collected from questionnaires, which involved imagined situations instead of real situations. Although we used the “cheap talk” method to try to close the potential bias gap between imaginary and real world scenarios, the uncertainties in this study’s results still exist and one should be cautious in quoting specific results.

rain regions of the world (Europe, America, and Asia), and the south of the Yangtze River comprises a strong acid rain center. Moreover, China’s acid rain is much more intense than that of the European Union and North America. Hence, the respondents consider the environmental cost of SO2 as the greatest cost, even greater than that of PM2.5 emission. On the other hand, the finding that China’s respondents prefer the reduction of acid rain and PM2.5 to CO2 reduction indicates that it is urgent for China’s government to arouse the public’s concern about CO2 and other kinds of GHG emission reduction through education and more rigid regulation. Moreover, it is found that female respondents have the willingness to pay higher amount for environmental improvement than male respondents do; the respondents with highest income have lower willingness to pay premium for green development of power industry than those with middle income. The reason is probably that the respondents with highest income would suffer less damage since they are more economically resilient to environmental deterioration. Furthermore, it is concluded that education level has a significant positive impact on respondent’s preference for environmental improvement; and generally the age variable will significantly affect respondents’ willingness to pay electricity premium for green development. Another interesting finding is that people without children have stronger willingness to pay for environmental improvement than those with children. In sum, we can conclude that the highly educated middle-income women with no children will express significant concerns about environmental change and pay a higher premium in electricity prices for environmental improvement than any other demographic groups. Finally, the WTP calculation results based on the RPL model show that the WTP per household for the best situation of environmental improvement is 40 RMB Yuan per month (480 RMB Yuan per year) or 0.30 yuan per kW h. The magnitude of externalities of coal-fired generation is less than that of nuclear power. According to Sun and Zhu (2014), the environmental cost of nuclear power plant in China is US$ 80.106–116.604 yearly for an average household (based on the data collected in 2013). The average RMB Yuan-US dollar exchange rate was 6.1932,6 meaning that the environmental cost of nuclear power plant is around 496–722 RMB Yuan. Our results are helpful for policy makers in guiding the power industry towards green development. First of all, the environmental cost of coal-fired power should be concerned in making an economic evaluation for building a coal-fired power plant. Currently, without considering environmental cost, the generation cost of coal-fired power in China is 0.373Yuan/kW h, and the generation cost of wind power is 0.44Yuan/kW h (Zhao and Wang, 2014). However, if the environmental cost of coal-fired power is considered, the generation cost of coal-fired power will increase to 0.673 Yuan/kW h. Then, the advantages of wind power or other kinds of renewable energy will surface. Secondly, the WTP power price (green power price) can be implemented in pilot cities with developed economies. This study’s results show that Chinese respondents have the willingness to pay premium for cleaner electricity supply. This indicates that in China, specifically in certain developed cities such as Beijing, Shanghai, Shenzhen, the policy of green power price can be implemented. With the increase of per capita income (the per capita income in Shanghai and Beijing has reached 47710 Yuan and 43910 Yuan respectively in 20147), China’s residents have paid increasing attention to environmental improvement. The implementation of green power price policy

Acknowledgments We would like to thank the National Natural Science Foundation of China for partially funding this study (Grant nos. 71373078 and 71573273), as well as the fund from China University of PetroleumBeijing (Grant no. 2462014YJRC052). Meanwhile, we appreciate professor Sufang Zhang’s correction on this work. References Aadland, D., Caplan, A.J., 2006. Cheap talk reconsidered: new evidence from CVM. J. Econ. Behav. Organ. 60, 562–578. Aravena, C., Hutchinson, W.G., Longo, A., 2012. Environmental pricing of externalities from different sources of electricity generation in Chile. Energy Econ. 34, 1214–1225. Bachmann, T.M., van der Kamp, J., 2014. Environmental cost-benefit analysis and the EU (European Union) Industrial Emissions Directive: exploring the societal efficiency of a DeNOx retrofit at a coal-fired power plant. Energy 68, 125–139. Bennett, J., Russell, B., 2001. The Choice Modelling Approach to Environmental Valuation. Edward Elgar Publishing, Cheltenham UK. Borchers, A.M., Duke, J.M., Parsons, G.R., 2007. Does willingness to pay for green energy differ by source? Energy Policy 35, 3327–3334. Brown, T.C., Kingsley, D., Peterson, G.L., Flores, N.E., Clarke, A., Birjulin, A., 2008. Reliability of individual valuations of public and private goods: choice consistency, response time, and preference refinement. J. Public Econ. 92, 1595–1606. Bulte, E., Gerking, S., List, J. a., De Zeeuw, A., 2005. The effect of varying the causes of environmental problems on stated WTP values: evidence from a field study. J. Environ. Econ. Manag. 49, 330–342. Carlsson, F., Frykblom, P., Lagerkvist, C., 2005. Using cheap talk as a test of validity in choice experiments. Econ. Lett. 89 (2), 147–152. Cerwick, D.M., Gkritza, K., Shaheed, M.S., Hans, Z., 2014. A comparison of the mixed logit and latent class methods for crash severity analysis. Anal. Methods Accid. Res. 3–4, 11–27. Colombo, S., Hanley, N., Louviere, J., 2008. Modelling preference heterogeneity in stated choice data : an analysis for public goods generated by agriculture. Sergio Colombo Stirling Economics Discussion Paper 2008–28 December 2008. Online at 〈http:// www.economics.stir.ac.uk〉. Cummings, R.L., Taylor, L., 1999. Unbiased value estimates for environmental goods: a cheap talk design for the contingent valuation method. Am. Econ. Rev. 89, 649–665. Diaz-Rainey, I., Ashton, J., 2007. Characteristics of UK Consumers' Willingness to Pay for Green Energy. Working Paper. Social Sciences Research Network. Ek, K., Persson, L., 2014. Wind farms – where and how to place them? A choice experiment approach to measure consumer preferences for characteristics of wind farm establishments in Sweden. Ecol. Econ. 105, 193–203. Ellis, G., Barry, J., Robinson, C., 2007. Many ways to say ‘No’, different ways to say ‘Yes’, applying q-methodology to understand public acceptance of wind farm proposals. J. Environ. Plan. Manag. 50 (4), 517–551. Faaij, A., Meuleman, B., Turkenburg, W., Van Wijk, A., Bauen, A., Rosillo-Calle, F., Hall, D., 1998. Externalities of biomass based electricity production compared with power generation from coal in the Netherlands. Biomass. Bioenergy 14, 125–147. García-Llorente, M., Martín-López, B., Nunes, P. a.L.D., Castro, a.J., Montes, C., 2012. A choice experiment study for land-use scenarios in semi-arid watershed environments. J. Arid Environ. 87, 219–230. Georgakellos, D., 2010. Impact of a possible environmental externalities internalisation on energy prices: the case of the greenhouse gases from the Greek electricity sector. Energy Econ. 32, 202–209. Gordon, J., Chapman, R., Blamey, R., 2001. The Chioce Modelling Approach to Environmental Valuation. Edward Elgar, Cheltenham, UK. Han, S.Y., Kwak, S.J., Yoo, S.H., 2008. Valuing environmental impacts of large dam

6 Data source: Statistics of average RMB Yuan to other main foreign currencies exchange rate (1985–2013), http://blog.sina.com.cn/s/blog_416ba4c90102ebfa.html. 2014, 5. 2. Last access 10.07.16. 7 Data source: Salary Development Report, (2013–2014). Ministry of Human Resources and Social Security (HRSS). http://www.cnrencai.com/salarinfo/147193.html, Last access 05.03.16.

316

Energy Policy 102 (2017) 307–317

X. Zhao et al.

change on rich and poor countries. Environ. Dev. Econ. 11 (2), 159–178. Sáez, R.M., Linares, P., Leal, J., 1998. Assessment of the externalities of biomass energy, and a comparison of its full costs with coal. Biomass Bioenergy 14, 469–478. Söderholm, P., Sundqvist, T., 2003. Pricing environmental externalities in the power sector: ethical limits and implications for social choice. Ecol. Econ. 46, 333–350. Sun, C.W., Zhu, X.T., 2014. Evaluating the public perceptions of nuclear power in China: evidence from a contingent valuation survey. Energy Policy 69, 397–405. Sun, C.W., Yuan, X., Xu, M.L., 2016a. The public perceptions and willingness to pay: from the perspective of the smog crisis in China. J. Clean. Prod. 112, 1635–1644. Sun, C.W., Yuan, X., Yao, X., 2016b. Social acceptance towards the air pollution in China: evidence from public's willingness to pay for smog mitigation. Energy Policy 92, 313–324. Sundqvist, T., 2004. What causes the disparity of electricity externality estimates? Energy Policy 32, 1753–1766. Susaeta, A., Lal, P., Alavalapati, J., Mercer, E., 2011. Random preferences towards bioenergy environmental externalities: a case study of woody biomass based electricity in the Southern United States. Energy Econ. 33, 1111–1118. Tang, J., 2011. Market structure and environmental pollution externality Govername. China population. Resour. Environ. 21, 1–4, (in Chinese). Train, K., 2009. Discrete Choice Methods with Simulation 2nd ed. Cambridge University Press, New York. Turner, R.K., Pearce, W., Bateman, I, D., 1994. Environmental Economics. Harvester Wheatsheaf, London. Valck, J. De, Vlaeminck, P., Broekx, S., Liekens, I., Aertsens, J., Chen, W., Vranken, L., 2014. Landscape and Urban Planning Benefits of clearing forest plantations to restore nature ? Evidence from a discrete choice experiment in Flanders, Belgium. Landsc. Urban Plan. 125, 65–75. Van der Kroon, B., Brouwer, R., van Beukering, P.J.H., 2014. The impact of the household decision environment on fuel choice behavior. Energy Econ. 44, 236–247. Wiser, R.H., 2007. Using contingent valuation to explore willingness to pay for renewable energy: a comparison of collective and voluntary payment vehicles. Ecol. Econ. 62 (3), 419–432. Xepapadeas, P., 1992. Environmental policy design and dynamic nonpoint-source pollution. J. Environ. Econ. Manag. 23, 22–39. Yoo, J., Ready, R.C., 2014. Preference heterogeneity for renewable energy technology. Energy Econ. 42, 101–114. Zarnikau, J., 2003. Consumer demand for "green energy" and energy efficiency. Energ. Policy. 31 (15), 1661–1672. Zhang, J., 2007. Discussion on characteristic index of steam coal used in power generation. Coal Qual. Technol. 3, 26–28, (in Chinese). Zhao, X.L., Wang, S.H., 2014. Economic evaluation of wind power generation based on benefits of carbon dioxide emission reduction. Electr. Power 47 (8), 154–160. Zhao, X.L., Zhang, S.F., Fan, C.Y., 2014. Environmental externality and inequality in China: current Status and future choices. Environ. Pollut. 190, 176–179. Zhao, X.L., Yin, H.T., Zhao, Y., 2015. Impact of environmental regulations on the efficiency and CO2 emissions of power plants in China. Appl. Energy 149 (1), 238–247.

construction in Korea: an application of choice experiments. Environ. Impact Assess. Rev. 28, 256–266. Hanemann, W.M., 1983. Marginal welfare measures for discrete choice models. Econ. Lett. 13, 129–136. Hanemann, W.M., 1984. Welfare evaluations in contigent valuation experiments with discrete response. Am. J. Agric. Econ. 66, 332–341. Hanley, N., Wright, R.E., Adamowicz, V., 1998. Using choice experiments to value the environment. Environ. Resour. Econ. 11, 413–428. Hanley, N., Mourato, S., Wright, R.E., 2002. Choice modelling approaches: a superior alternative for environmental valuation? J. Econ. Surv. 15 (3), 435–462. Hensher, D.A., Rose, J.M., Greene, W.H., 2005. Applied ChoiceAnalysis: a Primer. Jaeger, S.R., Rose, J.M., 2008. Stated choice experimentation, contextual influences and food choice: a case study. Food Qual. Prefer. 19, 539–564. Kosenius, A.K., Ollikainen, M., 2013. Valuation of environmental and societal trade-offs of renewable energy sources. Energy Policy 62, 1148–1156. Ku, S.J., Yoo, S.H., 2010. Willingness to pay for renewable energy investment in Korea: A choice experiment study. Renew. Sust. Energ. Rev. 14 (8), 2196–2201. Krinksy, I., Robb, A.L., 1986. On approximating the statistical properties of elasticities. Rev. Econ. Stat. 68, 715–719. Lancaster, K.J., 1966. A new approach to consumer theory. J. Political Econ. 74 (2), 132–157. Lee, J.-S., Yoo, S.-H., 2009. Measuring the environmental costs of tidal power plant construction: a choice experiment study. Energy Policy 37, 5069–5074. Li, W., Wei, P., Zhou, X., 2014. A cost-benefit analysis of power generation from commercial reinforced concrete solar chimney power plant. Energy Conversat. Manag. 79, 104–113. Lim, S.Y., Lim, K.M., Yoo, S.H., 2014. External benefits of waste-to-energy in Korea: a choice experiment study. Renew. Sustain. Energy Rev. 34, 588–595. List, John, Gallet, C., 2001. What experimental protocol influence disparities between actual and hypothetical stated values? Environ. Resour. Econ. 20, 241–254. Longo, A., Markandya, A., Petrucci, M., 2008. The internalization of externalities in the production of electricity: willingness to pay for the attributes of a policy for renewable energy. Ecol. Econ. 67, 140–152. Louviere, J., Hensher, D., 1982. On the design and analysis of simulated choice or allocation experiments in Travel Choice Modelling. Transp. Res. Rec. 890, 11–17. Louviere, J., Woodworth, G., 1983. Design and analysis of simulated consumer choice or allocation experiments: an approach based on aggregate data. J. Mark. Res. 20, 350–367. Louviere, J., Hensher, D., Swait, J.D., 2000. Stated choice methods: analysis and application. J. Appl. Econ. 17, 701–704. Mahapatra, D., Shukla, P., Dhar, S., 2012. External cost of coal based electricity generation: a tale of Ahmedabad city. Energy Policy 49, 253–265. Manski, C.F., 1977. The structure of random utility models. Theory Decis. 8, 229–254. Margolin, B.H., 1968. Orthogonal main-effect 23 designs and two-factor interaction aliasing. Technometrics 10, 559–573. Mcfadden, D., 1986. The choice theory approach to market research. Mark. Sci. 5, 275–297. Mendelsohn, R., Dinar, A., Williams, L., 2006. The distributional impact of climate

317