Measurement and Impactors of Tourism Carbon Dioxide Emission Efficiency in China

With tourism carbon dioxide emission efficiency (TCDEE) as an undesired output, this study establishes an index system based on the inputs and outputs of TCDEE and measures the provincial TCDEE of China in 2010–2018, using the epsilon-based measure (EBM). In addition, the impactors of TCDEE were tested by the Tobit model. The main results are as follows: China's TCDEEs had obvious provincial differences. Only six provinces reached the efficient frontier of TCDEE, namely, Beijing, Tianjin, Inner Mongolia, Shanghai, Jiangsu, and Guangdong. The other provinces failed to reach this state, leaving a room for improvement. Most eastern provinces had relatively high TCDEEs, while the central and western provinces had relatively low TCDEEs. In the sample period, the TCDEEs in eastern, central, and western parts all changed in the shape of letter N. The TCDEEs of the eastern part were much higher than those of the central and western parts. According to the results of the Tobit model, TCDEE is clearly enhanced by the urbanization level, strongly inhibited by industrial structure, technical progress, opening-up, and environmental regulation, and not significantly affected by the tourism level.


Introduction
Global warming, a result of the greenhouse effect, brings an unprecedented challenge to our survival and development. China pays much attention to this problem and takes multiple measures to reduce the emissions of carbon dioxide, such as environmental governance, energy saving, and green, lowcarbon technology. Every sector in China attaches great importance to reducing carbon dioxide emissions.
Tourism, a pillar industry of the national economy, is energy-intensive [1]. From the angle of carbon dioxide emissions, traveling is a healthy but high-carbon lifestyle. e carbon dioxide emitted by tourism has become a major inducer of environmental degradation. Statistics show that tourism contributes about 5% of carbon dioxide emitted around the world [2]. In recent years, tourism has developed rapidly at an annual rate of 10% in China. Despite its economic benefits, the extensive growth model of tourism consumes lots of energy and emits more and more carbon dioxide year by year. erefore, the efficiency and sustainable development of regional tourism hinge on the increase of tourism carbon dioxide emission efficiency (TCDEE).
To date, very few scholars have fully evaluated the lowcarbon tourism economy in China [3]. A few researchers have explored the carbon dioxide emission of tourism [4,5] but failed to fully understand the TCDEE in each province. Besides, very few scholars have discussed which factors that greatly affect the TCDEE. To fill up the gap, this paper scientifically evaluates the TCDEE in each province of China and further explores the influencing factors of this efficiency. e research findings help to fully understand the development level of low-carbon economy in tourism and facilitate the formulation of proper carbon reduction policies.

Literature Review
e estimation of tourism carbon dioxide emission has long attracted the attention of the academia. Early on, some scholars recognized that the carbon dioxide emitted by tourism is a major driver of global warming. For example, Scott et al. [6] regarded tourism as one of the main industries emitting greenhouse gases. Peterson and Dubois [7] found that the greenhouse gases generated by tourism are responsible for 4.4% of global warming, and the percentage grows by 3.2% each year in 2005-2035. Many scholars strive to estimate the carbon dioxide emitted by tourism by scientific methods. Currently, tourism carbon dioxide emissions can be measured by two approaches. e first approach is top-down analysis. Taking tourism as a part of national economy, the top-down analysis estimates the carbon dioxide emitted by tourism by the input-output method and social accounting matrix method. rough top-down analysis, Perch-Nielsen et al. [8] estimated the carbon dioxide emitted by domestic tourism of Switzerland as 6.62 Mt. Berners-Lee et al. [9] combined the input-output method and lifecycle process analysis to estimate the greenhouse gas emissions of northwestern England.
e second approach is the bottom-up analysis, which divides the tourism participants into tourists and tourism enterprises.
is approach estimates the carbon dioxide emitted by tourism, using lifecycle process analysis, sampling survey, and carbon footprints. For example, Gössling et al. [10] estimated tourism carbon dioxide emissions from three aspects: transportation, lodging, and tourism activities. Mayor and Tol [11] forecasted the trend of tourism carbon dioxide emissions through scenario analysis, according to index data such as tourist conditions, tourism energy consumption, and tourism income.
TCDEE is an important indicator of the relationship between the development of tourism economy and the carbon dioxide emitted by tourism. e effective measurement of TCDEE provides an important reference for determining the energy-saving and emission reduction level of regional tourism. e existing studies mainly concentrate on the efficiencies of tourism hotels [12,13], travel agencies [14,15], and tourism transportation [16,17]. Some scholars discussed the ecological efficiency of tourism under environmental constraints [18,19].
Because of the difficulty in estimating the TCDEE, the above estimations of TCDEE mainly focus on the national scale. e research on the scale of province, municipality directly under the central government, and autonomous region (hereinafter collectively referred to as the province) is severely lacking. What is worse, the literature on TCDEE mainly focuses on an aspect of tourism, such as tourism hotel and tourism transport. ere is little in the report that discusses the carbon efficiency of the entire tourism industry. In addition, none of the above studies recognizes the benefit of taking TCDEE as an undesired output: reflecting the constraint of environmental factors on tourism development and facilitating the accurate evaluation of the lowcarbon development level of tourism.
To fill up the research gap, this study adopts the bottomup method to estimate tourism carbon dioxide emission and builds an index system for TCDEE containing an undesired output. In addition, the epsilon-based measure (EBM) was adopted to estimate the TCDEE of each province in China, and the TCDEE impactors were verified using the Tobit model.

Methodology
3.1. EBM Model. Data envelopment analysis (DEA) is more adaptable and flexible than stochastic frontier analysis (SFA) because it can handle problems with many inputs and outputs, without needing a production function. In view of this, our research adopts DEA to measure TCDEEs in China. e earliest DEA models have constant or variable returns to scale [20,21]. e DEA model with constant returns to scale assumes that the returns to scale of inputs remain unchanged. However, this assumption goes against the reality, resulting in wrong efficiency measurement. Some scholars improved the DEA model with constant returns to scale into a DEA model with variable returns to scale. But the improved model still seeks the maximal output, failing to consider undesired outputs, which are inevitable in actual production. Hence, the improved model may seriously under/overestimate the efficiency. In addition, the above traditional DEA models ignore nonradial slack variables [22].
To overcome the defects of traditional radial DEA models, Tone [23] proposed the slack-based measure (SBM) model, which soon gained common recognition in the academia.
e greatest advantage of the model is the consideration of both undesired outputs and nonradial slack variables. As a result, the efficiency estimated by SBM is close to the actual situation of the production process. Nevertheless, SBM has several shortcomings in efficiency measurement. Firstly, the measured efficiency tends to be smaller than the actual level, for the model overlooks the proportion between the input and output targets and the actual values. e underestimation of efficiency increases the slackness of inputs and outputs, calling for significant changes to inputs and outputs for efficiency improvement. Secondly, the optimal slackness of zero values differs significantly from that of positive values, which further amplifies the error in efficiency measurement.
Tone and Tsutsui [24] extended the SBM into the EBM model, drawing on the merits of the DEA model with constant returns to scale. is hybrid model supports both radial and nonradial inputs and outputs, providing a new way to evaluate the efficiency of decision-makers. e main ideas of EBM are as follows: In a complete production system, there are n decisionmakers responsible for making decisions about production. e k-th decision-maker is denoted by DMU k � (x k , y k , b k ). e system can produce p desired outputs y and q undesired outputs b, after receiving r inputs x. For clarity, the inputs, desired outputs, and undesired outputs are denoted by X � (x 1 , x 2 , . . . , x n ) ∈ R r×n + , Y � (y 1 , y 2 , . . . , y n ) ∈ R p×n + , and us, all possible production scenarios can be expressed as T � (x, y, b): x can produce y and b . en, the EBM model can be established as: where 0 < θ * ≤ 1 is the TCDEE; s . t . is the model constraint; x, y, and b are the input, desired output, and undesired output, respectively; r, p, and q are the total number of inputs, desired outputs, and undesired outputs, respectively; i, o, and u are corresponding to an input or an output; ϕ is the programming parameter for the radial part of the model; σ x is the core parameter containing both the radial effect ϕ and nonradial effect s − i ; σ y and σ b are the core parameters containing nonradial effects s + o and s b− u , respectively; ρ is the nonradial programming parameter of the model; x ik , y ok , and b uk are the i-th input, oth desired output, and u-th undesired output of DMU k , respectively; w − i , w + o , and w b− u are the weights of the i-th input, oth desired output, and u-th undesired output, respectively; s − i , s + o , and s b− u are the slack terms of the i-th input, oth desired output, and u-th undesired output, respectively; t and λ are the decision-maker and it weight, respectively. x rt , y ot , and b ut are the input, desired output, and undesired output of t DMUs; k is the efficiency of the k-th DMU to be estimated.
If s − i , s + o , and s b− u are not zero, then θ * is smaller than 1. In this case, the decision-maker fails to achieve the optimal efficiency and needs to improve the inputs and outputs. If and only if s − i � s + o � s b− u � 0, the decision-maker achieves the efficiency of 1, reaching the efficient frontier. In this case, there is no need to improve inputs and outputs.
3.2. Index System. As a total factor, TCDEE refers to the maximum tourism output and minimum carbon dioxide emissions from constant inputs such as capital, labor, and energy. Referring to Zha et al. [25], this study builds up an index system of TCDEE based on relevant inputs and outputs. Notably, the desired outputs include the total tourism income and total number of tourists, while the undesired output is tourism carbon dioxide emissions. e definition of each index is given in Table 1.

Labor.
Many sectors, such as industry and agriculture, provide tangible products. Meanwhile, tourism provides intangible products called services. e services are provided by the employees of scenic areas, restaurants, travel agencies, etc. In general, the labor input can be best characterized by the number of tourism employees and their effective labor time. However, the effective labor time of tourism employees is not available in relevant statistical yearbooks. is study decides to measure the labor input by the number of tourism employees.

Capital.
In addition to labor, capital is an essential input of tourism development. Specifically, the fixed assets in tourism not only promote infrastructure construction but also improve the services of scenic areas, thereby enhancing tourism quality and attractiveness. Considering data availability, this study takes the original value of the fixed assets in tourism in each province as the capital index. To eliminate the effect of price-induced inflation, the nominal fixed assets in tourism were deflated to the actual fixed assets in tourism with 2005 as the base year, using the fixed asset price index.

Energy.
e tourism energy consumption is not provided in relevant statistical yearbooks. us, this study estimates the consumption with the tourism consumption stripping coefficient. ree sectors are closely associated with tourism, namely, transportation, warehousing, and postal industry; wholesale and retail industry; lodging and catering industry. e energy consumed in these sectors is partially related to tourism and must be stripped out by the right ratio. e tourism energy consumption can be estimated by where E it is the tourism energy consumption of the i-th province in the t-th year; E pq·t is the terminal consumption of the q-th energy in the p-th sector in the t-th year; α q is the coefficient of 10,000 tons of standard coal for the q-th energy; and R it is the tourism consumption stripping coefficient of the i-th province in the t-th year.

Desired Outputs.
e desired outputs reflect the yield of tourism development. Total tourism income and the total number of tourists are the two most important indicators of regional tourism development. Both were selected as desired outputs.
Total tourism income refers to the total operating income of tourism enterprises by providing tourists with tourism products and services. is index directly measures the economic value created by tourism development and sets a standard for measuring the scale of regional tourism. To eliminate the effect of price-induced inflation, the nominal total tourism income was deflated into the actual total tourism income with 2010 as the base year, using the consumer price index. e total number of tourists reflects the attractiveness of regional tourism, as well as the service quality and development scale of tourism in a region. erefore, it is very suitable to take this index as a desired output. For convenience, this study substitutes the total number of tourists with the number of inbound overnight tourists in each province.

Undesired Output.
e carbon dioxide emitted by tourism is not directly given in relevant statistical yearbooks. Drawing on the results of Becken et al. [26], Becken, and Patterson [27], this study employs the bottom-up method to estimate the tourism carbon dioxide emissions, which mainly come from three tourism-related sectors: tourism transportation, tourism lodging, and tourism activities. us, the carbon dioxide emitted by tourism can be estimated by where C is the total carbon dioxide emitted by tourism; C T , C H , and C R are the carbon dioxide emitted by tourism transportation, tourism lodging, and tourism activities, respectively.
where n is one of the four transportation modes, namely, railway, highway, waterway, and civil aviation; N ti is the passenger turnover of the i-th transportation mode in the tth year; g i is the proportion of the i-th transportation mode in passenger turnover; and λ i is the carbon dioxide emission factor of the i-th transportation mode.
where R i and Q i are the mean occupancy rate and number of beds of starred hotels in the i-th province, respectively; β is the carbon dioxide emission factor per bed per night of starred hotels in the i-th province [28].
where h is one of the five tourism activities, namely, sightseeing, vacation, business trip, visiting relatives/friends, and others; Z is the number of tourists; Y h and δ h are the tourist composition and carbon dioxide emission factor of the h-th tourism activity, respectively. After sorting out the relevant literature, the carbon dioxide emission factors of railway, highway, waterway, and civil aviation were set to 72 gCO 2 /pkm, 133 gCO 2 /pkm, 106 gCO 2 /pkm, and 137 gCO 2 /pkm, respectively; the carbon dioxide emission factor per bed per night of starred hotels in the i-th province was set to 2.458 gCO 2 /p visitor-night; the carbon dioxide emission factors of sightseeing, vacation, business trip, visiting relatives/friends, and others were set to 417 gCO 2 /p visitor, 1,670 gCO 2 /p visitor, 786 gCO 2 /p visitor, 591 gCO 2 /p visitor, and 172 gCO 2 /p visitor, respectively.

Tobit Model.
is study focuses on the impactors of TCDEE. Referring to Sun et al. [29], and considering data availability, the authors decided to explore how six factors, out of the various external impactors of tourism carbon emissions, affect TCDEE, namely, tourism level (TL), industrial structure (IS), technical progress (TP), opening-up (OU), urbanization level (UL), and environmental regulation (ES).

Tourism Level (TL).
Tourism level is closely related to regional eco-environment. On the one hand, the level of tourism directly manifests the economic benefits of that industry. On the other hand, the improvement of the regional tourism level leads to better tourism infrastructure and tourism services, which benefit the increase of TCDEE. Referring to Qiu et al. [30], this study measures the tourism level with per capita tourism income. e natural logarithm of the variable was added to the model.

Industrial Structure (IS).
e industrial structure reflects the proportion of primary, secondary, and tertiary industries in the national economy. e energy consumption and pollution emissions of tourism depend closely on the development of tertiary industries, namely, transportation, warehousing, and postal industry; wholesale and retail industry; lodging and catering industry. Hence, this study characterizes the industrial structure with the proportion of the tertiary industry in the gross domestic product (GDP) and thereby analyzes the relationship between the industrial structure and tourism carbon dioxide emissions.

Technical Progress (TP)
. Technical progress has a major impact on energy-saving, emission reduction, and environmental efficiency of a region. e introduction of new techniques, processes, and products to regional tourism enterprises helps to lower the energy consumption of regional tourism and promotes the sustainable growth of tourism. Considering data availability, this study represents technical progress by the ratio of regional expenditure on research and development (R&D) to GDP.

Opening-Up (OU).
As China further opens to the world, more and more tourists are attracted to this country. e growth of inbound tourists stimulates the demand for tourism services and pushes up the emissions of carbon dioxide. Meanwhile, opening-up makes it easier for China to import cutting-edge technology of environmental governance and advanced experience of tourism management. ese technology and experience promote energy conservation and emission reduction in tourism and thus improve TCDEE. Overall, the influence of opening-up over TCDEE remains uncertain and is yet to be verified.
is study converts USD into RMB by the mean exchange rate and characterizes opening-up with the ratio of actual foreign direct investment (FDI) utilized in each province to GDP.

Urbanization Level (UL).
e rising level of urbanization improves the infrastructure in towns and cities and promotes the quality of the service industry, laying a solid basis for tourism development.
e urbanization process also encourages the aggregation of talents and industries, which favors the carbon dioxide reduction of tourism. is study substitutes the urbanization level with the proportion of permanent urban residents in total regional population.

Environmental Regulation (ES).
Tourism, as an energy-intensive industry, inevitably emits a huge sum of carbon dioxide in its development. e government plays an important role in carbon reduction of tourism. By investing more in environmental pollution control, the government forces tourism enterprises to control pollutant discharge and reduce carbon dioxide emissions. If the investment is too high, however, the resource allocation mechanism of the market would be distorted, leading to the green paradox [31]. Considering data availability, this study characterizes environmental regulation with the proportion of the investment on environmental pollution control in GDP.
rough the above analysis, this study sets up a regression model for TCDEE impactors. As the explained variable, TCDEE ranges between 0 and 1. If the ordinary least squares (OLS) method is adopted to estimate TCDEE, the estimation would be biased, for the value range of the TCDEE violates the OLS assumption that the value of explained variable is unlimited. To solve the problem, Tobin [32] designed the Tobit model to handle censored explained variables. Hence, this study establishes a Tobit model for TCDEE impactors, which explains TCDEE from the angles of the tourism level, industrial structure, technical progress, opening-up, urbanization level, and environmental regulation where TCEE is the explained variable of TCDEE; TL, IS, TP, OU, UL, and ES are the tourism level, industrial structure, technical progress, opening-up, urbanization level, and environmental regulation, respectively (Table 2); β 1 , β 2 , β 3 , β 4 , β 5 , and β 6 are the coefficients of the tourism level, industrial structure, technical progress, opening-up, urbanization level, and environmental regulation, respectively; i and t are serial numbers of provinces and years, respectively; and ε is a random error. e coefficient of each explanatory variable represents the degree of influence of that variable over the explained variable.

Data
Sources. e main variables are from EBM and Tobit models. To ensure the completeness and availability of the data on these variables, this study chooses to examine the data about the provinces in China from 2010 to 2018. Note that Hongkong, Macao, Taiwan, and Tibet were excluded because the data on some variables in these provinces were missing for consecutive years. e original data were collected from the statistical yearbooks released by China and its provinces on tourism, energy, transportation, etc. Some missing data were completed through interpolation. e research data were generated no later than 2018. e data from 2018 to 2021 were not included. e reason is that this paper explores many provinces. e original data on tourism energy input and TCDEE indices of all provinces are not complete after 2018. In some provinces, the relevant data are partly missing in 2019-2021. us, the data samples were collected in the time interval of 2010-2018 for data comprehensiveness and availability.

Measured TCDEEs.
Based on the proposed index system for TCDEE, the data on various inputs and outputs were entered to MaxDEA. en, EBM was adopted to measure the provincial TCDEEs in 2010-2018. e results in Table 3 show that China's TCDEEs had obvious provincial differences.
In terms of the mean TCDEE, six provinces, namely, Beijing, Tianjin, Inner Mongolia, Shanghai, Jiangsu, and Guangdong achieved a mean TCDEE of 1 in the sample period. e TCDEEs of these provinces fell on the efficient frontier, reaching the DEA effective state. Geographically, all the six provinces, except for Inner Mongolia, belong to eastern coastal areas. e provinces with high TCDEEs mainly concentrate in eastern coastal areas, thanks to their superior geographical locations, convenient transportation, rich tourism resources, and pursuit of low-carbon tourism. e optimal TCDEE of Inner Mongolia has much to do with its rich tourism resources.
With mean TCDEEs between 0.8 and 0.1, Jilin, Fujian, Liaoning, Heilongjiang, Qinghai, Yunnan, and Zhejiang performed rather well. However, their TCDEEs did not reach the DEA effective state, waiting to be improved. is is because of the redundancy of input production factors: these provinces invest too much manpower, capital, and energy to develop tourism.
Shanxi, Anhui, Hainan, Xinjiang, Guizhou, Shaanxi, Guangxi, Jiangxi, and Hubei ranked in the middle of the country, in terms of the mean TCDEE (0.7-0.8). Located in central and western parts, these provinces witnessed rapid development of tourism in recent years. Nevertheless, the Journal of Environmental and Public Health tourism infrastructure in some areas is constructed repeatedly, and little attention is paid to save energy and reduce emissions of tourism.
at is why, their TCDEEs were way off the efficient frontier. e mean TCDEEs of Hunan, Shandong, Ningxia, Henan, Chongqing, Sichuan, Hebei, and Gansu were smaller than 0.7, ranking at the bottom of the country. ere is a huge potential for TCDEE improvement in these provinces. Among them, Shandong and Hebei lay in eastern coastal areas. e low TCDEEs of these two provinces are mainly attributable to the redundancy of tourism inputs. Except for Shandong and Hebei, the rest of this group of provinces performed undesirably in the TCDEE. An important reason is the backward transportation and poor infrastructure. In particular, the mean TCDEE of Gansu was merely 0.3029, which is largely associated with the low tourism output.
Overall, different provinces differed significantly in TCDEE. Most eastern provinces had relatively high TCDEEs, while the central and western provinces had relatively low TCDEEs. To promote low-carbon tourism, China must treat central and western parts as the focal points.
Furthermore, China was split into eastern, central, and western parts, according to the geographical location and economic level. e variations of TCDEEs in China and the three parts are displayed in Figure 1.
In the sample period, the TCDEEs in eastern, central, and western parts all changed in the shape of letter In addition, there were clear differences in TCDEE between the three parts. In the sample period, the mean TCDEE of the eastern part was as high as 0.8855, which is   (7), TCDEE impactors were estimated by the Tobit model, using Stata 12.0. Table 4 reports the coefficients of the variables and the significance test results. e estimation coefficient of the tourism level (TL) was positive but failed the significance test. us, the tourism level does not significantly affect TCDEE. e potential reason is as follows: although it brings more economic benefits, a rising tourism level does not truly promote energy conservation and emission reduction in tourism. Currently, the tourism in China is dominated by fashionable tourism.

Tobit Regression Result. Based on formula
is conventional model of tourism consumes too many energies and discharges lots of pollutants, including carbon dioxide.
erefore, TCDEE improvement hinges on the transformation of fashionable tourism to low-carbon tourism.
Industrial structure (IS) has a prominent negative effect on TCDEE on the level of 1%. is result is closely associated with the internal structure of the tertiary industry in China. At present, the tourism-related sectors account for a large proportion in the tertiary industry, namely, transportation, warehousing, and postal industry; wholesale and retail industry; lodging and catering industry. ese sectors consume more energies and emit more carbon dioxide than finance, computer, and software sectors. What is worse, the tertiary industry lacks motivation and vitality, owing to the defects in current institutions and the limited degree-offreedom for market mechanism. All these issues drag down the TCDEE.
As opposed to our expectation, technical progress (TP) clearly suppresses TCDEE, which is probably due to the direction of technical innovation. Acemoglu et al. [33] divided the technical R&D of enterprises into clean technology and polluting technology. If an enterprise firstly develops polluting technology, then technical innovation only adds to pollutant discharge. In reality, most tourism enterprises seek economic benefits rather than ecological benefits in technical R&D. ey prefer polluting technology that takes effect quickly, while creating lots of pollution. Against this backdrop, technical progress can hardly lead to energy conservation and emission reduction unless the government properly guides the R&D direction of enterprises.
Opening-up (OU) has a significant negative correlation with TCDEE. A possible reason is that FDI indeed eases the fund shortage of tourism development. But a lot of funds continue to flow towards the high-pollution tourism sectors. e cash flow pushes up pollutant discharge of tourism and thus lowers TCDEE. e urbanization level (UL) has a significant positive effect on TCDEE at the level of 1%. is means a high urbanization level is conducive to TCDEE improvement. Every 1% of growth in the urbanization level is followed by 2.697% of increase in TCDEE. As mentioned before, highly urbanized areas tend to have complete infrastructure, possess mature techniques of energy-saving and emission reduction, and make efficient use of resources. Compared with rural residents, urban residents are often well educated and fully aware of the importance of environment. ese factors obviously help to promote TCDEE.
Environmental regulation (ES) has an obvious negative impact on TCDEE. is result confirms our hypothesis: if the government investment in environmental pollution control is too high, the resource allocation mechanism of the market would be distorted, which hinders the efficiency improvement. erefore, the government should not solely rely on mandatory instruments of environmental regulation to better control tourism pollutants but adopt other types of environmental regulation instruments: market incentives, public participation, and voluntary actions.

Discussion.
e energy consumption and carbon dioxide emission of tourism grow year by year. As a result, low-carbon economy becomes the optimal choice for Chinese tourism to contribute to global energy conservation and emission reduction [3]. Improving the TCDEE is an important path for tourism to realize sustainable development. Chen et al. [3] claimed that it is reasonable and necessary to decompose the national goal of emission reduction and lowcarbon development to different regions in the light of the spatial imbalance of regional economy and tourism development. Focusing on the TCDEE of each Chinese province,   Note: * , * * , and * * * are the significance levels of 10%, 5%, and 1%, respectively. this paper discovers a significant provincial difference in TCDEE. Some provinces achieved the optimal TCDEE (efficiency of 1), and some provinces failed to realize the ideal TCDEE. e Chinese provinces had marked a difference in the development of low-carbon economy. us, it is necessary to prepare different tourism carbon reduction policies for different provinces. e research results echo with the ideas of Chen et al. [3].
In addition, the distribution of provincial TCDEEs is closely associated with economic growth. In general, the development of tourism economy spurs tourism carbon emission and negatively affects TCDEE. In reality, however, the Chinese provinces pay attention to ecological protection, while promoting tourism economy. is effort drives the low-carbon development of tourism. Taking Heilongjiang as an example, Tang and Huang [34] studied the detachment between carbon dioxide emission and tourism economy growth in 2019 and found that the detachment is basically benign. us, they concluded that energy conservation and emission reduction measures have achieved certain effects on promoting the tourism development in Heilongjiang. In this study, the TCDEE of Heilongjiang averaged 0.9007 in the sample period, which is close to the optimal frontier. erefore, our conclusion that Heilongjiang had a relatively high TCEDD can be strongly supported by Tang and Huang's results [34].

Conclusions
To realize sustainable development of tourism, the key lies in spurring the saving of energy and emission reduction in the industry. Taking the carbon dioxide emissions of tourism as the undesired output, this study constructs an index system for TCDEE, measures the TCDEEs of 30 provinces in China during 2010-2018, using the EBM model, and analyzes the regional differences of TCDEE. Finally, the TCDEE impactors were tested empirically by the Tobit model. e previous studies only tackle the TCEDD on the national scale. In this paper, the tourism carbon emission of each Chinese province is measured in the bottom-up manner and used as an undesired output to build a TCEDD index system for comprehensive evaluation. is study not only reduces the research scale of tourism low-carbon development but also compares the TCDEEs between provinces and identifies the factors influencing the provincial difference. ese innovative efforts help to formulate scientific policies. e main results are as follows: (1) During the sample period, only six provinces achieved DEA-optimal TCDEEs, namely, Beijing, Tianjin, Inner Mongolia, Shanghai, Jiangsu, and Guangdong. e other provinces failed to reach this state. Besides, there were significant provincial differences in TCDEE. Most provinces with high TCDEEs belong to eastern coastal areas, and those provinces with low TCDEEs concentrate in central and western areas. (2) In the sample period, the TCDEEs in eastern, central, and western parts all changed in the shape of letter N, i.e., the TCDEEs first increased, then declined and eventually rebounded. Furthermore, there were clearly phased variations in regional TCDEEs. In addition, the three parts diverged in terms of TCDEE. e eastern part had the highest TCDEE, followed by the central part, while the western part had the lowest TCDEE. In general, the TCDEE in China gradually decreased from the east to west. (3) According to the Tobit estimation results on TCDEE impactors, TCDEE is significantly enhanced by the urbanization level, greatly inhibited by industrial structure, technical progress, opening-up, and environmental regulation, and not significantly affected by the tourism level.
According to the above conclusions, the following policy suggestions were provided on the low-carbon development of tourism: (1) prepare specific policies on tourism carbon emission reduction for each region; (2) change the traditional fashionable tourism and pay efforts to develop green tourism under the low-carbon economy; (3) accelerate the effective integration of tourism and tertiary industry, optimize the internal structure of the tertiary industry, and reduce the energy consumption intensity of transportation, warehousing, and postal industry, as well as accommodation and catering industries; (4) strengthen the guidance on the direction of technological progress and vigorously promote green innovation; (5) further improve the environmental protection threshold of foreign investment and reduce the entry of foreign-funded enterprises with high pollution and high energy consumption; (6) step up the publicity of environmental protection and enhance people's awareness of environmental protection; (7) give full play to the government in environmental governance and adopt the marketoriented method to control the environmental pollution issues of tourism.
ere are still two limitations in this research. Firstly, the tourism carbon emission was estimated. is is a common practice in research when the government has not released the official data of carbon emissions in tourism. However, the estimation may affect the accuracy of the results to a certain extent. Secondly, the sample data were generated between 2010-2018 for the consideration of data availability and comprehensiveness. e short span of the data hinders the long-span analysis of TCDEE and the selection of relevant influencing factors. In the future, the Chinese government is expected to release more data on tourism carbon emission, as it pays more and more attention to carbon neutrality. en, the authors would carry out deeper analysis on the data released by the government. In addition, the authors would expand the channels of data acquisition. Apart from the statistical yearbooks published by the government, the authors would try to expand the time span of samples by querying relevant data on database websites.

Data Availability
e data used to support the findings of this study are available from the corresponding author upon request. 8 Journal of Environmental and Public Health

Conflicts of Interest
e authors declare that they have no conflicts of interest.