The CO₂ Emission Efficiency of China’s Hotel Industry under the Double Carbon Objectives and Homestay Growth

Abstract

Hotels emit large amounts of CO₂ when providing services. Carbon neutrality objectives and the growth of homestays have forced hotel managers to pay more attention to carbon reduction. First, this paper adopted the input–output method to calculate hotel CO₂ emissions. It was found that the amount of CO₂ emissions in the hotel industry decreased from the year of 2016 to 2018, and there are structural differences in the CO₂ emissions of the hotel industry in China. Furthermore, this paper adopted the three-stage SBM-DEA model to evaluate hotel CO₂ emission efficiency in 30 provinces in China. The results show that CO₂ emission efficiency increased significantly when environmental factors were eliminated. A close observation of different regions shows that the eastern region has a higher CO₂ emission efficiency than the western and middle regions. The emergence of homestays has led to increased input slacks in the labor and energy consumption of hotels, which has a negative effect on efficiency indirectly. Finally, implications and recommendations for the hotel industry are proposed.

1. Introduction

China is the world’s largest energy consumer and CO₂ emitter. The CO₂ emissions of China account for approximately 30% of global emissions (Data source: “Statistical Review of World Energy 2020” released by BP (British Petroleum)). Carbon neutrality and peak carbon dioxide emissions are dual strategic objectives (“the double carbon objectives”) for each industry in China. Carbon neutrality refers to enterprises, organizations or individuals offsetting the total amount of greenhouse gas emissions that they directly or indirectly produce within a certain period of time by means of plant afforestation, energy conservation, and emission reduction in order to achieve zero CO₂ emissions. Peak carbon dioxide emissions refer to China’s commitment to stop growing and gradually decrease emissions from their peak before 2030. Previous research on the CO₂ emission of high-energy-consuming industries has been relatively broad, and little research has focused on specific industries. Under the double carbon objectives, more refined research on this issue is necessary. This is one of the research gaps in CO₂ emissions.

As a pillar of the tourism industry, hotels are an essential source of CO₂ emissions, second only to the transportation industry. The CO₂ emissions of the tourism industry have increased rapidly to account for approximately 2.5% of China’s total CO₂ emissions and become a major factor affecting global climate change [1,2]. However, the hotel industry has been ignored in CO₂ emissions studies, and the green reform process of hotels has been dilatory for a long time.

From the industry perspective, hotel operators often fail to consider environmental issues [3]. Even with a green management mindset, hotel operators may easily ignore the CO₂ emissions problems caused by different stages and types of hotel investments. Thus, less effort has been made to perfect environmental plans, and energy consumption reduction has received little interest. From a governance perspective, the main reason for the ineffective regulation of hotel carbon emissions is that most of the policies focus on tourism as a whole. There is a lack of understanding of hotel industries specifically.

In October 2020, the Chinese government stated that “it is necessary to accelerate the development of green and low-carbon and comprehensively improve the efficiency of resource utilization”. At the 75th UN General Assembly, China also promised to realize carbon neutrality by 2060. This plan sets clear requirements for energy conservation and emissions reduction for heavy-emissions industries. An understanding of the CO₂ emissions in different regions and industries can inform the design of more specific and convenient emissions reduction measures and enable the government to create and implement precise policies. These are extrinsic reasons to calculate the concerning CO₂ emissions of the hotel industry.

In addition to these extrinsic reasons, the internal structure of the hotel industry has changed. The rise of the shared accommodation industry (“homestays”) is an intrinsic reason to effect a green reform of traditional hotels. As a shared economy industry, homestays provide an accommodation experience different from that of traditional hotels. Due to the use of idle housing resources, homestays are greener and more environmentally friendly than traditional hotels [4]. This new industry can trigger traditional hotel managers to change their business philosophy and pay more attention to energy conservation and emission reduction. Managers have realized that customers recognize hotels’ efforts to implement environmental protection measures and are more committed to environmentally friendly hotels [5]. The competition with homestays has forced the traditional hotel industry to implement green and low-carbon reforms.

Thus, the objective of this paper includes calculating the CO₂ emission and emission efficiency of the hotel industry, which can help us to accurately understand the current situation of hotel CO₂ emission in order to provide accurate policies to improve the efficiency of hotel CO₂ emissions in different regions.

In many carbon emissions studies, in addition to calculating the absolute value of CO₂ emissions, it is important to evaluate emissions efficiency. Data envelopment analysis (DEA) usually assumes that producing more outputs with fewer input resources is a criterion of efficiency. Air pollutants and hazardous waste have been widely recognized as undesirable outputs of production and social activities. Thus, the development of technologies with fewer undesirable outputs is an important subject of concern in every area of production [6]. In the presence of undesirable outputs, technologies yielding more good (desirable) outputs and fewer bad (undesirable) outputs relative to the amount of input resources are recognized as efficient. In this paper, CO₂ emissions are an undesirable output. Thus, we could evaluate the CO₂ emission efficiency of China’s hotel industry via the DEA approach. This is also a research gap in the DEA approach.

Based on the increased attention to lowering carbon emissions both inside and outside the industry, this paper conducts an in-depth study of the scale and efficiency of carbon emissions of hotels in China via the DEA approach. In the context of global warming, the hotel industry needs to realize the importance of environmental sustainability. From the inside, when hotel staff receive training on green concepts, they can better protect the environment [7]. When hotel managers implement green changes, it has a positive impact on the resilience of their teams [8]. All these show that the green hotel research is very important, our paper for the next step to formulate clear policies and guide the hotels in a green and low-carbon direction is to make contributions.

2. Literature Review

The literature on hotel CO₂ emissions has the following three aspects: calculation of the amount of CO₂ emissions, evaluation of emission efficiency, and identification of influencing factors.

Previous papers calculating the amount of CO₂ emissions have mostly used data related to the tourism industry as a whole. There are the following two calculation paths: top-down and bottom-up. The top-down approach is based on the input–output table [9]. For example, Meng (2017) calculated the direct and indirect CO₂ emissions (Indirect CO₂ emissions include those from the production, manufacturing and sale of tourism physical carriers and those generated by tourism management authorities in the normal operation of the tourism industry) of China’s accommodation and catering industry. The study calculated that there was 25.95 Mt of indirect emissions in the year of 2010, approximately 2.5 times greater than direct CO₂ emissions (generated from the activity processes of the tourism industry). The bottom-up method estimates energy consumption and green productivity [10,11]. Huang [12] estimated that the average annual growth rate of carbon emissions from tourism in the Yangtze River Economic Belt is 14.91%. Wu and Shi [13] found that the energy consumption and CO₂ emissions of China’s accommodation industry in 2008 were 96.80 Mt. In addition, Gössling [14] found obvious differences between energy consumption and CO₂ emissions among different accommodation types. For example, the energy consumption and CO₂ emissions of normal hotels rank first in the accommodation industry, far higher than those of camping grounds, holiday villages, and other subcategories. Given the limited data granularity, few studies specifically focus on the hotel industry or the provincial level.

Among the different methods of measuring emission efficiency, DEA is widely used. The three-stage DEA model proposed by Fried [15] and the SBM (slack-based measure)-undesirable model proposed by Tone et al. [6] have been widely used to measure CO₂ emission efficiency. However, research on CO₂ emission efficiency in the hotel industry remains scarce. Existing studies mainly use the DEA-CCR model proposed by Charnes, Cooper, and Rhodes [16] and the DEA-BCC model proposed by Banker, Charnes, and Cooper [17] to evaluate the “operating efficiency” of hotels. Other studies that use DEA to calculate CO₂ emissions consider the tourism or service industry as a whole. For example, Zha [18] found that the low-carbon development efficiency of tourism in China is highest in the middle region, followed by the eastern region and then the western region. Wang et al. [19] found that the carbon emission efficiency of the service industry in China is highest in the eastern region, followed by the middle region and then the western region. Regarding dynamic efficiency calculations, recent research has used the Malmquist–Luenberger index to calculate the evolution of tourism total factor carbon productivity [20].

Examining the environmental factors that affect hotel CO₂ emissions, Moutinho et al. [21] and Robaina-Alves et al. [22] studied tourism satellite accounts in Portugal by using decomposition analysis. Both studies found that the tourism activity effect, energy over fixed capital, and capital over labor productivity all have significant impacts on CO₂ emissions from the accommodation industry. This indicates that excessive energy consumption or excessive capital investment leads to increased CO₂ emissions. Other factors that affect CO₂ emissions include tourism consumption [23], number of tourist receptions (Wang et al., 2017), and per capita GDP [24].

In terms of current tourism management and low-carbon environmental protection studies, there is a well-established discussion of the CO₂ emissions amount and emission efficiency, but its application in the hotel industry is limited. Therefore, this manuscript wishes to expand the scope of the application of the SBM-DEA model and provide a more accurate understanding of the carbon emission efficiency of the hotel industry.

This paper’s structure is as follows. In Section 3, we use the input–output method to calculate the energy input and CO₂ emissions of the hotel industry’s regional heterogeneity in China. In Section 4, the three-stage DEA model and SBM-undesirable model are used to measure CO₂ emission efficiency, and the environmental slack effect is evaluated. Section 5 discusses the empirical conclusions. Finally, policy recommendations are provided to inform emission reduction policies as well as the expansion of tourism and even the service industry.

3. Calculation of Hotel CO₂ Emissions

In this section, we focus on the energy inputs of the hotel industry based on the 2015/2017 Chinese input–output table; then, the CO₂ emission amount of the hotel industry in China from 2016 to 2018 is calculated.

3.1. Data Source

The China Carbon Emission Accounts and Databases (CEADs) provided the sample data, which cover 30 provinces, municipalities, and autonomous regions (excluding the Tibet Autonomous Region, Hong Kong, Macao, and Taiwan (The input–output table provided by CEADs does not include data for Tibet, Hong Kong, Macau and Taiwan. When we searched a wider range of data, we found that the data in these areas have different statistical calibers. Therefore, in order to make the experiment scientific, data from different statistical sources cannot be placed in the same model.) of China. In the industrial classification for national economic activities of China, the corresponding sector of the hotel industry is the accommodation industry, but in the input–output table, the data of the accommodation industry and catering industry cannot be separated; therefore, this paper measures the hotel industry by the accommodation and catering industry. We chose the input–output data for the accommodation and catering industries to reflect the energy consumption of the hotel industry.

There are 42 economically productive sectors listed in the 2015/2017 Chinese input-output table. The energy products used in the hotel industry include products from the following four production sectors: (1) mining and washing of coal, (2) extraction of petroleum and natural gas, (3) processing of petroleum, coking, and (4) the production and distribution of gas. For energy choices, we refer to the relevant literature and combine it with energy data from China’s provincial energy inventory. Ten types of energy consumption products are selected, namely, raw coal, other coal washings, briquettes, coke, gasoline, kerosene, diesel oil, fuel oil, liquefied petroleum gas, and natural gas. Based on the industrial classification for national economic activities (GB/T4754-2017), the production sector corresponding to raw coal, other coal washings, and briquettes is coal mining and dressing; the sector corresponding to coke, gasoline, kerosene, diesel oil, and fuel oil is petroleum processing and coking; the sector corresponding to liquefied petroleum gas is gas production and supply; and that corresponding to natural gas is petroleum and natural gas extraction [25].

3.2. Calculation Process

We calculated the CO₂ emissions of the hotel industry by using the methods recommended by Liu and Hu [26] and Shan [27] on fossil fuel combustion and cement production. According to physical characteristics, the direct coefficient of carbon (c) conversion to CO₂ in fossil fuel combustion is 3.67 (According to the chemical reaction equation of carbon combustion C + O₂ = CO₂; the relative atomic mass of C is 12, of O is 16, and of CO₂ is 44; thus, the conversion of C to CO₂ is 44/12, which is approximately equal to 3.67), which can be used in Formula (1), which is as follows:

$$C{E}_{ij}=3.67\times A{D}_{ij}\times NC{V}_i\times C{C}_i\times O_i$$

(1)

where CE_ij represents the CO₂ emissions of the ith energy consumption product in the jth province (i.e., Beijing, Hubei, etc.); AD_ij represents the consumption of the ith energy consumption product in the jth province; NCV_i represents the net calorific value of the ith energy consumption product (i.e., the calorific value per unit of fossil fuel combustion); CC_i represents the carbon content of the ith energy consumption product; and O_i is the oxidation efficiency, which indicates the oxidation ratio of the ith energy consumption product as it burns.

Table 1. Fossil fuel parameter values.

Energy Consuming Products	NCVi 1	CCi 2	O
Raw Coal	0.21	26.32	74%
Other Washed Coal	0.15	26.32	74%
Briquettes	0.18	26.32	74%
Coke	0.28	31.38	89%
Gasoline	0.44	18.90	96%
Kerosene	0.44	19.60	96%
Diesel Oil	0.43	20.20	96%
Fuel Oil	0.43	21.10	96%
Liquefied petroleum gases	0.51	17.20	97%
Natural Gas	3.89	15.32	98%

¹ Data sources: China Carbon Emission Databases and Accounts consolidation. ¹ The unit of natural gas is PJ/10⁸ m³, and the other fossil fuels are measured in PJ/10⁴ t. ² The unit is tC/TJ.

shows the parameter values of the 10 energy-consuming products [28,29].

The total CO₂ emissions of the hotel industry in each region are calculated by summing the CO₂ emissions of each energy consumption product, as shown in Formula (2):

$$TC{E}_j={\displaystyle \sum }_{i=1}^{q}C{E}_{ij},q=1,2,\dots ,10$$

(2)

The results are shown in

Table 2. CO₂ emissions of China’s hotel industry in 2016–2018.

Eastern	CO2			Western	CO2			Middle	CO2
Province	2016	2017	2018	Province	2016	2017	2018	Province	2016	2017	2018
Beijing	1.465	1.349	1.011	Inner Mongolia	9.180	5.002	5.249	Shanxi	2.231	2.324	2.267
Tianjin	1.978	1.007	0.960	Guangxi	0.932	1.028	0.860	Jilin	2.804	2.237	0.340
Hebei	2.854	4.073	3.011	Chongqing	1.271	1.112	1.131	Heilongjiang	15.178	14.736	9.641
Liaoning	1.812	1.448	1.251	Sichuan	3.565	3.662	3.003	Anhui	1.857	0.881	0.535
Shanghai	0.643	0.629	0.594	Guizhou	9.297	9.660	7.887	Jiangxi	0.774	0.909	0.985
Jiangsu	0.363	0.143	0.145	Yunnan	2.100	2.472	2.514	Henan	0.194	1.233	2.131
Zhejiang	2.443	2.527	2.134	Shaanxi	2.181	2.109	1.852	Hubei	3.549	5.752	6.010
Fujian	0.115	0.476	0.487	Gansu	0.347	0.526	0.577	Hunan	7.643	8.032	8.685
Shandong	7.279	3.950	3.856	Qinghai	0.341	0.697	0.752	--	--	--	--
Guangdong	0.996	1.988	2.029	Ningxia	0.167	0.166	0.177	--	--	--	--
Hainan	0.194	0.225	0.143	Xinjiang	2.449	1.269	1.254	--	--	--	--
Mean	1.831	1.619	1.420	--	2.894	2.518	2.296	--	4.279	4.513	3.824
Total	20.142	17.812	15.622	--	31.830	27.703	25.256	--	34.229	36.105	30.594

Note: The unit of CO₂ emissions is 10⁶ tons (Mt).

.

In general, from the years 2016 to 2018, the CO₂ emissions of the hotel industry decreased year on year, at 86.202, 81.621, and 71.472 MT, respectively. The total CO₂ emissions in the years 2017 and 2018 were approximately 4.6 and 14.8 MT, respectively, less than those in the year 2016 indicating a trend of energy conservation and emissions reduction in the hotel industry.

From the perspective of regional heterogeneity, 60% of provinces have reduced their CO₂ emissions. Ranking the CO₂ emissions of the hotel industry in the 30 regions from high to low, the top five regions are Heilongjiang, Guizhou, Hunan, Hubei (Hubei ranked seventh in the year 2016, while Shandong ranked fifth), and Inner Mongolia. Eighteen provinces saw a decrease in CO₂ emissions over those three years, and the top five regions in terms of decreasing emissions were Heilongjiang, Inner Mongolia, Shandong, Jilin, and Guizhou. Although Heilongjiang is one of the regions with the highest carbon emissions, a carbon emission reduction was also evident in its hotel industry in the year 2018. In addition, the CO₂ emissions of the hotel industry increased in 12 provinces, including three in the east (Hebei, Fujian, and Guangdong), five in the middle (Shanxi, Jiangxi, Henan, Hubei, and Hunan), and four in the west (Yunnan, Gansu, Qinghai, and Ningxia). The top five regions in terms of increasing emissions are Hubei, Henan, Hunan, Guangdong, and Yunnan (Figure 1).

In this research, the 30 provinces of the research object are divided into three regions—the east, the middle, and the west. The reasons are as follows: (1) there are differences in the economic development level and the macro environment among the east region, the middle region and the west region of China. The highlight of the three-stage DEA model we adopted is that the environmental factors are considered in the second stage; therefore, the discussion of different regions is necessary. (2) The previous literature has analyzed a large number of regional differences in carbon emission efficiency, but mostly for the service industry and tourism industry. This classification covers too many sub-industries, which is not conducive to putting forward accurate carbon emission reduction targets. Note that the regions rank as follows in terms of hotel CO₂ emissions: middle region > western region > eastern region. This finding indicates that hotels in the eastern region contributed the least to overall carbon emissions. Combined with the level of economic development in the eastern region, hotels in the eastern region use carbon more efficiently. A more adequate assessment will be given later. In a previous study [17], the ranking of service industry CO₂ emissions was eastern region > western region > middle region, which proves that there are structural differences in carbon emissions within the service industry in different regions. This shows that it is necessary to study the hotel industry by region. We will discuss this point in the last chapter.

4. Evaluation of CO₂ Efficiency Using a Three-Stage SBM-DEA Model

According to the CO₂ emissions calculated above, this section uses a three-stage SBM-DEA model to measure the CO₂ emission efficiency of the Chinese hotel industry. The specific steps are as follows.

In the first stage, the input variables, output variables, and environmental variables were selected. The SBM-undesirable model was used to calculate the CO₂ emission efficiency. At this stage, the slack of the input variable in each province could be calculated simultaneously.

In the second stage, stochastic frontier analysis (SFA) was used to separate the effects of management inefficiency, environmental factors, and statistical noise on input slacks. At this stage, adjusted input values could be recalculated, while environmental impacts are excluded.

In the third stage, the adjusted input variable values from the second stage were used to replace the original variable values in the first stage, and the pure CO₂ emission efficiency was recalculated.

4.1. Variable Selection

Using Zhang et al.’s [30] quantitative research on carbon emissions and the efficiency of the construction industry for reference, we took unexpected CO₂ emissions as an undesirable output variable to help effectively calculate and evaluate CO₂ emission efficiency. According to the characteristics of the hotel industry and previous research [19], the original value of fixed assets was taken as the capital variable of the hotel industry, the number of employees at the end of the year as the labor variable, and total energy consumption as the energy consumption variable. The prime operating revenue of the hotel industry was selected as the desirable output index [31].

To determine the influence on the slack of input variables, the industrial structure, residents’ income level, urbanization process, and number of homestays were selected as environmental variables [32]. Variable definitions and data sources are shown in

Table 3. Variable definitions.

Variables	Symbol	Definitions	Data Sources
Input variables
Capital variable	K	Original value of fixed assets (108 yuan)	National Bureau of Statistics of China
Labor variable	L	Number of employees at the end of the year
Energy variable	E	Total energy consumption (104 yuan)	MRIO
Undesirable Output variables
CO2	TCE	CO2 emissions (104 tons)	In this paper 1.2
Desirable Output variables
Prime operating revenue	POR	prime operating revenue (108 yuan)	National Bureau of Statistics of China
Environmental variables
Industrial structure	Indstr.	Proportion of accommodation and catering industry in GDP	National Bureau of Statistics of China
Residents’ income level	PeGDP	GDP per capita (yuan)
Urbanization process	Urban.	Proportion of urban population in the total population
Homestays	Hstay.	Number of homestays	Website Statistics 1

¹ Data source website: https://www.tianyancha.com/ (accessed on 12 May 2021) the accumulated number of homestay related enterprises in each province from the year of 2016 to 2018 was sorted out.

.

4.2. Three-Stage SBM-DEA Model

4.2.1. Stage 1: CO₂ Emission Efficiency within Environmental Factors

In the first stage, the CO₂ emission efficiency within environmental factors of hotels in different regions from the year of 2016 to 2018 was calculated with the SBM-undesirable model.

Referencing Tone (2001, 2004), the SBM-undesirable model was constructed in the first stage. In this paper, there were n decision-making units, “DMUs”, n = 30 (30 provinces), each with the following three types of variables: input variables $x$, desirable outputs $y^g,$ and undesirable outputs $y^b$, represented by three vectors, $x{R}^m$, $y^g{R}^{s1}$ and $y^b{R}^{s2}$, where R means the matrix; $m$ means the number of input variables; s1 means the number of undesirable outputs; and s2 means the number of desirable outputs. In this paper, $m=3,s1=1,s2=1.$

We assumed that $X>0$, $Y^g>0$, $Y^b>0$, when $X=\left[x_1,\dots ,x_n\right]R^{m\times n},Y^g=\left[y^{g1},\dots ,y^{gn}\right]R^{s1\times n}$,$Y^b=\left[y^{b1},\dots ,y^{bn}\right]R^{s2\times n}$. The production possibility set (P) is shown in Formula (3), where $\lambda R^n$ is the intensity vector.

$$P=\left\{\left(x,y^g,y^b\right)|x\ge X\lambda ,y^g\le Y^g\lambda ,y^b\ge Y^b\lambda ,\lambda \ge 0\right\}$$

(3)

The definition of DMU₀ ($x_0,y_0^g,y_0^b$) is efficient in the presence of undesirable outputs if there is no vector ($x,y^g,y^b)\in P$ such that $x_0>x,y_0^g\le y^g$ and $y_0^b\ge y^b$ with at least one strict inequality.

The efficiency of undesirable output ${\rho }^{*}$ can be calculated as follows (the vector meaning in Appendix A,

Table A1. The vector notations used in Section 4.

Vectors	Meaning
x, X	Input vector of 30 provinces
$y^g, Y^g$	Desirable output vector of 30 provinces
$y^b,Y^b$	Undesirable output vector of 30 provinces
$\lambda$	Intensity vector
$s^{-}$	Excesses vector in inputs
$s^g$	Excesses vector in desirable outputs
$s^b$	Excesses vector in undesirable outputs

):

$${\rho }^{*}=min\frac{1-\frac{1}{m}{{\displaystyle \sum }}_{i=1}^m\frac{s_i^{-}}{x_{i0}}}{1+\frac{1}{s_1+s_2}\left({{\displaystyle \sum }}_{r=1}^{s_1}\frac{s_r^g}{y_{r0}^g}+{{\displaystyle \sum }}_{r=1}^{s_2}\frac{s_r^b}{y_{r0}^b}\right)}$$

(4)

$$\mathrm{St}:x_0=X\lambda +s^{-}$$

(5)

$$y_0^g=Y^g\lambda -s^g$$

(6)

$$y_0^b=Y^b\lambda +s^b$$

(7)

$$s^{-}\ge 0,s^g\ge 0,s^b\ge 0,\lambda \ge 0$$

(8)

The variable $s^{-},s^g,s^b$ is the difference between the actual input–output value and the target input–output value, $0\le {\rho }^{*}\le 1$. Subscript $i$ means the $i$th input variable, and subscript $r$ means the $r$th output variable. If ${\rho }^{*}=1$, that means the province’s carbon efficiency is efficient. If ${\rho }^{*}<1$, this means that the province’s carbon emission efficiency is inefficient, and the actual input value is greater than the target input value.

The results by region are shown in

Table 4. CO₂ emission efficiency of the hotel industry in 2016–2018 (stage 1).

Regions	CO2 Emission Efficiency			2016 to 2018 Efficiency Changes (%)
	2016	2017	2018	△2016–2017	△2017–2018	△2016–2018
Eastern	0.734	0.787	0.789	7.28%	0.27%	7.57%
Middle	0.439	0.421	0.512	−3.95%	21.61%	16.81%
Western	0.656	0.675	0.659	2.98%	−2.35%	0.56%
Mean	0.626	0.649	0.668	3.53%	3.07%	6.60%

.

The average efficiency of hotel industry CO₂ emissions from the year 2016 to 2018 was 0.626, 0.649 and 0.668, respectively, and the mean values rose steadily each year, with growth greater than 3%. However, the CO₂ emission efficiency of all three regions did not exceed 0.8, which indicates that the CO₂ emission efficiency of the hotel industry in China can be improved.

From the perspective of regional heterogeneity, the average emission efficiency increased in these three years, while the efficiency of the western regions increased the least. The eastern region had a significantly higher efficiency than the middle and western regions. In previous studies [19], the regions ranked as follows in terms of the CO₂ emission efficiency of the service industry: east > middle > west. Our results indicate that the hotel industry in the middle region has the lowest carbon emission efficiency. From the existing literature [18], the regions rank as follows on their CO₂ emission efficiency in tourism: middle > eastern > western. Combining with our results, the hotel industry in middle China performs worst overall in the tourism industry. Therefore, hotels in the middle region should pay more attention to energy conservation reform and emissions reduction under the double carbon objectives.

In a comparison of provinces and cities, hotels in Beijing, Shanghai, Guangdong, Chongqing, Qinghai, and Ningxia had the highest CO₂ emission efficiency. From the year 2016 to 2018, Hunan, Liaoning, and Anhui were the top three provinces in terms of increasing efficiency, especially Hunan, which saw an improvement of close to 200%. The efficiency values of Henan, Shandong, and Tianjin decreased the most, at over 30%.

4.2.2. Stage 2: Stochastic Frontier Analysis (SFA) of Environmental Factors

SFA can be used to clarify the influence of environmental variables on the slack values of the input variables. The slack values indicate the difference between the original input and the target input. A negative relationship between the environmental variable and the slack value indicates that the environmental factors can narrow the gap between the original input value and the actual value of the input variable, which is conducive to improving carbon emission efficiency. In the opposite case, environmental factors inhibit the improvement of carbon emission efficiency. According to Fried (2002), environmental factors, management inefficiency, and statistical noise all influence the efficiency value. In this paper, we assumed that the slack value of input variables was affected by four environmental variables, namely, the industrial structure, residents’ income level, the urbanization process, and homestays.

We gradually separated the effects of management inefficiency, statistical noise, and environmental factors and obtained the adjusted values of the input variables using the following formula:

$$X_{ij}^A=X_{ij}+\left[max\left(f\left(Z_j;{\widehat{\beta }}_i\right)\right)-f\left(f\left(Z_j;{\widehat{\beta }}_i\right)\right)\right]+\left[max\left(v_{ij}\right)-v_{ij}\right]$$

(9)

where $X_{ij}^A$ is an adjusted input variable, $X_{ij}$ is the ith original input variable in the jth province, $max\left(f\left(Z_j;{\widehat{\beta }}_i\right)\right)-f\left(f\left(Z_j;{\widehat{\beta }}_i\right)\right)$ is an adjustment to external environmental variables, and $max\left(v_{ij}\right)-v_{ij}$ is statistical noise. $Z_j$ means the independent external environmental variables, and ${\widehat{\beta }}_i$ means is a vector of the environmental factor parameters to be estimated.

With the slack values calculated for each input variable in stage one, Frontier 4.1 software was used to carry out the SFA by separating the influence of management inefficiency and statistical noise. The results are shown in

Table 5. SFA analysis results of CO₂ emission efficiency in China’s hotel industry.

Dependent Variable (Slack Variable)	Original Value of Fixed Assets		Number of Employees (Labor Force)		Total Energy Consumption
Independent variable	Coefficient value	T value	Coefficient value	T value	Coefficient value	T value
Industrial structure	5.463	1.095	−738.944 ***	−87.923	36.866	0.595
Residents’ income level	0.000	−0.92	0.071	0.581	0.156	0.279
Urbanization process	0.262	0.251	−794.666 ***	−6.519	−1943.184 ***	−3.265
Number of homestays	0.003	0.760	4.988 ***	3.363	17.845 **	2.505
Constant terms	−1.677	−0.037	40,281.975 ***	9249.882	81,320.747 ***	5248.644
σ2	6894.637 ***	3.125	438,581,500.000 ***	438,581,500.000	14,435,703,000.000 ***	14,435,703,000.000
γ	0.870 ***	16.825	0.692 ***	12.651	0.805 ***	23.347
log likelihood function	−456.564		−987.906		−1131.135
LR	21.057 ***		18.445 ***		34.434 ***

Note: *, ** and *** denote significance at the levels of 10%, 5% and 1%, respectively.

. Likelihood ratio (LR) tests of the one-sided error are 21.057, 18.445, and 34.434, indicating that the SFA model is reasonable. The parameters ${\sigma }^2$ and $\gamma$ are significant, indicating that management inefficiency is the main factor for the change in input-slack variables, and therefore, the use of three-stage DEA is necessary.

The analysis of the impact of environmental factors on the slack value of input variables is as follows.

(1)

Slack value of hotel industry fixed assets.

The results show that the slack value of fixed assets is not affected by environmental factors. This means that the hotel industry asset allocation is relatively fixed and does not change randomly with a change in the industrial structure, residents’ income, or urbanization in the short term. In addition, the large-scale spread of homestays has not yet affected the fixed assets of traditional hotels.

(2)

Slack value of hotel industry employees.

The industrial structure and urbanization level have a significant negative effect on the slack values of the labor force. That is, when the proportion of the hotel industry in regional GDP is higher, the allocation of labor force resources is more effective. The higher the level of urbanization is, the less labor slack the hotel has, which is conducive to improving the carbon emission efficiency of the hotel. In contrast, the number of homestays increases the employee slack of hotels. This indicates that increases in homestays have increased hotels’ labor slacks, thereby indirectly reducing hotels’ carbon emission efficiency.

(3)

Slack value of hotel industry energy consumption.

The urbanization level negatively affects the energy consumption slack of hotels. This means that the higher the proportion of urban residents is, the more significantly hotel energy consumption could be reduced and the more the CO₂ emission efficiency of the hotel industry could be improved. However, the number of homestays has significantly increased the energy consumption slack of hotels, indirectly reducing hotel carbon emission efficiency.

To sum up, environmental factors have different degrees of impact on the hotel’s capital, labor, energy input. The increase in the industrial structure and urbanization process will indirectly increase the true level of the carbon emission efficiency of hotels; on the contrary, the outbreak of homestays will decrease the true level of the carbon emission efficiency of hotels. In summary, a visualization of the SFA analysis results is shown in Figure 2.

4.2.3. Stage 3: Exclusion of Environmental Factors

In the third stage, the original variable value in the first stage was replaced by the adjusted input variable value in the second stage, and the CO₂ emission efficiency was recalculated. The results at this stage reflected the pure CO₂ emission efficiency value after the elimination of management inefficiency, statistical noise, and environmental factors.

As

Table 6. CO₂ emission efficiency of China’s hotel industry in 2016–2018 (stages 1 and 3).

Province	2016		2017		2018
	Stage 1	Stage 3	Stage 1	Stage 3	Stage 1	Stage 3
Beijing	1.000	1.000	1.000	1.000	1.000	1.000
Tianjin	1.000	0.661	0.835	0.697	0.644	0.675
Hebei	0.275	0.546	0.270	0.517	0.276	0.528
Liaoning	0.510	0.630	0.682	0.668	1.000	1.000
Shanghai	1.000	1.000	1.000	1.000	1.000	1.000
Jiangsu	0.663	0.841	1.000	1.000	1.000	1.000
Zhejiang	0.458	0.596	0.440	0.591	0.429	0.611
Fujian	1.000	1.000	1.000	1.000	1.000	1.000
Shandong	0.542	0.648	0.431	0.587	0.332	0.511
Guangdong	1.000	1.000	1.000	1.000	1.000	1.000
Hainan	0.623	0.813	1.000	1.000	1.000	1.000
Shanxi	0.261	0.495	0.271	0.497	0.292	0.521
Jilin	0.388	0.569	0.345	0.562	0.393	0.678
Heilongjiang	0.385	0.561	0.378	0.570	0.363	0.576
Anhui	0.394	0.618	0.507	0.765	0.705	1.000
Jiangxi	0.500	1.000	0.515	0.702	0.550	0.713
Henan	0.681	0.868	0.414	0.592	0.301	0.495
Hubei	0.556	0.668	0.510	0.661	0.495	1.000
Hunan	0.343	0.521	0.431	0.600	1.000	1.000
Inner Mongolia	0.348	0.561	0.376	0.576	0.451	0.605
Guangxi	0.373	0.625	0.383	0.625	0.377	0.640
Chongqing	1.000	1.000	1.000	1.000	1.000	0.753
Sichuan	0.498	0.649	0.426	0.621	0.415	0.613
Guizhou	0.437	0.644	0.443	0.655	0.427	0.657
Yunnan	0.304	0.549	0.342	0.568	0.374	0.618
Shaanxi	0.420	0.597	0.458	0.649	0.457	0.681
Gansu	0.832	1.000	1.000	1.000	1.000	1.000
Qinghai	1.000	1.000	1.000	1.000	1.000	1.000
Ningxia	1.000	1.000	1.000	1.000	1.000	1.000
Xinjiang	1.000	1.000	1.000	1.000	0.752	1.000
Mean of eastern	0.734	0.794	0.787	0.824	0.789	0.848
Mean of middle	0.439	0.663	0.421	0.619	0.512	0.748
Mean of western	0.656	0.784	0.675	0.790	0.659	0.779
Mean of total	0.626	0.755	0.649	0.757	0.668	0.796

shows, the overall mean value of CO₂ emission efficiency from the year 2016 to 2018 in the hotel industry (0.755, 0.757, 0.796) was higher than the overall mean value in the first stage (0.626, 0.649, 0.668) after adjustment in stage two. The pure CO₂ emission efficiency of hotels in China increased by approximately 20%. This indicates that external environmental factors have a significant negative impact on the CO₂ emission efficiency of the hotel industry.

In terms of regional differences, the ranking of CO₂ emission efficiency of the hotel industry is as follows (as in stage one): eastern region > western region > middle region. Specifically, the efficiency of the middle region increased by nearly 50%, that of the western region by nearly 20%, and that of the eastern region by approximately 8%. This shows that the hotel industry in the middle region is in the most unfavorable external environment for green reform, while the hotel industry in the eastern region is less affected by the external environment.

From the perspective of provinces and cities, the pure CO₂ emission efficiency of most provinces and cities increased significantly. The details are as follows: ① the CO₂ emission efficiency in Beijing, Shanghai, Guangdong, Qinghai and Ningxia did not change either before or after the adjustment; the efficiency value remained at one, indicating that the CO₂ emission efficiency in these provinces was less affected by the environment and had reached the optimal frontier. ② The CO₂ emission efficiency in Tianjin, Liaoning, and Chongqing decreased the most from the year 2016 to 2018. This finding reflects that the external environment of these provinces is conducive to hotel carbon reduction. ③ The rising CO₂ emission efficiency in other regions indicates that external environmental factors limit the improvement of carbon emission efficiency. For example, the CO₂ emission efficiency in Hubei had the highest increase and reached a value of one in the year 2018, indicating that the external environment in Hubei significantly affected hotels’ energy conservation and emissions reduction. In addition, the CO₂ emission efficiency in some regions, such as Hebei, rose sharply for two consecutive years; however, the efficiency value was less than 0.55 after adjustment, significantly less than that of other provinces and cities in the eastern region. This indicates that there is still much room for improvement in these regions.

In summary, the CO₂ emission efficiency of the hotel industry is significantly negatively affected by external environmental factors and has obvious regional heterogeneity. Therefore, the intensity of environmental improvement should be different for each region. The middle region in particular needs to create a better environment to improve hotels’ CO₂ emission efficiency.

To sum up, we used readability graphs (Figure 3) to reflect the temporal and spatial variation in CO₂ emission efficiency from the year 2016 to 2018 between stage one and stage three.

5. Discussion, Conclusions, and Recommendations

5.1. Discussion

Compared with hotels, homestays provide different accommodation experiences to tourists. Homestays have attracted tourists from hotels to some extent, resulting in slack in the labor and energy consumption of hotels, but have not had a significant direct impact on capital slack. While industrial structure optimization and urbanization rationalize the allocation of the labor force and energy consumption in the hotel industry, homestays may cannibalize this allocation.

Although previous studies have studied the competitive relationship between hotels and homestays from the perspective of business management [32], it is not clear whether the relationship between the hotel industry and homestays is complementary or alternative from the perspective of sustainable development. At the same time, homestays increase the labor and energy consumption slacks of hotels in China and has a double negative impact on hotel management and carbon emission efficiency.

In addition, an increase in the proportion of the hotel industry in regional GDP improves CO₂ emission efficiency, indicating that hotel carbon emissions are sensitive to agglomeration effects. In previous studies [19,26], the improvement of the industrial structure will also increase CO₂ emission efficiency, whether in China as a whole or in the service industry of China. Moreover, with a higher proportion of urban residents, a hotel can reduce its labor slack. Thus, hotel managers should improve their staff and service quality. This can achieve an optimal outcome in terms of economic and green benefits.

5.2. Conclusions

Based on the input–output method and a three-stage SBM-DEA model, this paper calculated the CO₂ emission amount and efficiency of the hotel industry in 30 regions of China. The main conclusions are as follows:

(1)

From the year 2016 to 2018, the total amount of CO₂ emissions of the hotel industry continued to decrease, with a decrease of 5.3% in 2017 and 12.4% in 2018, and the average carbon emission efficiency increased year on year but did not reach 0.8. From the perspective of regional heterogeneity, hotels in the eastern region have lower carbon emissions and are more efficient, followed by the western region and the middle region. This ranking is different from the regional ranking of the emission efficiency of tourism and service industries, which shows that there is obvious heterogeneity in carbon emissions within the industry.

(2)

CO₂ emission efficiency was significantly improved after excluding the influence of external environmental factors, especially in the middle and western regions. It shows that the external environment in the eastern region is more conducive to the implementation of carbon emission reduction measures by hotels. The external environment in the middle region is the most unfavorable for hotels to improve carbon emission reduction efficiency.

(3)

The large-scale proliferation of homestay facilities has a significant positive impact on slacks of labor force and energy consumption, negatively affecting hotels’ CO₂ emission efficiency indirectly. Increases in the proportion of the hotel industry in GDP and the urbanization ratio are conducive to reductions in the hotel labor force and energy consumption slacks and improvements to the carbon emission efficiency of the hotel industry.

5.3. Recommendations

Based on the double carbon objectives, the environmental efficiency of hotels is as important as their operational efficiency. Hotels must focus on improving their dual economic and environmental efficiencies. According to the above conclusions, the recommendations are as follows.

From the perspective of hotel industry’s development:

(1)

Continue to promote the green development of the hotel. In recent years, the hotel’s carbon emission reduction has achieved initial results. In the context of energy conservation and emission reduction, awareness of green environmental protection should be enhanced, and the CO₂ emissions of hotels should be reduced. This reform requires more hotels to participate, especially hotels in the middle and west regions.

(2)

Optimize the hotel’s energy input structure. The labor and energy consumption inputs of the hotel industry should be rationally planned and not blindly expanded. It is possible to appropriately decrease the labor input and direct the energy input of hotels in areas with relatively lower levels of urbanization.

(3)

Adopt a differentiated competitive strategy with homestays. In order to reduce the input slacks caused by the increase in homestays, a hotel could provide diversified services to attract customers, such as more detailed room service, special catering services.

From the perspective of government management:

(1)

Pay more attention to the environmental protection conditions of hotels. Simultaneously, hotel facilities should be checked regularly, and energy-saving facilities should be replaced in a timely manner.

(2)

Measures that aim to reduce carbon emissions in different regions and different industries need to be more targeted. For example, environmental protection departments in the middle regions should introduce more policies to encourage hotels to implement carbon emission reductions.

(3)

Encourage the development of diversification, specialization, and differentiation of homestays and hotels, and guide homestays and hotels to attract customer groups with different goals. Departments of hotel management should try to avoid disgusting competition between homestays and hotels.

This paper discussed the amount of CO₂ emissions and efficiency of China’s hotel industry from 2016 to 2018 in the context of homestay growth and with consideration of other environmental factors. To achieve the double carbon objectives, a more accurate understanding of the carbon emissions of each subindustry is necessary. This article showed the entire calculation process of carbon emissions and emission efficiency in a subindustry, including the decomposition of energy input, the selection of environmental factors, and the application of model methods. This approach can provide a meaningful reference for future research in other tourism subindustries both in China and in other countries. At the same time, the results of this article provide comparable basic data for other studies.

The limitations in the paper include two main aspects. First, the data have not been obtained in a long time series to study the changing trend of CO₂ emissions and efficiency in the hotel industry. Future research will use a dynamic DEA model to evaluate carbon emission efficiency with long-term series. Second, the relationship between hotels and homestays has not been clarified. This issue is expected to be addressed with a more complete econometric model in future research. In the future, it is suggested that factors affecting the green efficiency of hotels can be studied from the aspect of the green management of hotels.