Measuring industrial green total factor productivity in the context of the “two‐carbon” target: A case study of energy revolution reform pilot area, China

Since the “double carbon” target was proposed in 2020, emissions reduction has become the focus of the Chinese government's work. Industrial development is a crucial factor contributing to the increase in carbon emissions. Therefore, implementing carbon reduction in the industrial sector is an important part of China's goal of “double carbon.” In this context, this paper uses a three‐stage data envelopment analysis model to calculate the green total factor productivity (GTFP) of 18 industry sectors in Shanxi Province. The results show that: (1) the external environment has influence on the GTFP of energy revolution reform pilot area industrial sectors, and the efficiency value of two industries is overestimated before excluding environmental factors; (2) the industry difference of GTFP is significant, among which pharmaceutical manufacturing industry has the greatest promoting effect on GTFP, and ferrous metal mining and dressing industry has the greatest inhibiting effect on GTFP; (3) there is a serious redundancy in R&D investment of energy revolution reform pilot area in China, and only six industries belong to zero redundancy. (4) To reflect the GTFP of various industrial sectors more clearly, this paper divides four industries: “Double‐high” industry, “High‐low” industry, “Low‐high” industry, and “Double‐low” industry.

which is the highest growth rate in the past 7 years. China's carbon dioxide emissions from energy use increased by 2.2%, showing a significant rebound trend compared with the average growth rate of 0.5% in the past 5 years. Energy has transcended national boundaries and become a global issue of concern to all countries in the world, to speed up the establishment of a clean, lowcarbon, safe, and efficient modern energy system. To combat global climate change, Xi Jinping proposed the "double carbon" target at the 2020 United Nations General Assembly, that is, China's carbon dioxide emissions should peak by 2030 and achieve carbon neutrality by 2060.
Because Shanxi is a coal-rich province in China, the energy consumption will continue to face tremendous pressure to control carbon emissions. Shanxi's gross domestic product (GDP), energy consumption, and coal consumption all showed an overall upward trend in 1985-2014. In 2014, the energy consumption was 4.8 times that of 1985. From the perspective of energy production structure, Shanxi Province is a "high carbon" energy structure province dominated by coal. Among them, the production capacity and output of raw coal, coke, and electricity are among the highest in China. From the perspective of energy consumption structure, Shanxi's coal production accounts for a quarter of the country's total. Of the total energy consumption of the province, 94.7% comes from coal, while wind, hydropower, and coal-bed methane resources account for only 2.17%. From the perspective of energy industry structure, the added value of the first, second, and third industries in Shanxi Province accounted for the provincial GDP proportion of 6.2%, 56.8%, and 37.0%, respectively, the proportion of secondary industry more than the national average of 10%. The contribution of coal, coke, metallurgy, and electric power to economic growth accounted for nearly 90%. Therefore, the State Council recognized Shanxi Province as an energy revolution reform pilot area in May 2019 due to its coal consumption industry characteristics. Shanxi Province, as "the vanguard of the national energy revolution," will assume important responsibilities in improving the quality and efficiency of the energy supply system, building a clean and lowcarbon energy consumption model, promoting energy technology innovation, deepening energy system reform, and expanding energy external cooperation. The key to achieving sustainable development in the energy revolution reform pilot area is to improve industrial green total factor productivity (GTFP) 1 and achieve a win-win situation for energy conservation and green development. So, measuring the GTFP of the energy revolution reform pilot area has important practical significance.
Some scholars used environmental factors in the study of total factor productivity (TFP) to form GTFP. Mohtadi 1 took pollutant emissions as unpaid inputs, introduced them into the production function and measured the GTFP. Chung et al. 2 used directional distance function (DDF) and Malmquist-Luenberger (ML) productivity index to measure GTFP of Swedish pulp mills by taking pollution emissions as undesired output. After the publication of "Transforming Our World: The 2030 Agenda for Sustainable Development" by the United Nations in 2015, the research on GTFP is gradually increasing. Wang and Xie 3 and Sun and Yang 4 measured the GTFP of each province in China and conducted an empirical analysis of its influencing factors. Li and Liu 5 measured the GTFP of three city groups in China, analyzed its spatial correlation and investigated its influencing factors. Ni et al. 6 calculated the GTFP of 41 cities in the Yangtze River Delta region and analyzed the characteristics of temporal and spatial evolution. Parts of scholars have studied the GTFP of some specific industrial sectors. Liang and Long, 7 Guo and Liu, 8 and Liu et al. 9 measured the agricultural GTFP in China and analyzed its influencing factors; Ding et al. 10 measured the GTFP of Marine economy in 11 coastal areas of China; Chen et al. 11 measured GTFP of 36 industrial sectors in China; Teng 12 studied the spatial differentiation and driving factors of GTFP growth in China's service industry.
From the measurement of GTFP, Solow residual method, stochastic frontier analysis (SFA) method, and data envelopment analysis (DEA) method are often adopted by scholars. Some scholars use the Solow residual method to measure GTFP. Chen 13 estimated the change of TFP of China's industrial sectors with the Solow residual method considering CO 2 emissions. Some scholars use SFA to measure GTFP. Zhu et al. 14 constructed the environmental composite index to measure the relative green GDP and analyzed the efficiency of China's economic growth and its influencing factors under environmental constraints by using the SFA model. Kuang and Peng 15 studied China's environmental TFP by incorporating environmental factors into green economic growth accounting under the framework of SFA. Wang and Huang 16 studied GTFP in China from 2000 to 2010 by using the method of parameterized random boundary and Luenberger productivity index. More scholars use DEA to measure GTFP. Chen 17 used DDF and ML productivity index to measure industrial GTFP under resource and environment constraints, and studied the influencing factors of industrial GTFP 1 GTFP refers to green total factor productivity, which is referred to as GTFP in this paper. through a dynamic panel model. Lu et al. 18 21 measured the GTFP of 35 industrial industries in China by using the DEA-SBM model and analyzed its influencing factors.
Among them, Solow residual method has strong constraints on the function and cannot consider the inefficiency in the production process, which may lead to biased conclusions. 22 The SFA method is subjective and the validity of parameter estimation depends on a large number of sample data. Although the traditional DEA method has the advantage of being more objective and not requiring large-scale sample data, random error, and external environmental influence are ignored. This paper adopts the three-stage DEA model proposed by Fried et al. 23 This measurement has more advantages than other methods in efficiency measurement and influencing factor analysis, but its disadvantage is that environmental factors are not well considered.
In this paper, we use the three-stage DEA model to measure and analyze the GTFP of 18 industries in Shanxi Province. On the basis of this measurement, we also provide energy transformation policies for different industries to provide ideas for the transformation of other resource-based regions. The innovation is that the research area is considering the differences of energy consumption and output efficiency in the energy revolution reform pilot area.

| Modeling
Referring to the three-stage DEA model put forward by Fried et al., 23 the traditional three-stage DEA model is widely used in the measurement of GTFP. But, there are the following disadvantages: (1) DEA method only evaluates the relative efficiency of the decision-making unit (DMU), not the absolute efficiency. It cannot completely replace the analysis of absolute efficiency with the traditional ratio analysis method. (2) DEA method cannot measure the negative output. The linear model hypothesis simplifies DEA analysis, but the positive output is the premise of solving linear programming. If the output is negative, it cannot be measured by this method. This paper studies the relative efficiency of DMU and does not involve the negative output of industry.
This paper revises the GTFP efficiency of Shanxi industrial industry, so as to evaluate the efficiency of DMU.
The first stage: The traditional DEA model. The DEA-BCC model with variable-scale returns was proposed by Banker et al. 24 We used to this model obtain the initial GTFP value and input relaxation value of energy revolution reform pilot area in Shanxi Province, and the model is shown in formula (1).
Among them, X λ k k is the optimal mapping value of the nth input value of the ith DMU on the efficiency frontier. Y λ k k is the optimal mapping value of the nth output value of the ith DMU on the efficiency frontier. S + indicates a positive output indicator, and S − indicates a negative output indicator.
represents the m factor input of the k industrial industry. represents the s output index of the k industrial sector. θ is the total efficiency value, among, 0 < θ < 1, only when θ = 1, the industrial sector is at the forefront of efficiency, which indicates that the energy in this sector is used efficiently, otherwise it will not be used efficiently.
The second stage: The adjustment of input-output variables by the SFA model.
The SFA model is used to adjust the input. The purpose of the adjustment is to put all DMUs in the same external environment and random disturbance. This paper establishes a similar SFA regression model to separate the related factors that affect the efficiency value, and separate the slack variables of GTFP input into the functions of three independent variables, namely, environmental variables, random interference terms, and management inefficiency. The functions are as follows: Among, S ik represents the relaxation variable of the i input of the k industry. Z z z z = ( , , …, ) interference terms and management inefficiency terms, approaches to zero infinitely, which indicates that the efficiency value is greatly influenced by random errors. Random error and management inefficiency can be effectively separated from mixed error by the following formula: 1. The estimated values of β i , σ 2 , and γ are obtained by the maximum-likelihood estimation method.
where φ represents the density function of standard normal distribution, Ф is the distribution function of standard normal distribution, and e k is the error term. 3. Calculate the estimated value of random interference term V ik .
4. On the basis of the most effective DMU, adjust the investment of industries that have not achieved the ideal efficiency value.
among, y* ik and y ik are the adjusted value and the initial value of item i of the k industry.  β i is the estimated ik ik is to adjust the random errors of all industries to the same level. After adjustment, all industries are in the same external environment and random interference state.
The third stage: This stage measures the GTFP of all industries under the same research background. By using the traditional DEA model to analyze the input variables and initial output variables after SFA adjustment, we can get the real efficiency value of the industry excluding the influence of external environment and random interference.

| Variable selection
1. Input variable: The input variable selected is enterprise research and development input (R&D). The proportion coefficient obtained by formula (2) is multiplied by the total R&D input of industrial enterprises above the designated size, and the industrial R&D input is obtained. 2. Output variable: The output variable selected in this paper is industrial structure adjustment (ir). According to the practice of Wang et al., 25 the industrial structure (IS) adjustment is measured by the proportion of the industrial added value of different industries to regional GDP. 3. Environment variable: The environmental variable selected is the intensity of environmental regulation (er). According to the practice of Gray 26 and Lanoie et al., 27 the intensity of environmental regulation is measured by the proportion of pollution control costs to the total cost or total output value of enterprises. In this paper, the proportion of the total output value of a single industry to the total output value of all industries is calculated, and then multiplied by the total pollution control cost of the industry to calculate the pollution control cost of a single industry. Finally, the environmental regulation intensity of the industry is obtained by dividing the pollution control cost of a single industry by the industrial added value of the industry. 4. Technological progress: The technological progress variable selected is the investment amount of environmental pollution control (invest). To achieve the goal of energy saving and emission reduction, enterprises will increase investment in environmental pollution control and improve production technology. This chapter uses the investment amount of environmental pollution control by industry to measure the investment amount of environmental pollution control in this industry. 5. Industrial structure: The secondary industry is a high energy-consuming industry among the three major industries, and the development of each industry has the most direct influence on the energy utilization in this region. Therefore, this study takes the proportion of the secondary industry in GDP as an index to measure the IS.

| Variable description
Descriptions of related variables mentioned in the selection of variables are shown in Table 1.

| Data source
Considering the completeness and accessibility of data collection, this paper takes the input-output data of 18 industrial sectors in Shanxi Province in 2019 as the DMU. All data are from Shanxi Statistical Yearbook.

| Results of the first-stage efficiency measurement
On the basis of the input-output data of 18 industrial industries in Shanxi Province, the GTFP of these industries was decomposed into technical efficiency change index (EFFCH), technological progress change index (TECHCH), pure technical efficiency change index (PECH), scale efficiency change index (SECH), and return to scale (RTS). The results are shown in Table 2. Table 2 is the measurement result of GTFP without excluding the influence of external environment and random error factors. It can be seen that the GTFP of Shanxi's industrial sectors is at the middle level, with the average GTFP of 1.057, the average comprehensive technical efficiency of 1.048, the average pure technical efficiency of 1.008, and the average scale efficiency of 1.000, which means that the actual output level of R&D input of 18 industrial enterprises is higher than the optimal level. The results are as follows: 1. From the perspective of comprehensive technical efficiency, the comprehensive technical efficiency values of medium-polluting industry and light-polluting industry are 1.066 and 1.090, respectively, both of which are at the forefront of production, that is, these two industries are DEA efficient. However, the heavy-polluting industry is in the DEA inefficiency. The ineffective comprehensive technical efficiency is caused by the irrational allocation of resources, which is particularly evident from the mining and dressing of ferrous metals in the heavy-polluting industry. 2. From the pure technical efficiency, medium-polluting industry and light-polluting industry fully highlight good internal management level in the R&D input of enterprises. Among them, the automobile manufacturing is the best level. The agricultural and sideline food processing industry is in need of improving the pure technical efficiency. 3. From the perspective of scale efficiency, the heavypolluting industry and medium-polluting industry among the three major industries achieve the optimal state of scale efficiency. From the industrial, except for nonferrous metal mining and dressing industry, wine, beverage, and refined tea manufacturing, agricultural, and sideline food processing industry and general equipment manufacturing, all the other 14 industries have achieved the optimal state of scale efficiency. 4. From the perspective of scale returns, coal mining and washing industry, metal products industry, electricity, heat, gas, water production and supply industry, computer, communication, and other electronic equipment manufacturing, are in the stage of constant scale returns. Nonferrous metal mining and dressing industry, wine, beverage and refined tea manufacturing, and general equipment manufacturing are in the stage of diminishing returns to scale, which shows that the production allocation of enterprises in these industries is unreasonable, and too much investment has not brought ideal output. The remaining 11 industries are in the stage of increasing returns to scale.

| Results of industry category analysis
As shown in perspective of the industry, the light-polluting industry has the highest average GTFP, with an efficiency of 1.096, which is four percentage points higher than the average growth of all industries, it is related to its characteristics of high technology, high added value, and low emission. Among them, the efficiency value of computer communication and other electronic equipment manufacturing is the highest (1.174). Careful observation of the decomposition index shows that the TECHCH index contributes the most to the growth of GTFP in the light-polluting industry. This phenomenon is consistent with other subsectors of the light-polluting industry, and the TECHCH index dominates the growth of GTFP. The average GTFP of heavy-polluting industries dominated by heavy chemical industries and pollution-intensive industries is 0.994, among which the average GTFP of ferrous metal mining and dressing industry is the lowest (0.935). From the decomposition term, the main reason that restricts the growth of GTFP of heavypolluting industry is the low index of TECHCH (0.986). The index of TECHCH mainly reflects the additional increase of output and benefit brought by technological innovation. Therefore, in the future development of heavy-polluting industry in Shanxi Province, we should focus on strengthening technological innovation and improving the utilization rate of GTFP, so as to improve the GTFP and promote high-quality economic growth.
The average GTFP (1.089) of medium-polluting industries manufacturing life service products and some heavy-polluting industry products is slightly lower than that of light-polluting industries, and it is in the middle T A B L E 2 Results of the first-stage efficiency measurement. level of the three major industries. From the perspective of decomposition terms, the index value of TECHCH is 1.066, which mainly drives the growth of GTFP of medium-polluting industries. In summary, it is found that the technological progress index contributes the most to the GTFP growth of 18 industrial sectors in Shanxi Province, while the scale efficiency index is the lowest. Light-polluting industry should maintain technological innovation. Focus on the development of computer, communication, and other electronic equipment manufacturing. Heavypolluting industry should strengthen technological innovation and rely on new technologies and processes to improve production efficiency. While promoting technological innovation, medium-polluting industry should adjust the structure of the industrial production scale, and rely on expanding production scale and improving production management to promote the green growth of industrial economy.

| Results of the second-stage SFA regression
At this stage, the SFA method is adopted to eliminate the influence of environmental factors and random errors on GTFP, and only the influence of internal management is considered. Adjust the original input values of various industrial sectors in Shanxi Province, and get the GTFP level under the same environment. Taking environmental variables as explanatory variables and input relaxation variables obtained in the first stage as explanatory variables, the results are shown in Table 3.
As shown in Table 2, all likelihood ratios have passed the significance test at the level of 1%, that is, there is a significant correlation between input relaxation variables and environmental variables, which show that SFA regression, can be used to eliminate the influence of external environmental variables of various industries on GTFP. In addition, both sigmasquared and gamma values have passed the significance test, which indicates that environmental factors have a dominant influence on the GTFP of various industrial sectors compared with random errors. At the same time, most of the estimated values of each parameter are significant, which shows that environmental variables have a significant influence on the redundancy of R&D investment of enterprises in various industries. On the basis of this, it is necessary to strip off the influence of these environmental factors when measuring the GTFP of various industrial sectors in Shanxi Province. The influence of environmental variables on input relaxation variables can be seen from the coefficient values of environmental variables. The regression coefficient is positive, and the environmental variable is positively correlated with the input redundancy variable, which indicates that the increase of the environmental variable will produce more input redundancy, but not conducive to the promotion of the GTFP. If the regression coefficient is negative, it means that the increase of environmental variables will reduce the input redundancy, which is conducive to the improvement of GTFP in various industries.
1. Technological progress: This variable has a significant positive correlation with enterprise R&D input variables. The regression coefficient of the relaxation variable of enterprise R&D input is positive, which is inconsistent with the theoretical expectation. The possible reason is the "rebound effect" of R&D, it means that technological progress should improve the transformation efficiency of R&D achievements and save R&D costs, but at the same time, technological progress will also promote the economic development, then generate new demands for R&D, which partially or even completely offset the R&D costs saved by technological progress. In addition to this, technological progress should enhance the competitiveness and influence of enterprises, promote the scale expansion of enterprises, and thus create more employment opportunities for the industrial industry. 2. Industrial structure: The regression coefficients of this variable to the slack variable of enterprise R&D investment are all positive, which indicates that the increase of industrial industry in GDP will result in the waste of resources, which is not conducive to the promotion of GTFP. *, **, and *** are significant at the levels of 10%, 5%, and 1%, respectively.

| Results of the third-stage efficiency measurement
The SFA regression model is used to strip off the environmental factors and random errors that affect the GTFP of various industries. The input variables are adjusted by formula (5), the new input variables and initial output are substituted into the DEA-BCC model, and the efficiency levels of different DMUs are remeasured, and the efficiency measurement results of the third stage can be obtained, as shown in Table 4. The statistical chart of GTFP values of 18 industries is shown in Figure 1.
Comparing the results in Tables 2 and 4, we can see that the relative GTFP value and its structure of various industries have changed greatly before and after the input revision. In general, first, after the input correction, the average GTFP becomes 1.037 (Table 3), which is lower than the 1.056 before the input correction (Table 1), indicating that environmental variables have an impact on the GTFP. In addition, it can be seen from Table 4 that the GTFP of about 50% of industries is lower than the average of 0.706, which indicates that the efficiency in these industries is low, especially in heavypolluting industries. Second, it can be seen from Table 3 that the TECHCH index of various industries has changed obviously. Before the input correction, the average value of the TECHCH index was 1.048 and after the input correction was 1.004. Significant changes have taken place in the TECHCH index of various industries. Before the input correction, the average value of the SECH index was 1.048, and after the input correction, it T A B L E 4 Results of the third-stage efficiency measurement.

Category
Industry GTFP EFFCH TECHCH PECH SECH RTS was 0.982. The average pure technical efficiency decreased from 1.000 before the input correction to 0.988 after the input correction. This result shows that if environmental factors and random errors are not taken into account, the change of technological progress in energy utilization will be overestimated, and the change of scale efficiency will also be overestimated. The specific results are as follows: 1. From the analysis of GTFP, it can be seen that before and after the adjustment of input variables, the industries in the front of production have obviously changed, and the industries in the front of efficiency have decreased from 13 before the adjustment of input to 11 after the adjustment of input, among which 11 industries, such as food manufacturing, wine and beverage and refined tea manufacturing industry, computer, communication, and other electronic equipment manufacturing, are still in the front of efficiency, which indicates that the GTFP of these industries is relatively high, which is not mainly determined by their superior external environment. After the input revision, the coal mining and washing industry and metal products industry withdrew from the front of efficiency, indicating that the true GTFP level of these industries was overestimated. 2. From the analysis of TECHCH index, there is no industry whose pure TECHCH index value is the same before and after adjustment, which indicates that the external environmental factors will affect its TECHCH, and the main reason for its TECHCH index is external reasons. After the adjustment, the TECHCH index of ferrous metal mining and dressing industry, nonferrous metal mining and dressing industry, petroleum processing, coking and nuclear fuel processing industry, chemical raw materials, and chemical manufacturing industry increased, indicating that it was greatly influenced by the external environment, the TECHCH index was underestimated, and the real level was higher than before the adjustment. The TECHCH index values of most other industries are overestimated.

| DISCUSSION
To reflect the GTFP of various industrial sectors more clearly, this paper divides the industrial sectors under study into the following categories with the GTFP mean and pure technical efficiency mean as critical values: based on the GTFP mean, "high" industries refer to those with GTFP value higher than 1.037, and "low" industries refer to those with GTFP value lower than 1.037; based on the average value of pure technical efficiency, "high" industry refers to industries with pure technical efficiency higher than 0.988, and "low" industry refers to industries with scale efficiency lower than 0.988, as shown in Table 5. As shown in Table 5: "Double-high" industry: It divides regional types by taking the mean GTFP and the mean pure technical efficiency as critical values, and can express industries whose mean GTFP and the mean pure technical efficiency are higher than their mean values as "double-high" industries. These industries have relatively high GTFP, which is a relatively effective utilization mode of GTFP. 28 The GTFP and pure technical efficiency of these industries are higher than those of other industries due to their higher economic development level and technological advantages.
"High-low" industry: This kind of industry refers to the industries where the GTFP value is higher than the average value, but the pure technical efficiency value is lower than the average value. There are two industries, which show that these industries have a relatively high level of technology for GTFP, and the improvement of comprehensive technical efficiency mainly depends on the breakthrough of technological innovation. 29 "Low-high" industry: "Low-high" industry refers to the industries where the GTFP value is lower than the average value and the pure technical efficiency value is higher than the average value. There are two industries. The utilization situation of GTFP in these industries shows that the bottleneck of improving GTFP mainly lies in the small scale of R&D investment of enterprises. 30 It is necessary to expand the R&D investment of these industries to obtain the technical level of GTFP utilization.
"Double-low" industry: The remaining eight industrial sectors refer to "double-low" industry where the GTFP and pure technical efficiency are lower than the average value. The GTFP and pure technical efficiency of these industrial sectors have great room for improvement. In view of the future development, we should not only make a breakthrough in management innovation, but also make reasonable resource allocation and appropriately adjust the scale of enterprise R&D technology investment to promote the improvement of GTFP. 31

| CONCLUSION
In this study, the three-stage DEA model is used to measure the GTFP of industrial industry in the energy revolution reform pilot area, and the following conclusions are drawn: first, the investment of enterprises in agriculture, forestry, animal husbandry and fishery in energy revolution reform pilot area has promoted the growth of GTFP. Second, the investment of enterprises in heavily polluting industries has inhibited the growth of GTFP. Among them, the investment in coal mining and washing industry has the greatest inhibitory effect on GTFP, followed by the investment in nonmetallic mining, ferrous mining and nonferrous mining. Third, the six industries included in the manufacturing investment inhibit the increase of GTFP, the investment in petroleum processing, coking, and nuclear fuel processing industry has the greatest inhibition on GTFP, and the investment in the metal products industry has the least inhibitory effect on GTFP. In the last, investment in the service industry is divided into low-end producer service industry and high-end producer service industry, low-end producer service industry includes two major industry investments which promote the increase of GTFP. The financial industry investment included in high-end producer services has no significant impact on GTFP, but the investment in scientific and technological services has a positive effect on GTFP. Investment in information transmission, software, and information services is to restrain the increase of GTFP.
On the basis of the difference of pure technical efficiency average, 18 industrial industries are divided into four industries. Contribute to the introduction of different industry policies: "Double-high" industry optimizes R&D investment and reduces the demand for cheap labor. "High-low" industry reduces R&D investment and increases capital investment. "Lowhigh" industry achieves economies of scale and restrains T A B L E 5 Distribution of GTFP in various industrial sectors

Industry type
Industrial classification "Double-high" industry Rubber and plastic products industry, Automotive manufacturing, Pharmaceutical manufacturing industry, Instrument manufacturing, Computer, communication, and other electronic equipment manufacturing, and Electrical machinery equipment manufacturing "High-low" industry Tobacco industry, electricity, heat, gas, water production, and supply industry "Low-high" industry Food manufacturing and Metal products industry "Double-low" industry General equipment manufacturing, Agricultural and sideline food processing industry, Coal mining and washing industry, Ferrous metal mining and dressing industry, Nonferrous metal mining and dressing industry, Petroleum processing, coking and nuclear fuel processing industry, Wine, beverage and refined tea manufacturing industry, and Chemical raw materials and chemicals manufacturing industry Abbreviation: GTFP, green total factor productivity. excessive investment of resources. "Double-low" industry improves energy utilization efficiency.