The impact of industry integration on industrial energy efficiency in China

Abstract

This study examined the impact of industrial integration on industrial energy efficiency in China. We measured industrial energy efficiency using stochastic frontier analysis with Mundlak auxiliary equations and distinguished between effective and ineffective energy use. We empirically examined the causal relationship between industrial integration and industrial energy efficiency using Chinese provincial industrial sector data for 2004 to 2021. The estimates showed that increasing the industry integration level by 1% reduced industrial inefficient energy use by 1.640%, which demonstrates that industry integration improved industrial energy efficiency significantly. The mechanism test showed that industrial integration improved industrial energy consumption through the scale effect. However, industrial integration can improve production efficiency through the technique effect, promote industrial upgrading through the composition effect, and reduce industrial energy intensity. Additionally, the causal relationship between industrial integration and industrial energy efficiency showed regional heterogeneity. The scale effect caused by industrial integration, rebound effect of technological progress, and transregional industrial transfer weakened the energy-saving effect of industrial integration.

Keywords

industry integration industrial inefficient energy use industrial energy efficiency stochastic frontier analysis industrial energy consumption

Introduction

Since its reform and opening up, China has experienced profound economic transformation. The nation's contribution to the global economic aggregate has escalated from 1.7% in 1978 to a substantial 15.9% in 2022. Furthermore, China's aggregate industrial output has consistently been ranked top of international rankings for a consecutive 13-year period, underscoring its prominent position in the global economic landscape. However, the traditional pattern of economic growth comes at the cost of the environment; such growth cannot be sustained because it has resulted in problems of rapid energy consumption and high carbon emissions.^1,2 From the Statistical Yearbook of World Energy in 2022, China's total energy consumption increased from 17.18 exajoules in 1979 to 157.65 exajoules in 2021, ranking first among countries in total energy consumption. As the primary consumer of energy, the industrial sector has historically accounted for a large share of consumption.³ Between 1980 and 2021, industrial energy consumption as a proportion of overall energy consumption increased from 64.7% to 66.8%. As illustrated in Figure 1, China’s energy intensity and industrial energy intensity maintain the same downward trend. However, the ratio of industrial energy intensity to total energy intensity has been expanding, increasing from 1.78 in 1998 to 2.05 in 2021. This highlights the importance of reducing emissions and saving energy in the industrial sector, which remains the main goal of energy conservation and emissions reduction to achieve green development.

Faced with the double challenges of slowing economic growth and rising energy consumption, China has increased its efforts to upgrade the service industry to enhance energy efficiency.^4–7 The China Statistical Yearbook 2022 reveals remarkable growth in the added value of China's tertiary industry. Between 1980 and 2021, the sector's contribution to GDP increased from 96.6 billion yuan to 61,447.6 billion yuan, representing 54.9% of GDP in 2021. Among these services, producer services developed faster than others, accounting for 31.20% of total GDP. Given the escalating significance of producer services in China’s economic expansion, the government has outlined a strategy to prioritize the optimization of the economic structure within the service sector. This strategy aims to foster the comprehensive integration of advanced manufacturing and modern service industries. Industry integration, a novel development model for China's industrial economy, adds value to industrial products by incorporating professional services throughout the industrial production process. The integration of producer services and industry contributes to diminished production costs for industrial enterprises, providing favorable external conditions for industrial transformation and upgrading and promoting productivity.^8–10 The available literature primarily concentrates on the environmental effects of manufacturing servitization and producer service agglomeration, with limited research exploring the energy-saving potential of industrial integration. Investigating the latter and its underlying mechanisms would provide support for China to realize peak carbon emissions and attain carbon neutrality.

Thus, to address the gaps in the current literature, this study employed the industrial sector in China to investigate the influence of industrial integration on industrial energy efficiency. Moreover, we examined the specific mechanism linking industrial integration and industrial energy efficiency and investigated regional differences in causality between industrial integration and industrial energy efficiency. The main goal of this study is to provide Chinese industry with a useful guide for improving energy efficiency, developing a paradigm for low-carbon and green energy consumption, and assisting in industrial transformation and upgrading.

This study contributes to the existing literature in three ways. First, we used a coupling coordination model to quantify industrial integration and systematically evaluated the interaction between the industrial sector and producer services from the perspective of stocks and increments. Second, we constructed an industrial energy demand model using stochastic frontier analysis (SFA) and calculated energy efficiency as the ratio of actual to optimal energy consumption. Concurrently, we added auxiliary Mundlak equations to the regression model to provide a more accurate estimate of inefficient industrial energy use. Third, we decomposed the mechanisms that influence industrial integration and industrial energy consumption into three parts (scale, technique, and composition effects) and validated them using empirical data. This is beneficial for augmenting relevant theories concerning the energy-saving effects of industrial integration and provides a reference for the formulation of effective energy-saving policies.

The remainder of this paper is organized as follows. In the next section, we review the extant literature. Section “Model construction” describes the modeling framework and identification strategy in detail. Section “Data sources and description” describes the data used in this study. In Section “Results and discussion,” we discuss the results. Section “Conclusions” summarizes the findings and discusses policy implications.

Literature review

This study examined industrial integration and efficiency in the industrial sector. Thus, we review the relevant literature from these two perspectives.

Regarding industrial integration, the current literature concentrates on the interactive connection between producer services and manufacturing, measurement of the degree of integration of producer services and manufacturing, and the environmental effects of the integration of manufacturing and producer services. For example, existing literature examining the nexus between producer services and manufacturing can be categorized into four groups. The first group is based on the supposition that the manufacturing industry is in a basic position and producer services are in a supplementary position, which is a prerequisite for the emergence and development of producer services.^11,12 The second group emphasizes the influence of producer services, positing that producer services are the basis for improving manufacturing efficiency. Producer services can provide support for manufacturing industry to form a strong manufacturing force.^13,14 In the third group, the nexus between them is proposed to be interdependent, which jointly facilitates the coordination level of the two industries.^15–17 In the fourth group, researchers posit that as the information technology sector advances, producer services and the manufacturing sector persist in complementing and extending each other. The original industrial boundary will be gradually eliminated, and a trend of integrated development will be exhibited in the two industries.^18,19

To examine the level of producer services and manufacturing integration, researchers have predominantly employed methods involving correlation coefficients, input-output analysis, coupling evaluation models, the E-G industrial agglomeration index, and the Herfindahl index. For example, Ellison and Glaeser²⁰ introduced the Concentration Index of Industrial Space, which they used to quantify the degree of interindustry agglomeration. Gambardella and Torrisi²¹ used the Herfindahl index to investigate technology integration within the electronic information sector, utilizing patent data from various industries as a proxy. Fai and Tunzelman²² adopted correlation coefficients to assess the extent of technological integration of the chemical, electronic, mechanical, and transportation sectors in the United States, based on data from American enterprises. Xing et al.²³ measured the integration level of China's ICT industry using an input-output approach. Zhou et al.²⁴ applied a coupling evaluation model to reveal a coupling relationship between the automotive and digital industries.

Regarding the environmental effects of integrating the manufacturing sector with producer services, many researchers have focused on manufacturing and industrial collaborative agglomeration. White et al.²⁵ were the first to study the environmental impacts of servitization. They found that product-based services improve the environmental performance of products and reduce resource consumption and environmental pollution. Doni et al.²⁶ found that servitization led to an improvement in energy consumption that promoted environmental performance in a dataset of 208 European-listed manufacturing companies. Meng and Xu²⁷ discovered an inverse N-type correlation between industrial collaborative agglomeration and carbon intensity. In addition, studies have examined the energy-saving effects of producer service agglomeration, suggesting that it can improve energy efficiency and reduce carbon emissions intensity.^28,29

The current literature on industrial efficiency encompasses two primary dimensions: industry efficiency measurements and factors that influence efficiency. The available literature primarily uses two methods for industry efficiency measurements: data envelopment analysis (DEA) and stochastic frontier analysis (SFA). For instance, Alsaleh et al.³⁰ investigated the technical efficiency of the bioenergy sector across 28 European Union countries and used DEA to partition technical efficiency into pure technical efficiency and scale efficiency. Yang et al.³¹ utilized DEA to examine the influence of producer service agglomeration on energy efficiency. This measure has also been used by other researchers.^32–34 This approach eliminates the need to develop function equations, avoids structural biases caused by model and setting errors, and divides industry efficiency into technical and scale efficiencies. In contrast, other researchers have used SFA to measure industrial efficiency, assuming that random error will affect results; including technical inefficiencies and a random error term is hypothesized to improve the accuracy of technical efficiency assessment.^35–37

In terms of the factors that influence industrial efficiency, Li and Shi³⁸ examined the determinants of industrial energy efficiency in China and demonstrated that industrial structure, capital deepening, and R&D collectively contribute to energy efficiency. Alsaleh et al.³⁹ identified labor input, capital input, gross domestic product, and other factors that influence industrial technical efficiency. Wang et al.⁴⁰ found that industrial energy efficiency in Beijing was significantly affected by the energy consumption structure, R&D investment, and capital deepening. Liao and He⁴¹ analyzed the effects of R&D, energy consumption structure, and firm size on industrial energy efficiency. Guo and Yuan⁴² concluded that environmental regulation, R&D, urbanization, exports, and industrial structure affect industrial energy efficiency. Furthermore, some studies have suggested that industrial scale, energy prices, and marketization affect energy efficiency.^43–47

In summary, many researchers have investigated the nexus between producer services and the manufacturing industry from different perspectives. However, there is scope for further improvement. First, the extant literature primarily investigated energy-saving effects from the perspectives of manufacturing servitization and producer service agglomeration and rarely delved into the realm of industrial integration involving producer services and industry. As a product of the advanced stages of manufacturing servitization, industrial integration has a broader meaning and richer expression than manufacturing servitization. Second, the existing literature could not systematically evaluate the coordination relationship between two industries from the perspective of stock and increment because of inaccurate and incomplete information owing to small sample sizes. Therefore, we utilized an improved entropy method and coupling coordination model to quantify the level of industry integration between producer services and industry. Third, existing studies neglected the existence of random error when calculating energy efficiency using DEA, which led to discrepancies among results. SFA can include a random error term, which can permit accurate measurements of technical inefficiency terms. Moreover, the SFA method with the Mundlak auxiliary equation can distinguish inefficiency from cross-sectional heterogeneity, so that the inefficiency term can be measured more accurately.

Model construction

The IPAT model is an important method for examining the environmental effects of human activities, with representations of environmental impact (I), population (P), affluence (A), and technology (T).⁴⁸ Although the IPAT model facilitates the examination of the environmental effects resulting from different human activities, its framework is neither quantitative nor can be measured directly. Dietz and Rosa⁴⁹ advanced the IPAT model to develop the STRIPAT model. The model is configured as follows:

E = α \times P_{i t}^{β_{1}} \times A_{i t}^{β_{1}} \times T_{i t}^{β_{1}} \times ε_{i t}

(1)

where E represents environmental issues, such as industrial energy consumption, P denotes population, A denotes affluence, T indicates technology, and ε denotes the error term. We generate the following energy demand function by taking the natural logarithm of Equation (1):

Ln E = Ln α + β_{1} Ln P_{i t} + β_{2} Ln A_{i t} + β_{3} Ln T_{i t} + Ln ε_{i t}

(2)

Based on Equation (2), we referred to Sheng and Guo⁵⁰ to construct the following industrial energy demand function:

Ln (E_{i t}) = α + β_{1} Ln (Y_{i t}) + β_{2} Ln (P_{i t}) + β_{3} Ln ({KS}_{i t}) + β_{4} Ln ({EP}_{i t}) + {IE}_{i t} + ε_{i t}

(3)

where i denotes provinces, t represents the year, E refers to consumption demand of industrial energy, Y indicates industrial scale, P denotes population size, KS represents capital deepening, EP is the energy price, IE is inefficient use of industrial energy, and ε is the error term.

Following the relevant literature, we added environmental regulation (ER), trade (TR), internal R&D expenditure (RD), and marketization (MR) as control variables to reduce the influence of omitted variables on energy efficiency.^51,52 The inefficient use of industrial energy can be specified as follows:

{IE}_{i t} = λ + θ_{1} Ln ({CP}_{i t}) + θ_{2} Ln ({ER}_{i t}) + θ_{3} Ln ({TR}_{i t}) + θ_{4} Ln ({RD}_{i t}) + θ_{5} Ln ({MR}_{i t}) + ξ_{i t}

(4)

where IE represents the inefficient use of industrial energy, CP represents the integration level of industry and producer services, ER is environmental regulation, TR denotes trade, RD is internal R&D expenditure, and MR represents marketization. The SFA method was applied to estimate equation (3) and equation (4) simultaneously, and the model was set as follows:

Ln (E_{i t}) = α + β_{1} Ln (Y_{i t}) + β_{2} Ln (P_{i t}) + β_{3} L n ({KS}_{i t}) + β_{4} Ln ({EP}_{i t}) + μ_{i t} + ν_{i t}

(5)

Equation (5) includes two independent components. In this equation, μ is presumed to adhere to a half-normal distribution, representing inefficient energy consumption. The term ν reflects the effect that is unobserved in the model and follows a normal distribution. The inefficiency term in the model is expressed as follows:

{IE}_{i t} = \frac{p e_{i t} - p e_{i t}^{F}}{p e_{i t}} = \exp (- {\hat{μ}}_{i t})

(6)

where pe_it^F is the minimum energy use in the ith province at time t. When pe_it^F is equal to pe_it, industrial energy consumption reaches the production frontier and is optimal.

We utilized a pooled ordinary least square method with random effects to estimate Equation (5).^53,54 Nonetheless, both of these methodologies fail to differentiate the individual-specific effects from the inefficiency term, leading to inaccurate estimations.³⁶ To address this problem, Greene⁵⁵ proposed a true fixed effects model, disentangling time-varying inefficiency from unobserved heterogeneity. However, this method does not consider inefficiencies that do not change over time. In addition, in cases for which explanatory variables are associated with unobserved heterogeneity, estimates are biased.⁵⁶ Therefore, we added the Mundlak term to the random effects method, which efficiently distinguishes inefficiency from cross-sectional heterogeneity.⁵⁷ The Mundlak specification was set as follows:

\begin{aligned} ν_{i t} = {AX}_{i} π + γ_{i t}; {AX}_{i} = \frac{1}{T} \sum_{t = 1}^{T} X_{i t}; \\ γ_{i t} \sim N (0, δ_{γ}^{2}) \end{aligned}

(7)

where X_i denotes the vector of explanatory variables and AX _i signifies the average of the explanatory variables across the study period.

The two-step approach to estimating inefficient terms tends to produce biased estimates, because the variables that affect inefficient terms change the dispersion of the inefficiency term and the frontier function.^58,59 Therefore, equations (4), (5), and (7) may be estimated simultaneously.⁶⁰ The specific simultaneous estimation is shown in equation (8):

\begin{aligned} Ln (E_{i t}) = α + β_{1} Ln (Y_{i t}) + β_{2} Ln (P_{i t}) + β_{3} Ln ({KS}_{i t}) + β_{4} Ln ({EP}_{i t}) + μ_{i t} + ν_{i t} \\ μ_{i t} = λ + θ_{1} Ln ({CP}_{i t}) + θ_{2} Ln ({ER}_{i t}) + θ_{3} Ln ({TR}_{i t}) + θ_{4} Ln ({RD}_{i t}) + θ_{5} Ln ({MR}_{i t}) + ξ_{i t} \\ ν_{i t} = {AX}_{i} π + γ_{i t}; {AX}_{i} = \frac{1}{T} \sum_{t = 1}^{T} X_{i t} \\ ξ_{i t} \sim N (0, σ_{θ}^{2}); γ_{i t} \sim N (0, δ_{γ}^{2}) \end{aligned}

(8)

where θ₁ denotes the industrial integration influence coefficient, which quantifies the degree to which industry integration influences industrial energy efficiency. A negative coefficient indicates that industrial integration positively contributes to increasing industrial energy efficiency.

Data sources and description

Data

The panel dataset used in this study covered 30 Chinese provinces between 2004 and 2021. However, Tibet, Hong Kong, Macao, and Taiwan were excluded from the data selection because statistical information from these areas was unavailable. The dataset was collected from the China Energy Statistical Yearbook, China Industry Statistical Yearbook, and China Statistical Yearbook.

Dependent variable

We used the terminal energy consumption of the industrial sector to calculate the energy consumption. We considered the effects of industrial integration on industrial energy consumption from three perspectives—scale, technique, and composition—building on the work of Copeland and Taylor¹⁰ and Zhao et al.⁶¹ The scale effect was equal to industrial added value, the technique effect was calculated as the ratio between industrial energy consumption and industrial output value, and the composition effect was represented as the ratio of industrial added value to GDP.

Independent variable

We use the coupling coordination model, as suggested by Zhou et al.'s study, to assess the level of industry and producer service integration and development.²⁴ The formula was set as follows:

C = \frac{2 \sqrt{(X_{1} \times X_{2})}}{(X_{1} + X_{2})}

(9)

where C denotes the coupling degree, and X₁ and X₂ are the development levels of the producer services and industrial industries, respectively. Because of the availability of data, we chose the “wholesale and retail trades industry,” “transport, storage and post-industry,” “financial intermediation industry,” and “real estate industry” as the producer services industries. Referring to the criteria of comprehensiveness, scientific rigor, and operability, we built a comprehensive index system consisting of three primary and six secondary indicators (Table 1). The three first-level indicators were industrial scale, economic benefit, and development potential.

Table 1.

Quantitative evaluation system for industry and producer services.

First-level indicators	Second-level indicators	Definition	Unit
Industrial scale	The ratio of X added value	The ratio of X added value to GDP	%
Industrial scale	The proportion of X employment population	The ratio of X employment population to total employment population	%
Economic benefits	Industrial productivity	The ratio of added value to industrial employees	Yuan/person
Economic benefits	Average labor remuneration of employees	The ratio of total wages to employment population	Yuan
Development potential	The growth rate of fixed assets	The ratio of current fixed assets this year compared with last year	%
Development potential	Employment growth rate	The ratio of the number of people employed this year compared with last year	%

Note: X denotes industry or producer services.

It was necessary to standardize the original data because of differences in the nature, units, and orders of magnitude of the evaluation indicators. Therefore, we calculated the comprehensive evaluation level and allocated weights using an entropy method. The standardized equations are as follows:

x_{i j} = \frac{X_{i j} - X_{i j \min}}{X_{i j \max} - X_{i j \min}} positive indicators

(10)

x_{i j} = \frac{X_{i j \max} - X_{i j}}{X_{i j \max} - X_{i j \min}} negative indicators

(11)

where C measures the coupling level between industry and producer services in Equation (9). This can only reflect the proximity of the two industries and offers limited insight into their development levels. To examine the coupling coordination relationship between these industries more extensively, a model was constructed as follows:

\begin{aligned} D = & \sqrt{C \times T} \\ T = & a X_{1} + b X_{2} \end{aligned}

(12)

where D denotes the coupling coordination level between two industries, which reflects the integration level. T represents a comprehensive assessment index reflecting the total degree of development of the two industries. In addition, α and β indicate the importance of these industries to the coordination degree. Considering the equal importance in both industry coupling coordination models, we set both α and β to a value of 0.5.

Control variables

Based on the extant literature, we added the following control variables: industrial scale was represented by industrial added value, which represents the level of industrial development. Additionally, population was expressed in terms of employment in industry. Capital deepening was denoted by the proportion of net fixed assets of industrial enterprises to industrial employees. Energy price was expressed as the purchasing price of fuel and power in the industrial sector. Environmental regulation was expressed by dividing sulfur dioxide emissions by the gross regional product. Trade was quantified by computing the total imports and exports as a percentage of GDP. R&D was calculated by dividing industrial R&D expenditure by industrial added value. Marketization was measured as the share of the main business income between state-owned enterprises and industrial enterprises above the designated size. We took the natural logarithms of all variables (Table 2).

Table 2.

Summary statistics.

Variables	Mean	Max.	Min.	S. D.
Industrial energy consumption	8.809	10.407	6.050	0.780
Industry integration	−0.616	−0.124	−1.087	0.186
Industrial scale	8.181	10.549	4.589	1.134
Population	5.929	7.767	3.399	1.028
Capital deepening	2.621	4.209	0.843	0.617
Energy price	4.649	4.947	4.224	0.104
Environmental regulation	−3.870	−3.424	−4.606	0.206
Trade	−1.694	0.537	−4.876	0.983
R&D	−3.699	−2.661	−5.864	0.552
Marketization	−1.048	−0.177	−2.345	0.533
Industrial energy intensity	0.627	2.282	−1.291	0.683
Industry structure	−1.069	−0.556	−2.301	0.287

Industry integration and industrial energy intensity

We adopted a scatter diagram to study the correlation between industrial integration and industrial energy intensity (Figure 2). An apparent negative relationship existed between the level of industrial integration and industrial energy intensity. Furthermore, we used Pearson and Spearman correlation coefficients to test the relationship between industrial integration and industrial energy intensity. The Pearson and Spearman correlation coefficients were significant at the 1% significance level, with values of −0.606 and −0.611, respectively. The typical features of Figure 2 show that industrial integration may be conducive to industrial energy efficiency. Therefore, we examined the energy-saving effect of industry integration using regression analysis.

Results and discussion

Results

To estimate Equation (8), we used the following methods: ordinary least squares (OLS), random effects (RE), fixed effects (FE), SFA, and stochastic frontier analysis with auxiliary equations (SFM). As shown in Table 3, the estimates sigma-squared by SFA and SFM were significant, which indicates that the OLS, RE, and FE methods produced biased estimates. The 1% significance of lambda values based on SFA and SFM supports the conclusion that the inefficiency term is sufficiently large relative to the random error term. Meanwhile, the auxiliary equation variable coefficients estimated using the SFM method were significant, indicating that the explanatory variables were closely related to unobservable effects. Hence, SFM was more effective than SFA at distinguishing between the cross-sectional heterogeneity effect and the inefficiency term.

Table 3.

Tests for the SFM estimations.

Variables	OLS	RE	FE	SFA	SFM
Variables	OLS	RE	FE	SFA	Master equation	Auxiliary equation
Industrial scale	−0.060 (0.043)	0.260*** (0.027)	0.251*** (0.027)	−0.060 (0.043)	0.055* (0.033)	0.112 (0.466)
Population	0.862*** (0.057)	0.431*** (0.045)	0.440*** (0.042)	0.862*** (0.056)	0.774*** (0.042)	−0.580 (0.651)
Capital deepening	0.660*** (0.050)	0.269*** (0.035)	0.281*** (0.034)	0.660*** (0.049)	0.582*** (0.037)	2.821*** (0.672)
Energy price	0.444*** (0.164)	0.152** (0.066)	0.159** (0.067)	0.444*** (0.163)	0.448*** (0.101)	−1.843*** (0.540)
Constant	0.389 (0.825)	2.715*** (0.427)	2.673*** (0.415)	0.389 (1.749)	−0.157 (0.512)
Sigma²				0.264***(0.028)	0.261***(0.029)
Lambda				1.592***(0.058)	1.498***(0.061)
AIC	502.231	−576.181	.	506.231	268.338

Note: ***, **, and * represent p < 0.01, p < 0.05, and p < 0.1, respectively.

We estimated Equation (8) using the SFM method based on the model selection results above. We added control variables to the model in turn to examine its robustness. The results are robust because there were no discernible changes in the sign or significance of the model variables (Table 4). Furthermore, the coefficients of capital deepening and energy price variables in the auxiliary equation were significant, suggesting that the explanatory variables were closely related to unobservable effects. However, the coefficients of the auxiliary equation have no economic meaning.⁶² Compared with the results of the four groups of estimations, the estimation of SFM (4) is better than the other results according to the Akaike information criterion.

Table 4.

SFM estimations for industrial energy consumption.

Variables	SFM (1)		SFM (2)		SFM (3)		SFM (4)
Variables	Master equation	Auxiliary equation	Master equation	Auxiliary equation	Master equation	Auxiliary equation	Master equation	Auxiliary equation
Industrial Efficient Energy Use
Industrial scale	0.143***(0.033)	−0.142(0.530)	0.141***(0.034)	−0.168(0.535)	0.145***(0.033)	−0.156(0.527)	0.150***(0.033)	−0.132(0.532)
Population	0.689***(0.045)	−0.301(0.763)	0.692***(0.045)	−0.265(0.771)	0.689***(0.044)	−0.277(0.759)	0.696***(0.044)	−0.219(0.766)
Capital deepening	0.559***(0.036)	2.931***(0.730)	0.559***(0.036)	2.961***(0.742)	0.550***(0.036)	2.938***(0.733)	0.537***(0.034)	3.104***(0.754)
Energy price	0.399***(0.093)	−1.834***(0.560)	0.396***(0.092)	−1.852***(0.569)	0.389***(0.092)	−1.847***(0.563)	0.374***(0.090)	−2.078***(0.576)
Constant	−0.080(0.481)		−0.070(0.480)		−0.036(0.478)		−0.027(0.456)
Industrial Inefficient Energy Use
Industry integration	−2.984***(0.615)		−3.089***(0.620)		−2.558***(0.663)		−1.640**(0.834)
Environment regulation	2.562***(0.530)		2.561***(0.528)		2.411***(0.520)		2.255***(0.539)
Trade			−0.139*(0.075)		−0.142*(0.074)		−0.118(0.073)
R&D					−0.308*(0.173)		−0.293*(0.172)
Marketization							0.642**(0.249)
Constant	5.931***(1.880)		5.621***(1.878)		4.260**(1.961)		5.119***(1.865)
Sigma²	0.279***(0.030)		0.279***(0.030)		0.277***(0.030)		0.272***(0.030)
Lambda	1.663***(0.058)		1.654***(0.058)		1.629***(0.059)		1.549***(0.061)
AIC	222.971		221.555		220.863		214.471

Note: ***, **, and * represent p < 0.01, p < 0.05, and p < 0.1, respectively.

The estimation of SFM (4) shows that industry integration had an adverse effect on inefficient energy consumption in the industrial sector. The industrial integration coefficient was −1.640, which is statistically significant. This suggests that a 1% increase in industry integration reduced inefficient energy consumption by 1.640%. Industrial integration can reduce inefficient industrial energy consumption without reducing economic output, thereby improving industrial energy efficiency. Our findings are consistent with those of Yang et al.⁶³ and Zhu et al.⁶⁴ That is, industrial integration benefits the environment by enhancing energy efficiency and realizing energy savings and emissions reduction.

Generally, industrial integration has three impacts on inefficient industrial energy consumption: technological progress, resource allocation, and industrial internal structure optimization. First, industrial integration breaks the boundaries of the original industrial development space, accelerates the transregional flow of knowledge and technology, and reduces enterprises’ transaction costs. Influenced by advanced knowledge and technology, industrial integration promotes the dissemination of knowledge and technology by gathering technical talent, thereby accelerating the application of green technology in enterprise production.^29,47 Therefore, industrial integration reduces energy use intensity and enhances energy efficiency through technological progress. Second, industrial integration provides professional intermediate services for industrial enterprises and reduces transaction costs. Industrial integration can form an industrial division of labor and core competitive advantage, and change the original organizational framework and factor allocation combination of enterprises. Industrial integration has continuously embedded service factors such as knowledge capital and talent capital into enterprises, optimized the original factor input structure, reduced dependence on energy factors, and thus reduced the intensity of energy consumption. Third, as crucial support for industrial upgrading, industrial integration optimizes the pivotal role of information technology in transformation and upgrading. Driven by information technology, industrial integration can refine an industry's internal structure, decrease the ratio of high-consumption energy and heavy-pollution industries, and achieve green and low-carbon transformation.⁶⁵

However, it should be noted that the resource allocation effect and industrial upgrading effect of industrial integration are essentially caused by technological innovation. Technological innovation caused by industrial integration changes the original factor input structure, and results in a resource allocation effect through the substitution of physical production factors. Similarly, industrial integration helps strategic emerging sectors grow faster by spreading knowledge and technology more rapidly. This makes it easier to optimize an industry's internal structure.

Furthermore, we analyzed the effects of efficient industrial energy use, which cannot be reduced without reducing economic output. The industrial scale coefficient was significantly positive, indicating that the causality between industrial development level and energy consumption was positive. Decoupling has not yet been realized.⁶⁶ The coefficient of population was positive and significant, indicating a positive correlation between the industrial population and efficient industrial energy use. This may be due to the higher energy consumption of labor-intensive industries compared to capital-intensive industries, indicating that increasing the proportion of the former could lead to an increase in overall energy consumption. The coefficient of capital deepening was positive and significant, indicating that excessive capital deepening may not promote energy efficiency enhancement but may aggravate energy consumption. The results indicate that it is difficult to reduce energy consumption by increasing energy prices. The main reason for this is that China's energy market is a seller's market, and rising energy consumption lags rising energy prices.

We also investigated the effects of inefficient industrial energy use, which can be reduced without reducing economic output. The environmental regulation coefficient indicates that environmental regulations can reduce inefficient industrial energy consumption. This finding suggests that environmental regulations force firms to innovate by increasing production costs, thereby improving energy efficiency. The trade coefficient indicates that trade was conducive to industrial energy efficiency. The negative coefficient of R&D indicated a significant reduction in industrial energy use. An increase in R&D helped enterprises generate new technologies and information, facilitate the application of green technologies in industrial production, and improve energy efficiency. Marketization appears to increase inefficient industrial energy use, as indicated by the positive and significant marketization coefficient. One possible reason is that state-owned enterprises are monopolistic in nature. These entities exhibit decreased production efficiency in comparison to private enterprises, coupled with a higher proportion of capital, labor, and energy inputs per unit of output, which leads to inefficiencies in the production process.

Mechanism test

This study examined the precise influence of industrial integration on the improvement of industrial energy efficiency to investigate the energy-saving effects of industrial integration. According to research conducted by Copeland and Taylor¹⁰ and Zhao et al.,⁶¹ we examined the impact of industrial integration on industrial energy consumption from three different perspectives: scale effect, technique effect, and composition effect. To scientifically and effectively analyze the three effects of industrial integration, this analysis used least squares and three-stage least squares methods for regression analysis (Table 5). As shown in Table 5, the results of the ordinary least squares method were consistent with those of the three-stage least squares method. Considering that the three-stage least squares method can avoid potential endogeneity problems, we interpreted the regression results based on this method. As can be seen from Table 5, there existed positive causality between industrial integration and industrial added value, while a negative relationship prevailed between industrial integration and both industrial energy intensity and industrial added value as a share of GDP. This shows that industrial integration can promote industrial development through the scale effect and can also promote industrial structure upgrading through the composition effect. Industrial integration can simultaneously decrease energy consumption per unit of industry-added value through technological effects, thereby improving energy efficiency.

Table 5.

Mechanism test of the impact of industry integration on industrial energy consumption.

	OLS			3SlS
	lnScale	lnTech	lnComp	lnScale	lnTech	lnComp
Industry integration	2.547***(0.065)	−1.513***(0.135)	−0.261***(0.075)	2.428***(0.066)	−1.719***(0.130)	−0.267***(0.075)
Trade	−0.014(0.012)	−0.031(0.021)	0.026**(0.012)	−0.002(0.012)	−0.031(0.021)	0.025**(0.012)
Industrial energy consumption	0.259***(0.024)			0.472***(0.022)
Energy price	0.011(0.111)	0.021(0.204)	0.327***(0.114)	0.014(0.110)	0.010(0.203)	0.326***(0.113)
Population	0.720***(0.018)			0.538***(0.016)
R&D		−0.463***(0.045)	0.075***(0.025)		−0.331***(0.039)	0.078***(0.025)
Environment regulation	−0.064(0.060)	0.323***(0.103)	0.416***(0.057)	−0.193***(0.058)	0.369***(0.102)	0.417***(0.057)
Constant	3.123***(0.537)	−3.048***(0.898)	−0.821(0.576)	1.510**(0.592)	−0.324(1.023)	−0.803(0.572)

Note: ***, **, and * represent p < 0.01, p < 0.05, and p < 0.1, respectively.

Specifically, in the scale effect model, a 1% increase in industrial integration increased industrial added value by 2.428%. As a new development mode of industrial economy, industrial integration further stimulates industrial agglomeration and produces a scale economy in the process of development.⁶⁷ Under the effect of scale economy, industries can improve the degree of information asymmetry between industries, reduce transaction costs, and promote the rapid growth of industrial output scale.⁶⁸ However, the increase in the scale of industrial output leads to the demand for physical elements and increases the total energy consumption.

In terms of technique effect, industrial energy intensity decreased by 1.719% for a 1% increase in industrial integration. This indicates that industrial integration can reduce industrial energy intensity through technical effects. Industrial integration breaks the boundaries of the original industrial development space, contributes to the cross-regional flow of knowledge and technology, reduces transaction costs, and raises the level of enterprise innovation.¹⁰ Influenced by advanced knowledge and technology, industrial integration facilitates the dissemination of knowledge and technology by gathering technical talent. Industrial integration promotes the application of green technologies in industrial production, reduces the intensity of industrial energy use, and improves an enterprise's energy efficiency.

Regarding the composition effect model, a 1% increase in industrial integration reduced industrial structure by 0.267%. Industrial integration can reduce energy consumption through industrial upgrades. This shows that industrial integration is conducive to promoting industrial transformation and upgrading and driving the transfer of the industrial industry from resource-consuming and labor-intensive to knowledge-intensive and capital-intensive industries. That is, industrial integration contributes to facilitating the evolution from high-emission, low-value-added industrial production chains towards low-emission, high-value-added scenarios, diminishing high-polluting and energy-intensive enterprises, while continuously elevating industrial energy efficiency.^69,70

Analysis of regional heterogeneity

Given these results, we assessed the industrial energy efficiency of China's provinces based on the SFM (4) estimates in Table 4. As shown in Figure 3, China’s industrial energy efficiency in 2021 was 68.8%, indicating that China could reduce industrial energy consumption by 31.2% without reducing economic development. In addition, industrial energy efficiency also declined slightly during 2004–2021, which stood at 70.1% in 2004. At the same time, industrial energy efficiency differed across provinces (Figure 3). In 2021, five provinces had industrial energy efficiencies of more than 90%. However, the provinces with the worst industrial energy efficiency performances were Shanxi, Hebei, Ningxia, Liaoning, Shandong, and Inner Mongolia, all of which were below 55 percent. The energy-intensive and high-pollution industrial structure of these provinces explains their low industrial energy efficiencies. From 2004 to 2021, industrial energy efficiency improved in 17 provinces, including Guizhou, Henan, and Hunan. In contrast, industrial energy efficiency decreased in the remaining 13 provinces, with Shandong, Liaoning, and Fujian experiencing the largest declines. Finally, we conclude that industrial energy efficiency has not improved in China as a whole, and that there are 13 provinces with no significant improvement in industrial energy efficiency between 2004 and 2021.

Based on the SFM (4) estimates in Table 4, we also studied the causality between industrial integration and industrial energy efficiency in China from 2004 to 2021 (Figure 4). As shown in Figure 4, the causality between industrial integration and industrial energy efficiency presented clear regional heterogeneity. The regression coefficient associated with the integration of industries and industrial energy efficiency across the 19 provinces showed a positive trend, indicating that industrial integration contributed to the enhancement of industrial energy efficiency in Beijing, Shanghai, Guangdong, and other provinces. This result is consistent with the results reported earlier in this section. Conversely, the regression coefficients for industrial integration and industrial energy efficiency in 11 provinces were negative. These provinces included Tianjin, Jiangsu, and Zhejiang, which have good economic development levels, and Heilongjiang, Shanxi, and Xinjiang, which have poor economic development levels. For the former, these provinces are superior to the central and western provinces in terms of GDP, industrial structure, marketization level, and per capita disposable income, with a higher level of industrial integration. There are two reasons for this negative causality between industrial integration and energy efficiency. From one perspective, high industrial development and opening levels amplify the scale economy effect engendered by industrial integration, expanding the total amount of industry and increasing industrial energy consumption. In contrast, technological advancement imposes downward pressure on energy prices, reducing the relative cost of energy, thereby stimulating businesses to allocate increased energy resources as substitutes for other production factors, leading to elevated energy intensity. That is, technological progress in these provinces triggered a rebound effect, thereby increasing energy intensity. For the latter, these provinces were predominately concentrated in coal, steel, chemical, and other energy-consuming industries. The level of industrial integration stimulates economies of scale in the industrial sector and increases the total industrial output, resulting in a growing demand for energy. Meanwhile, the eastern coastal provinces transferred high-energy and high-pollution industries to these provinces, further increasing their industrial energy intensity.

Thus, industrial integration can reduce industrial energy intensity through technological progress and composition effects, producing a positive spillover impact on industrial energy efficiency. However, the causality between industrial integration and industrial energy efficiency shows clear regional heterogeneity. The scale effect caused by industrial integration, the rebound effect of technological progress, and industrial transfer in other provinces weaken the energy-saving effect of industrial integration.

Conclusions

We examined the impact of industrial integration on industrial energy efficiency in China from 2004–2021. For this purpose, we utilized an improved entropy method and a coupling coordination model to quantify the integration level of industry and producer services. We empirically tested the spillover effect of industry integration on industrial energy efficiency using stochastic frontier analysis. In addition, we used a three-stage least squares approach to study the impact path in industrial integration to improve industrial energy efficiency. Finally, we analyzed the impact of industrial integration on industrial energy efficiency in different provinces.

The main conclusions are as follows. (1) Industrial integration had a positive impact on industrial energy efficiency. Every province achieved a 1.640% decrease in inefficient energy consumption with a 1% increase in industrial integration. Industrial integration can reduce inefficient industrial energy use and improve industrial energy efficiency, without reducing economic output. (2) The mechanism test results showed that industrial integration helped increase industrial economic scale and energy consumption through the scale effect. However, the effects of the technical progress and composition caused by industrial integration reduced industrial energy consumption. (3) China's industrial energy efficiency in 2021 was 68.8%. Industrial energy efficiency improved in 17 provinces from 2004 to 2021, and industrial integration ensured energy conservation in the industrial sector. However, the industrial energy efficiency of 13 provinces showed a downward trend, and the industrial sector still has considerable room for energy conservation. (4) The causal relationship between industrial integration and industrial energy efficiency presented clear regional heterogeneity. The scale effect caused by industrial integration, the rebound effect of technological progress, and industrial transfer in other provinces weakened the energy-saving effect of industrial integration.

Our results provide an economic rationale for policymakers seeking to coordinate sustainable economic development through energy conservation, as follows. (1) Accelerating the integration of modern service sectors with advanced manufacturing is imperative, as evidenced by the positive impact of industrial integration on industrial energy efficiency. The government should actively facilitate the industrial integration of producer services and industry and introduce corresponding policies to support its rapid development. It is necessary to accelerate stripping of the service functions of manufacturing enterprises, especially advanced manufacturing enterprises, and encourage them to integrate their resource advantages. (2) Every province should promote an integrated level of producer services and industry in different ways according to its industrial structure characteristics. The eastern region should combine its factor endowment, comparative advantage, manufacturing structure, and the demand of leading industries and rationally divide and scientifically plan productive service functions to make full use of the energy-saving impact of industrial integration. The central and western provinces should formulate a development plan for producer services based on the development needs of leading industries and facilitate the integration of industries with specialized agglomerations of characteristic producer services to improve industrial efficiency. (3) Industrial integration in the eastern provinces facilitates industrial upgrading and accelerates the transregional transfer of industries. The central and western provinces actively undertake industrial transfer from the eastern provinces to stimulate economic growth. However, spurred by growth-oriented local government performance evaluation systems, governments have relaxed environmental regulations to attract investment in energy-intensive industries and increase energy intensity. Thus, the central and western provinces can establish a coordination mechanism between regional investment and environmental governance and strictly formulate the environmental assessment standards of local governments in the process of investment introduction to achieve sustainable economic development in the region and form a mutually beneficial pattern of investment introduction. (4) All provinces should facilitate coordinated low-carbon development of the energy supply and consumption sides according to their resource endowment characteristics. The eastern region should reasonably transfer the energy-rebounding effect caused by technological progress to the central and western provinces. This could not only reduce the rebound effect of the eastern provinces, but also reduce energy intensity by using abundant renewable energy in the central and western provinces.

This study has some limitations. First, the producer services selected in this study did not include information technology services because statistical data were not available, which may have had an unknown impact on the results. Second, industrial integration can promote technological progress and bring about a rebound effect, but limited by the availability of data, its specific role is difficult to estimate directly.

Footnotes

Acknowledgements

The authors would like to thank Thomas A. Gavin, Professor Emeritus, Cornell University, for help with editing this paper.

Author contributions

Haohao Wei did funding acquisition, conceptualization, data curation, methodology, software, validation, visualization, writing—original draft.

Sheng Ding did supervision, software, visualization, writing—review and editing.

Code availability

Stata 16.

ORCID iDs

Haohao Wei

Sheng Ding

Data availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Graduate Talent Program of Henan University (grant numbers SYL20060105, SYLYC2022010)

Haohao Wei is currently pursuing the PhD degree with the School of Economics, Henan University, Kaifeng, China. His research interests include energy economics, industrial economics, and environmental management.

Sheng Ding is currently pursuing the PhD degree with the School of Economics, Henan University, Kaifeng, China. His research interests include new energy vehicles, digital economics, and big data analysis.

References

. Energy investment, economic growth and carbon emissions in China—empirical analysis based on spatial Durbin model. Energy Policy 2020; 140: 111425.

Shahbaz

Song

Ahmad

, et al. Does economic growth stimulate energy consumption? The role of human capital and R&D expenditures in China. Energy Econ 2022; 105: 105662.

Xiong

. The impact of industrial structure efficiency on provincial industrial energy efficiency in China. J Clean Prod 2019; 215: 952–962.

Favi

Germani

Landi

, et al. Comparative life cycle assessment of cooking appliances in Italian kitchens. J Clean Prod 2018; 186: 430–449.

Landi

Gigli

Germani

, et al. Investigating the feasibility of a reuse scenario for textile fibres recovered from end-of-life tyres. Waste Manage 2018; 75: 187–204.

Tian

Bai

Jia

, et al. Realizing low-carbon development in a developing and industrializing region: impacts of industrial structure change on CO2 emissions in southwest China. J Environ Manage 2019; 233: 728–738.

Wang

Song

Duan

. Coupling between energy efficiency and industrial structure: an urban agglomeration case. Energy 2021; 234: 121304.

Arnold

Javorcik

Mattoo

. The productivity effects of services liberalization: evidence from the Czech Republic. World Bank Working Paper 2006: 1–38.

Goldhar

Berg

. Blurring the boundary: convergence of factory and service processes. J Manuf Technol Mana 2010; 21: 341–354.

10.

Zhao

Dong

. How does producer services’ agglomeration promote carbon reduction?: the case of China. Econ Model 2021; 104: 105624.

11.

Francois

Woerz

. Producer services, manufacturing linkages, and trade. J Ind Compet Trade 2008; 8: 199–229.

12.

Macpherson

. Producer service linkages and industrial innovation: results of a twelve-year tracking study of New York state manufacturers. Growth Change 2008; 39: 1–23.

13.

Eswaran

Kotwal

. The role of the service sector in the process of industrialization. J Dev Econ 2002; 68: 401–420.

14.

Restuccia

Yang

Zhu

. Agriculture and aggregate productivity: a quantitative cross-country analysis. J Monetary Econ 2008; 55: 234–250.

15.

Fuentes

. On the limits of the post-industrial society structural change and service sector employment in Spain. Int Rev Appl Econ 1999; 13: 111–123.

16.

Preissl

. The German service gap or: re-organising the manufacturing––services puzzle. Metroeconomica 2010; 58: 457–478.

17.

Francois

Hoekman

. Services trade and policy. J Econ Lit 2010; 48: 642–692.

18.

Eberts

Randall

. Producer services, labor market segmentation and peripheral regions: the case of Saskatchewan. Growth Change 1998; 29: 401–422.

19.

Goe

. Factors associated with the development of nonmetropolitan growth nodes in producer services industries, 1980–1990. Rural Sociol 2010; 67: 416–441.

20.

Ellison

Glaeser

. Geographic concentration in US manufacturing industries: a dartboard approach. J Polit Econ 1997; 105: 889–927.

21.

Gambardella

Torrisi

. Does technological convergence imply convergence in markets? Evidence from the electronics industry. Res Policy 1998; 27: 445–463.

22.

Fai

Von

. Industry-specific competencies and converging technological systems: evidence from patents. Struct Change Econ D 2001; 12: 141–170.

23.

Xing

Kui

. Measuring convergence of China’s ICT industry: an input–output analysis. Telecommun Policy 2011; 35: 301–313.

24.

Zhou

Wang

. Innovative coupling and coordination: automobile and digital industries. Technol Forecast Soc 2022; 176: 121497.

25.

White

Stoughton

Feng

. Servicizing: the quiet transition to extended product responsibility. Tellus Instit Boston 1999; 97: 1–89.

26.

Doni

Corvino

Martini

. Servitization and sustainability actions. Evidence from European manufacturing companies. J Environ Manage 2019; 234: 367–378.

27.

Meng

. Can industrial collaborative agglomeration reduce carbon intensity? Empirical evidence based on Chinese provincial panel data. Environ Sci Pollut Res 2022; 29: 61012–61026.

28.

Yang

Zhang

. Exploring the effect of producer services agglomeration on China’s energy efficiency under environmental constraints. J Clean Prod 2020; 263: 121320.

29.

Liu

Zhang

Sun

. Does the agglomeration of urban producer services promote carbon efficiency of manufacturing industry? Land Use Policy 2022; 120: 106264.

30.

Alsaleh

Abdul-Rahim

Mohd-Shahwahid

, et al. An empirical analysis for technical efficiency of bioenergy industry in EU28 region based on data envelopment analysis method. Int J Energy Econ Policy 2016; 6: 290–304.

31.

Yang

Zhang

. Exploring the effect of producer services agglomeration on China’s energy efficiency under environmental constraints. J Clean Prod 2020; 263: 121320.

32.

Alsaleh

Abdul-Rahim

Mohd-Shahwahid

, et al. Frontier production function and cost efficiency empirical analysis of bioenergy industry in EU28 region. Soc Sci 2016; 11: 5318–5332.

33.

Alsaleh

Abdul-Rahim

Mohd-Shahwahid

. Determinants of technical efficiency in the bioenergy industry in the EU28 region. Renew Sust Energ Rev 2017; 78: 1331–1349.

34.

Lin

Zhou

. Does energy efficiency make sense in China? Based on the perspective of economic growth quality. Sci Total Environ 2022; 804: 149895.

35.

Xie

Bai

Wang

. Accessing provincial energy efficiencies in China’s transport sector. Energy Policy 2018; 123: 525–532.

36.

Sheng

Zhai

Zhang

, et al. The effects of urbanization on household wastewater emissions in China: efficient- and inefficient-emissions. Environ Pollut 2020; 267: 115350.

37.

Tan

. How does economic agglomeration affect energy efficiency in China?: evidence from endogenous stochastic frontier approach. Energy Econ 2022; 108: 105901.

38.

Shi

. Energy efficiency analysis on Chinese industrial sectors: an improved super-SBM model with undesirable outputs. J Clean Prod 2014; 65: 97–107.

39.

Alsaleh

Abdul-Rahim

. Determinants of cost efficiency of bioenergy industry: evidence from EU28 countries. Renew Energy 2018; 127: 746–762.

40.

Wang

Shi

Zhang

. Energy efficiency and influencing factors analysis on Beijing industrial sectors. J Clean Prod 2017; 167: 653–664.

41.

Liao

. Exploring the effects of influencing factors on energy efficiency in industrial sector using cluster analysis and panel regression model. Energy 2018; 158: 782–795.

42.

Guo

Yuan

. Different types of environmental regulations and heterogeneous influence on energy efficiency in the industrial sector: evidence from Chinese provincial data. Energy Policy 2020; 145: 111747.

43.

Fan

Liao

Wei

. Can market oriented economic reforms contribute to energy efficiency improvement? Evidence from China. Energy Policy 2007; 35: 2287–2295.

44.

Herrerias

Cuadros

Orts

. Energy intensity and investment ownership across Chinese provinces. Energy Econ 2013; 36: 286–298.

45.

Filipović

Verbič

Radovanović

. Determinants of energy intensity in the European Union: a panel data analysis. Energy 2015; 92: 547–555.

46.

Peng

Zhang

Wang

, et al. Energy efficiency and influencing factor analysis in the overall Chinese textile industry. Energy 2015; 93: 1222–1229.

47.

Lin

. How to promote energy efficiency through technological progress in China? Energy 2018; 143: 812–821.

48.

Ehrlich

Holdren

. Impact of population growth. Science 1971; 171: 1212–1217.

49.

Dietz

Rosa

. Effects of population and affluence on CO2 emissions. Proc Natl Acad Sci USA 1997; 94: 175–179.

50.

Sheng

Guo

. Energy consumption associated with urbanization in China: efficient-and inefficient-use. Energy 2018; 165: 118–125.

51.

Zhao

Lin

. Impact of foreign trade on energy efficiency in China’s textile industry. J Clean Prod 2020; 245: 118878.

52.

Lin

Zhu

. Energy efficiency of the mining sector in China, what are the main influence factors? Resour Conserv Recycl 2021; 167: 105321.

53.

Aigner

Lovell

Schmidt

. Formulation and estimation of stochastic frontier production function models. J Econometr 1977; 6: 21–37.

54.

Pitt

Lee

. The measurement and sources of technical inefficiency in the Indonesian weaving industry. J Dev Econ 1981; 9: 43–64.

55.

Greene

. Reconsidering heterogeneity in panel data estimators of the stochastic frontier model. J Econometr 2005; 126: 269–303.

56.

Farsi

Filippini

Kuenzle

. Unobserved heterogeneity in stochastic cost frontier models: an application to Swiss nursing homes. Appl Econ 2005; 37: 2127–2141.

57.

Mundlak

. On the pooling of time series and cross section data. Econometrica 1978; 46: 69–85.

58.

Kumbhakar

Lovell

. Stochastic frontier analysis. 1st ed. New York: Cambridge University Press, 2002.

59.

Greene

. Econometric analysis. 7th ed. Upper Saddle River, NY: Prentice Hall, 2012.

60.

Wang

Schmidt

. One-step and two-step estimation of the effects of exogenous variables on technical efficiency levels. J Prod Anal 2002; 18: 129–144.

61.

Copeland

Taylor

. North-South trade and the environment. Quart J Econ 1994; 109: 755–787.

62.

Baltagi

. An alternative derivation of Mundlak’s fixed effects results using system estimation. Econometr Theor 2006; 22: 1191–1194.

63.

Yang

Zhang

. Exploring the effect of producer services and manufacturing industrial co-agglomeration on the ecological environment pollution control in China. Environ Dev Sustain 2021; 23: 1–26.

64.

Zhu

Zhang

. Does industrial collaborative agglomeration improve environmental efficiency? Insights from China’s population structure. Environ Sci Pollut Res 2022; 29: 5072–5091.

65.

You

Zhang

. How heterogeneous technological progress promotes industrial structure upgrading and industrial carbon efficiency? Evidence from China’s industries. Energy 2022; 247: 123386.

66.

Sheng

Zhai

, et al. Coupling of economic growth and reduction in carbon emissions at the efficiency level: evidence from China. Energy 2020; 213: 118747.

67.

Krugman

. Increasing returns and economic geography. J Polit Econ 1991; 99: 483–499.

68.

Daniels

. The locational geography of advanced producer services firms in the United Kingdom. Prog Plann 1995; 43: 123–138.

69.

Han

Zhang

Huang

, et al. Does industrial upgrading promote eco-efficiency? – A panel space estimation based on Chinese evidence. Energy Policy 2021; 154: 112286.

70.

Xue

, et al. Impacts of industrial structure adjustment, upgrade and coordination on energy efficiency: empirical research based on the extended STIRPAT model. Energy Strateg Rev 2022; 43: 100911.