Has Trade Liberalization Promoted Energy Efficiency in Enterprises?

Abstract

Improving energy efficiency is a way for China’s economy to achieve sustainable development. Using data on listed industrial enterprises in China from 2000 to 2022, we investigate the impact of trade liberalization on the energy efficiency of Chinese firms and its mechanism of action at the micro-firm level. Our findings suggest that trade liberalization has greatly contributed to firms’ energy efficiency and the main conclusions in this paper still hold after a series of robustness tests. Further research shows that two channels that contribute to the improvement of firms’ energy efficiency are the expansion of the firm’s size and the improvement of production efficiency induced by trade liberalization. Another significant route according to which trade liberalization influences energy efficiency improvements by firms is the dynamic decomposition of industrial energy efficiency, which shows that 70.07% of industrial energy efficiency enhancement is due to structural adjustments. The heterogeneity analysis results show that the energy saving impact of trade liberalization is stronger for non-exporting, foreign-funded, growth-oriented enterprises that are based in the eastern region. Our study presents new ideas for the realization of energy conservation and emissions reduction from the perspective of trade liberalization, which is instructive in regard to the improvement of sustainable development of enterprises.

1. Introduction

Since 1978, China’s economy has developed significantly, but, at the same time, the conflict between energy supply and energy demand is more and more obvious. Economic development still relies on the “high-consuming, low-producing” development model [1]. According to the latest data from the BP Statistical Yearbook of World Energy 2022, China is the world’s largest energy consumer, ranking first in terms of total energy consumption for 17 consecutive years, with China’s growth in regard to its energy consumption accounting for one-third of the world’s growth in terms of energy consumption in 2021 alone. However, China’s energy efficiency is still relatively low when compared with developed countries. China’s excessive consumption of the world’s total energy, and especially its low level of energy efficiency, has caused a series of ecological and environmental problems. The country’s poor energy efficiency is becoming a key factor limiting China’s sustainable economic development. How to improve energy efficiency is one of the most significant and urgent issues for China’s economy, in order to reduce energy consumption and achieve its green development goals. Faced with such a serious contradiction between energy supply and demand, the Chinese government clearly proposed to reduce the country’s energy consumption, per unit of GDP, by 20%, as early as in the 11th Five-Year Plan, and, once again, clearly proposed to reduce the country’s energy consumption, per unit of GDP, by 15% by 2020, on the basis of the reduction achieved in 2015, according to the 13th Five-Year Plan in 2017. On the other hand, since China’s accession to the WTO, China’s trade liberalization process has further accelerated. In terms of the average tariffs in the industry, they have declined from 17.4% in 1998 to 7.1% in 2022. This raises a question that deserves our attention, namely has trade liberalization improved the energy efficiency of firms? And, if so, what is the mechanism of effect of trade liberalization? This will be the next question to be explored in-depth in this paper.

Research on the trade–environment relationship has a long history. Interdisciplinary concerns, such as environmental economics and international economics, have been a focus and a hotspot, but research on this issue has not reached a consistent conclusion. According to the existing literature, there are three different perspectives on the relationship between trade and the environment: the first perspective supports the “pollution haven hypothesis”, which argues that tougher environmental laws in developed nations lead to polluting industries relocating to developing nations where environmental laws are weaker, thus leading to environmental degradation in developing countries [2,3]. The second view is that the technological effects of trade reduce pollution in developing countries, contrary to the pollution sanctuary hypothesis [4]. The third view falls in-between the previous two views, that the relationship between trade and environmental pollution is not a simple linear relationship, but that it is an inverted U-shaped relationship. In particular, in the initial stage of economic development, trade leads to increased pollution, but then with further economic development, people’s awareness of environmental protection increases and, thus, this stage of trade looks to reduce pollution emissions, the so-called “Environmental Kuznets Curve” [5].

In addition, another strand of the literature relevant to this paper is research on energy efficiency. From the perspective of the existing literature on China’s energy impacts, the initial research focuses on exploring and improving the methods for measuring energy efficiency [6]. There are two main aspects to the development of energy efficiency measurement methods. One is to use the single-factor energy efficiency index to express the energy efficiency of enterprises, specifically the ratio of the total industrial output value of the enterprise to the energy input or the ratio of the energy input to the total industrial output value, and the representative literature, such as the work by Lyubich et al., has used this method to calculate energy efficiency [7]. The single-factor energy efficiency method is simple to calculate, easy to understand, and easy to operate. It allows different levels of decomposition in order to examine the contribution of various factors to changes in energy efficiency. The disadvantage is that it only considers energy input and does not include production factors that affect the output, such as labor and capital, which has been criticized by some scholars [8]. In order to overcome the shortcomings of the single-factor energy efficiency method, Li et al. proposed measuring energy efficiency by using an all-factor energy efficiency approach, in which they simultaneously incorporated productivity factors, such as energy, labor, and capital, into the framework of classical production theory to measure energy efficiency, while taking into account the substitution effects between various input production factors, which is characterized by an integrated multidimensional approach [9]. As a result, the proposed method seems to be more comprehensive in measuring energy efficiency and has gradually begun to be adopted by scholars, such as Geng, Zhang, and Zhou, who use the all-factor energy efficiency method to measure energy efficiency [10,11]. With the improvement and maturity of energy efficiency measurement methods and the availability of data, scholars have begun to turn to the study of factors that impact energy efficiency.

Regarding the influencing factors on energy efficiency, the current research by scholars mainly focuses on the following aspects: first, in regard to the impact of environmental regulation research, existing research on whether environmental regulation promotes energy efficiency has not formed a consistent conclusion; based on Porter’s theory, some scholars have found through the use of a large number of case studies that, subject to being reasonably designed, the stringency of environmental regulation and enterprise energy efficiency usually have a positive relationship [12,13,14]. However, according to the cost of compliance theory proposed by the neoclassical school of economics, it is argued that environmental regulations can increase the cost of the associated clean up and compliance with institutional requirements, thus inhibiting their energy efficiency [15]. Second, research focuses on the impact of technological innovation, which enhances energy efficiency by improving the progress made in regard to green technologies, promotes upgrades to industrial structures, and also enhances the energy efficiency of neighboring cities through spillover effects [16,17,18]. Third, the impact of international trade research is an area of focus, mainly including global value chains and intermediate goods imports [19]. In this regard, there are also some scholars who analyze energy efficiency from the point of view of trade liberalization, which influences enterprises’ energy efficiency mainly through the technological effect, scale effect, and the effect of resource reallocation [20]. However, the empirical results of trade, on whether liberalization can improve energy efficiency, are inconsistent. In addition, the existing studies and measurements on trade liberalization and energy efficiency are at the macro level, which may distort the true causal relationship between the variables [21]. Therefore, there is a need to adjust the level of study to further discuss, in-depth, the influence of trade liberalization on the elements of enterprises’ energy efficiency.

Existing studies on energy efficiency in China have three obvious shortcomings: first, the energy efficiencies used are derived from macro-level aggregated data at the provincial or industry level, which do not reflect firm heterogeneity and may be subject to the fallacy of aggregation, and there are few studies using micro-level data. Second, none of the above studies involved testing the impact mechanism or dynamic decomposition of energy efficiency at the aggregate level, making it difficult to understand the intrinsic dynamics of international trade on energy efficiency improvement at a deeper level, from the perspective of resource reallocation. Third, most of the limited micro-level data studies use pollution databases on Chinese industrial firms, which were produced a long time ago and may not be a realistic representation of the current situation.

The main objective of this paper is to identify the impact of trade liberalization on the energy efficiency of firms. In comparison with the literature to date, the following aspects of this paper can complement previous studies: (1) measuring enterprises’ energy efficiency and examining the effects of trade liberalization on firms’ energy efficiency and the related internal mechanism, using micro-level data at the enterprise level in China. It is found that the energy efficiency of companies is greatly promoted by trade liberalization. (2) The results of this study suggest that the two routes to enhancing enterprises’ energy efficiency are the technology effect and the scale effect brought about by trade liberalization. (3) In order to quantify the contribution of structural adjustments induced by trade liberalization, to decompose energy efficiency at the industry level, the dynamic decomposition method by Melitz and Polance is used [22]. The decomposition findings suggest that 70.07% of the improvement in energy efficiency at the industry level can be attributed to the resource reallocation effect. (4) Our study uses data on listed industrial enterprises in China from 2000 to 2022 to overcome the problem concerning the sample period in the relevant literature which is relatively long, so that the conclusions in this study are more realistic and have guiding significance.

2. Research Design

2.1. Theoretical Hypotheses

Trade liberalization refers to the reduction of tariff barriers, the suppression of non-tariff barriers, the elimination of obstacles and discrimination in international trade, and the facilitation of the exchange and production of goods and services around the world, which is most directly reflected in the significant reduction of tariff rates through the adoption of trade treaties and agreements between countries, either multilaterally or bilaterally [23]. In the context of globalization, free trade can affect an enterprise’s energy efficiency in many ways and the factors present the complexity of working together, so it is worthwhile for scholars to study the relationship between free trade and the energy efficiency of enterprises from different perspectives. From the existing literature, trade liberalization mainly influences the firms’ energy efficiency through three channels, namely the technology effect, scale effect, and resource reallocation effect [24].

2.1.1. Technological Effect

Increased income from trade liberalization can increase people’s awareness of energy conservation [25], while the process of free trade is often accompanied by the flow of advanced management experience and energy-saving technologies across regions, and the advancement of technology and management will lead to more efficient use of inputs, thus enhancing enterprises’ energy efficiency, which is known as the “technology effect” [26,27,28]. Arias and Colmenarez used industry-level data in the U.S. and found that energy efficiency improvements in the U.S. are driven by technological progress in firms [29]. He and Huang used time-series data in China and found that energy efficiency improvements, since the reform and opening up of China, has been driven by technological progress, which is the main channel for energy efficiency improvement [30]. Casey conducted a similar study and also confirmed that technological advances are a key driver of energy efficiency improvements [31].

2.1.2. Scale Effect

With the expansion of economic scale brought about by trade liberalization, companies are taking advantage of economies of scale to purchase expensive energy-efficient equipment, reduce heat sinks, and recycle waste heat to improve their energy efficiency, the so-called “scale effect” [32]. Under the conditions of free trade, the size of the foreign market is much larger than the size of the domestic market, and various factors and products necessary for production are exchanged in the international market, which further expands the scale of enterprises [33]. So far, most of the literature on free trade and energy efficiency is based on the “scale effect” perspective. Meșter et al. found that energy efficiency improvements benefited from the increase in market size after trade liberalization, by studying EU trade agreements [34]. Similar studies by Huang and Pei also confirmed the significant contribution of economies of scale to energy efficiency [35].

Hypothesis 1.

Trade liberalization promotes corporate energy efficiency through “scale effects” and “technology effects”.

2.1.3. Structural Adjustment Effect

Trade liberalization leads to a reallocation of resources through structural adjustments at the sectoral level, which in turn affects energy efficiency [36]. Cherniwchan discussed the influence of trade liberalization-induced firm entry and exit on pollution emissions, based on U.S. data, to test for the existence of the resource reallocation effect [37]. The influence of resource reallocation effects on the energy efficiency of firms is derived from the structural dividend hypothesis, according to which there are differences in productivity, competitiveness, market share, and the allocation and transfer of various resources among different firms, that will lead to changes in the energy efficiency of the whole economy [26]. Specifically, the resource reallocation effects induced by trade liberalization include the intra-firm effect, inter-firm effect, firm entry effect, and firm exit effect [32]. The intra-firm effect refers to the fact that trade liberalization can optimize the allocation of resources within individual firms, changing the technological level, competitiveness, and other characteristics of surviving firms. The inter-firm effect refers to the fact that trade liberalization can adjust the allocation of resources among firms and, through changes in market shares, increase the flow of resources to more productive and competitive firms, and as the market share of such firms expands, the energy efficiency of the entire sector will increase [38]. The firm entry effect and the firm exit effect refer to the fact that trade liberalization can change the entry and exit thresholds of firms, leading to the entry of new firms and the exit of existing firms. Typically, there are significant differences in the energy efficiency of firms with different levels of productivity and competitiveness. As a result, trade liberalization induces firms to enter and exit the market, which in turn causes a reallocation of resources and affects the energy efficiency of the entire industry. The energy efficiency of the entire sector will increase if trade liberalization leads to the entry of more productive and competitive firms and the exit of less productive and competitive firms [24]. In summary, through resource reallocation effects, including intra-firm, inter-firm, entry, and exit effects, trade liberalization also affects the energy efficiency of firms and even entire industries. Therefore, the following hypothesis is proposed in our study:

Hypothesis 2.

Trade liberalization leads to structural shifts at the industry level, resulting in resource reallocation effects that impact industries’ energy efficiency, but the effects are uncertain and depend on a combination of intra-firm, inter-firm, entry, and exit effects.

2.2. Model Selection

Examining the effects of trade liberalization on a firm’s energy efficiency is the core of the research in our study. Based on the existing literature, this paper sets up the econometric model as follows:

$${tfee}_{,fit}={\alpha }_0+{\alpha }_1{Tariff}_{it}+{\gamma X}_{fit}+{\beta }_r+{\beta }_f+{\beta }_t+{\beta }_i+{\epsilon }_{fit}$$

(1)

where subscript f represents the firm, i represents three quarters of the industry, and t represents the year. Moreover, ${tfee}_{,fit}$denotes the energy efficiency of enterprise f in period t. The energy efficiency measurement method is the all-factor method and the specific measurement process is detailed below. ${Tariff}_{it}$ is the core explanatory variable in this paper, which indicates how much trade has been liberalized in three quarters of the industry, and the specific method used to calculate this will also be explained below. $X_{fit}$ denotes the firm-level control variables, including the financing constraints (loan), using the interest expense and fixed assets of the ratio +1 to make the natural logarithm represent the financing constraints; environmental regulation (regulation), using the wastewater treatment costs and the ratio of industrial output value +1 to make the natural logarithm represent the environmental regulation; capital density (KL) as fixed assets and the number of employees in terms of the ratio +1 to make the natural logarithm represent the capital density; corporate profitability (profit), with corporate profits and the industrial output value ratio +1 used to make the natural logarithm represent the corporate profitability; and the state-owned enterprises dummy variable (state), where 1 indicates state-owned enterprises and 0 for non-state-owned enterprises.

2.3. Key Variable Design

2.3.1. Measuring All-Factor Energy Efficiency

In this paper, coal consumption, capital, and labor are selected as inputs, and the firm’s total industrial output is taken as the consensual output. In order to measure the all-factor energy efficiency of a firm, the firm’s production function is assumed to be the Cobb–Douglas productivity function:

$$Y_{ft}=A_{ft}{L}_{ft}^{\alpha }{K}_{ft}^{\beta }{E}_{ft}^{\theta }$$

(2)

In Equation (2), $Y_{ft}$, $L_{ft}$, $K_{ft},$and $E_{ft}$ represent the output, labor input, capital input, and the firm’s energy input, respectively, and $A_{ft}$ represents the all-factor energy efficiency (tfee) of the enterprise. Simultaneous logarithmic treatment of Equation (2) yields the following linear regression equation:

$$y_{ft}=l_{ft}+k_{ft}+e_{ft}+u_{ft}$$

(3)

In Equation (3), the logarithmic forms of $Y_{ft}$, $L_{ft}$, $K_{ft},$and $E_{ft}$ are denoted by $y_{ft}{,l}_{ft}{,k}_{ft}$, and $e_{ft}$, respectively, and $u_{ft}$ is the residual term that contains information about the enterprise’s all-factor energy efficiency (tfee) after calculating the logarithm. The calculation of Equation (3) can provide an estimate of the tfee at the enterprise level, but the method inevitably suffers from measurement problems involving sample selection bias and simultaneity bias in estimating the all-factor energy efficiency of enterprises. In order to overcome the abovementioned problems, scholars have proposed a variety of solutions, which have led to the derivation of a variety of estimation methods for all-factor energy efficiency, such as the Levinsohn–Petrin method (LP), and the method involving fixed effects estimation techniques (FE). In this paper, coal consumption, capital, and labor are selected as inputs, and the firm’s output is taken as the consensus output, and the LP method is used to measure the enterprise’s TFE. Of course, this paper also measures energy efficiency using the fixed effects estimation technique method (FE), the OLS estimation method, and the single-factor method, which form the robustness tests in this paper.

2.3.2. Trade Liberalization Calculations

Final goods tariffs are used as a measure in the benchmark regression analysis in this paper. Following Amiti and Konings, this paper uses final goods tariffs to measure a three-quarter industry-level indicator of trade liberalization, using the following formula [39]:

$${Tariff}_{it}={\sum }_{k\in {\varnothing }_i}{num}_{kt}{importtax}_{kt}/{\sum }_{k\in {\varnothing }_i}{num}_{kt}$$

(4)

where i represents the industry, t represents the year, k represents the HS96-coded product, ${\varnothing }_i$ is the industry’s product collection, ${importtax}_{kt}$ is the import tariff rate for HS96-coded products, and ${num}_{kt}$ is the tariff line for HS96-coded products.

2.3.3. Data Description

The research in this paper mainly involves two sets of data: the first set of data is from industrial enterprises listed in China (2000–2022), which covers some key indicators in terms of the enterprises, such as the statement of assets and liabilities, the statement of cash flow, and other key information. The second data set is import tariff data from the WTO’s Tariff Download Facility database, which is mainly used to calculate trade liberalization indicators. The study in this paper requires the fusion of two data sets, which is conducted by drawing on the research idea by Brandt to merge the tariff rates at the product level to the three-quarter industry level and merge the tariff data into the panel data of listed companies based on the adjusted three-quarter industry codes, which ultimately results in the panel data required for the research in this paper (

Table 1. Description of the variables used in this study.

Symbol	Variable	Definition
tfee	Energy efficiency	Calculated by LP, +1 equates to the natural logarithm
tariff	Trade liberalization	Calculated from tariff data, +1 equates to the natural logarithm
loan	Financial constraint	The ratio of interest expense to fixed assets, +1 equates to the natural logarithm
regulation	Environmental regulation	The ratio of wastewater treatment costs to the total industrial output value, +1 equates to the natural logarithm
KL	Capital intensity	The ratio of fixed assets to the number of employees, +1 equates to the natural logarithm
profit	Enterprise profitability	The ratio of corporate profits to the total industrial output, +1 equates to the natural logarithm
state	State-owned enterprise	Value 1 when state owned, otherwise the value is 0

and

Table 2. Summary of the statistics.

Variable	Obs	Mean	SD	Min	Max
tfee	33,628	1.556	0.280	0.847	2.351
tariff	33,628	2.293	0.482	0.076	4.105
loan	33,628	2.462	0.770	0.693	4.232
regulation	33,628	0.127	0.130	0.886	1.347
KL	33,628	4.287	1.253	0.004	12.671
profit	33,628	1.293	0.129	0.008	2.643
state	33,628	0.168	0.374	0.000	1.000

) [40].

3. Test Results

3.1. Benchmark Regression Results

This section empirically tests trade liberalization’s energy efficiency effects, considering that differences in the firm, year, industry, and region may impact the firm’s energy efficiency. To reduce the impact of these factors on the regression results, our study controls for firm-fixed, year-fixed, industry-fixed, and region-fixed effects in all the regression test models. To help ensure the accuracy of the regression results, the benchmark regression process uses a stepwise addition of firm-level control variables in regard to the regression.

Table 3. Benchmark regression results.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
tariff	−0.0289 ***	−0.0291 ***	−0.0299 ***	−0.0296 ***	−0.0285 ***	−0.0283 ***
	(−7.01)	(−7.07)	(−7.28)	(−7.23)	(−7.03)	(−7.02)
loan		0.0076 ***	0.0068 ***	0.0070 ***	0.0062 ***	0.0064 ***
		(5.01)	(4.55)	(4.65)	(4.20)	(4.30)
regulation			0.1975 ***	0.1962 ***	0.1799 ***	0.1794 ***
			(42.23)	(41.89)	(38.83)	(38.81)
KL				0.0041 ***	0.0025 ***	0.0025 ***
				(5.56)	(3.42)	(3.39)
profit					0.2772 ***	0.2773 ***
					(67.56)	(67.57)
state						−0.0058 **
						(−2.07)
cons	1.6221 ***	1.6040 ***	1.5824 ***	1.5645 ***	1.5641 ***	1.5648 ***
	(171.51)	(158.46)	(156.85)	(147.69)	(149.41)	(149.40)
j. R-sq	33,628	33,628	33,628	33,628	33,628	33,628
	0.558	0.558	0.563	0.563	0.573	0.573

Note: **, and *** indicate significance at 5% and 1% levels, respectively. The t statistic values are reported in parentheses. All regressions are controlling firm-fixed, year-fixed, industry-fixed, and region-fixed effects.

shows the results obtained from the regression that is based on Equation (1), as the base model. Column (1) adds only trade liberalization to the baseline regression equation and the regression test results suggest that the regression coefficient of trade liberalization is significantly negative at the 1% level, implying that trade liberalization contributes significantly to companies’ energy efficiency. In column (2), financing constraints are added to the regression equation based on column (1). The findings suggest that the coefficient of financing restrictions is significantly positive at the 1% level, indicating that financing restrictions promote energy efficiency improvements in enterprises, with the reason being that enterprises get loans from banks to relieve the pressure of financing restrictions faced by enterprises, so that enterprises have funds to purchase new production equipment to improve the energy utilization of the enterprise. Based on column (2), column (3) introduces an environmental regulation coefficient. The regression results suggest that the regression coefficient of environmental regulation is significantly positive at the 1% level, which indicates that the implementation of environmental regulation contributes to the improvement of the energy efficiency of enterprises. In order to manage the environmental pressures from environmental regulations, firms actively carry out technological innovation, eliminate inefficient production equipment, and use clean energy, thus helping to improve their energy efficiency. Capital intensity is added to column (3) for the regression in column (4). The findings suggest the capital intensity coefficient is significantly positive at the 1% level, indicating that more capital-intensive firms are more energy efficient. Column (5) is regressed based on column (4), with the addition of the firm’s profitability. The results suggest that the regression coefficient of profitability is significantly positive at the 1% level, implying that the higher the profit, the higher the energy efficiency of the company. Column (6) adds a dummy variable as to whether the firm is a state-owned enterprise and the regression is carried out based on column (5). The findings suggest that the coefficient of state-owned enterprise is significantly positive at the 1% level, implying that state-owned enterprises have a significant inhibiting effect on the energy efficiency of enterprises. This is because state-owned enterprises have poor management, lack incentives and sovereignty, and lack any impetus to improve energy efficiency. On the other hand, China’s energy market is not yet fully marketized and state-owned enterprises (SOEs), which have the advantage of strong political affiliations, pay very low costs to acquire energy resources. They are more likely to use low-cost energy as a substitute for labor and capital, leading to wasted energy [41].

3.2. Endogenous Test

3.2.1. Sample Selectivity Bias

In this paper, we use Heckman’s two-stage endogeneity testing method to obtain consistent estimates, by overcoming the problem of sample selection bias [42]. The method consists of two steps: the first step is to select the equation, which is estimated using the Probit model, and the Mills ratio is obtained by estimating the values described below. In the second step, the determinants of the firm’s energy efficiency are analyzed and the Mills ratio is added to the baseline regression equation. The model is designed based on existing relevant studies, as follows:

$$Probit\left({port}_{fit}=1\right)={\alpha }_0+{\gamma X}_{fit}+{\beta }_r+{\beta }_f+{\beta }_t+{\beta }_i+{\epsilon }_{fit}$$

(5)

In Equation (5), ${port}_{fit}$ denotes the dummy variable on whether the match is on or off (1 for firms matched, 0 for firms not matched). The meanings of the other control variables are consistent with the previous benchmark model. Columns (1) and (2) in

Table 4. Endogeneity test results.

Variable	(1)	(2)	(3)	(4)	(5)
	Heckman Two-Stage Method		IV-2SLS
	Probit		IV	Key Explanatory Variable in Lag	All Variables in Lag
tariff	−0.6435 ***	−0.0300 ***	−0.1780 *	−0.0437 **	−0.0509 **
	(−175.40)	(−7.38)	(−1.52)	(−2.18)	(−2.50)
loan	0.2778 ***	0.0082 ***	0.0070 ***	−0.0011	−0.0008
	(123.54)	(5.26)	(4.47)	(−0.52)	(−0.42)
regulation	−0.3051 ***	0.1788 ***	0.1807 ***	0.1821 ***	0.0046
	(−21.91)	(38.52)	(38.55)	(29.25)	(0.68)
KL	0.1052 ***	0.0028 ***	0.0025 ***	0.0070 ***	0.0067 ***
	(69.18)	(3.80)	(3.40)	(6.88)	(6.16)
profit	−0.1242 ***	0.2771 ***	0.2765 ***	0.2529 ***	0.0808 ***
	(−10.31)	(67.50)	(66.47)	(43.06)	(13.16)
state	0.4790	−0.0038	−0.0058 **	−0.0067 *	0.0045
	(92.12)	(−1.32)	(−2.05)	(−1.71)	(1.03)
Mills Ratio		0.0095 ***
		(3.96)
KP-LM			418.3110	1.1 × 104	1.0 × 104
			[0.0000]	[0.0000]	[0.0000]
Wald rk F			230.0560	8914.5440	8916.6040
			[0.0000]	[0.0000]	[0.0000]
cons	−1.9612 ***	1.5487 ***
	(−65.47)	(137.81)
N	46,791	33,628	33,628	21,032	21,032
j. R-sq	0.232	0.573	0.343	0.337	0.377

Note: *, **, and *** indicate significance at 10%, 5%, and 1% levels, respectively. t statistic values are reported in parentheses.

present the findings of Heckman’s two-stage regression test. The findings suggest that the core conclusions in this paper still hold, as the regression coefficients for trade liberalization pass the 1% significance test and are negative after controlling for potential sample selection bias issues. In addition, the regression results show that the regression coefficients of the inverse Mills ratio pass the 1% significance test, indicating that there is a sample selection bias problem in the baseline regression analysis in this paper, suggesting that it is reasonable to consider the effect of sample selection bias.

3.2.2. Instrumental Variable Two-Stage Regression (IV-2SLS)

The endogeneity of trade liberalization can arise from several sources: first, endogeneity may occur due to omitted variables that bias the econometric estimates. Second, endogeneity may result from reverse causality between the liberalization of trade and the energy efficiency of firms. On the one hand, trade liberalization affects firms’ energy efficiency, while, on the other hand, firms with low energy efficiency may try to lobby interest groups and governments to seek trade protection, thus forcing governments to maintain higher tariff levels in relevant international tariff negotiations or to adopt some temporary protection measures. Moreover, the government has a clear political bias when it comes to setting tariffs. In terms of the reality of China’s tariff concessions, almost all of China’s industries have experienced tariff reductions of varying degrees and unequal proportions over the sample period and, especially, since China joined the World Trade Organization. There is a serious endogeneity problem associated with trade liberalization that can lead to biased and inconsistent estimates in terms of the benchmark regressions in this paper. To ensure the reliability of the estimation results, this paper adopts the two-stage least squares (2SLS) method to address the potential endogeneity problem. Constructing an appropriate instrumental variable is the key to this method. To construct instrumental variables for trade liberalization, this paper uses a two-step approach. In the first step, industry-level tariffs on final goods in 2001 are used as the dependent variable in a cross-sectional regression of industry-level variables, such as the rate of change in industry wages, the rate of change in sales, and the rate of growth in output over the period 1998–2000, to obtain the adjusted values of tariffs on final goods in 2001. In the second step, tariffs on final goods for subsequent years are projected according to the common rules on tariff reductions, in conjunction with the tariff levels committed to in China’s WTO accession agreement. Column (3) presents the findings of the two-stage least squares (2SLS) approach. It shows that the estimated coefficient on final tariffs, after accounting for endogeneity issues, is significantly negative at the 1% level, suggesting that trade liberalization significantly improves firms’ energy efficiency. Furthermore, for robustness purposes, we choose one lag in terms of trade liberalization and one lag in terms of all the control variables as instrumental variables. The findings also show that trade liberalization contributes significantly to firms’ energy efficiency. In regard to instrumental variable validity assurance, this paper uses Kleibergen and Paap’s method to perform various tests to select the instrumental variables [43]. The result shows that the original hypothesis of under identification of instrumental variables was rejected by the KP-LM statistic test and is significant at the 1% level.

3.3. Robustness Test

3.3.1. Changing the Measurement of Trade Liberalization

In the benchmark regression analysis above, tariffs on the final product are used to calculate the index of trade liberalization. For robustness purposes, this paper attempts to construct an indicator of trade liberalization using tariffs on intermediate goods. The regression results are presented in column (1) of

Table 5. Robustness test.

Variable	(1)	(2)	(3)	(4)	(5)
	Intermediate Goods Tariffs	OLS	FE	Single-Factor Energy Efficiency	DID
tariff	−0.2231 ***	−0.1185 ***	−0.0952 ***	−0.0522 ***
	(−7.18)	(−7.94)	(−6.42)	(−4.14)
triaff01 × post01					0.0232 ***
					(6.99)
loan	0.0064 ***	0.0193 ***	0.0611 ***	0.0006	0.0064 ***
	(4.30)	(3.52)	(11.23)	(0.15)	(4.28)
regulation	0.1800 ***	0.8282 ***	0.8712 ***	−0.0084	0.1801 ***
	(38.84)	(48.54)	(51.38)	(−0.62)	(38.86)
KL	0.0025 ***	−0.0280 ***	0.1857 ***	0.0010	0.0024 ***
	(3.37)	(−10.30)	(68.80)	(0.47)	(3.24)
profit	0.2772 ***	1.0760 ***	1.1461 ***	−0.0765 ***	0.2770 ***
	(67.55)	(71.19)	(76.32)	(−7.33)	(67.48)
state	−0.0058 **	−0.0355 ***	−0.0004	0.0269 ***	−0.0059 **
	(−2.07)	(−3.43)	(−0.04)	(3.54)	(−2.11)
cons	1.9423 ***	3.4119 ***	5.9510 ***	0.6858 ***	1.4584 ***
	(31.39)	(88.46)	(155.29)	(22.23)	(189.75)
N	33,628	33,628	33,628	33,628	33,628
j. R-sq	0.573	0.678	0.752	0.719	0.573

Note: ** and *** indicate significance at 5% and 1% levels, respectively. t statistic values are reported in parentheses.

, which shows that the regression coefficient for tariffs on intermediate goods is significantly negative at the 1% level. In other words, the previous conclusion that trade liberalization contributes to firms’ improved energy efficiency is not affected by replacing the measurement of the trade liberalization indicator.

3.3.2. Changing the Measurement of Energy Efficiency

In the benchmark regression, in our study, the LP method is used for the measurement of the all-factor energy efficiency (tfee) of enterprises and, to consider the robustness of the test results, this part of the study also includes the robustness test of the tfee measured using the OLS and FE method, respectively. The results are presented in columns (2) and (3) of Table 5 and the results show that the regression coefficients of the results of both the OLS and FE methods are significantly negative at the 1% level, which suggests that the all-factor energy efficiency obtained by different methods will not change the core conclusions. Additionally, the single-factor energy efficiency (total industrial output/coal consumption), also known as energy productivity, which measures the unit energy output, is measured in this paper. The regression results are presented in column (4) of Table 5 and the results show that the regression coefficient of the single-factor energy efficiency is significantly negative at the 1% level, indicating that trade liberalization plays an important role in the promotion of energy efficiency of enterprises. In other words, replacing the measurement of the energy efficiency indicator does not affect the previous conclusion that trade liberalization contributes to improving firms’ energy efficiency.

3.3.3. Double-Differential Distribution (DID)

A DID model is constructed to identify the impact of trade liberalization on firms’ energy efficiency in order to mitigate the endogeneity problem prevalent in the OLS regression in regard to benchmarking analysis:

$${tfee}_{,fit}=\beta ·∆{Tariff}_{2001,i}\times {post}_{2001}+\alpha X+{\delta }_t+{\theta }_f+{\vartheta }_r+{\beta }_i+{\delta }_{frit}$$

(6)

In Equation (6), f, i, and t denote the firm, industry, and year. $∆{Tariff}_{2001,i}$ denotes the decline in tariff rates at the industry level from 2001 to 2002, and its magnitude indicates the degree of trade liberalization. Moreover, ${post}_{2001}$ is a dummy variable for the year of China’s accession to the World Trade Organization; if ${post}_{2001}$ = 1, the year is greater than or equal to 2001, and if ${post}_{2001}$ = 0, the year is less than 2001. The core explanatory variable in the double-difference model is $∆{Tariff}_{2001,i}\times {post}_{2001}$. β reflects the effect of trade liberalization on the energy efficiency of firms; if β is significantly negative, it indicates that trade liberalization inhibits the improvement of the energy efficiency of firms, and vice versa, it suggests that trade liberalization promotes the improvement of the energy efficiency of firms. Column (5) presents the findings of the DID model, which shows that the β coefficient is significantly positive at the 1% level. This result shows that the conclusion that trade liberalization can significantly contribute to the improvement of the energy efficiency of firms does not vary depending on the estimation method.

3.4. Heterogeneity Analysis

In the analysis above, the effects of trade liberalization on firms’ energy efficiency have only been tested at the aggregate level and have not yet been tested and analyzed for different subsamples. Next, this paper will further test for differences in the effects of trade liberalization on the energy efficiency of a firm along four different dimensions: exporter or non-exporter, firm type, growth stage, and region.

3.4.1. Exporting or Non-Exporting Heterogeneity

According to the analysis by Xiang et al., there is a significant difference in productivity between exporting and non-exporting firms, which may lead to differences in the energy efficiency of firms [6]. Therefore, the total sample is divided into exporting and non-exporting firms to test the differing effects of liberalizing trade on firms’ energy efficiency. The finding suggests trade liberalization improves the energy efficiency of both exporting and non-exporting firms. A comparison of the absolute magnitude of their regression coefficients shows that trade liberalization promotes non-exporting firms’ energy efficiency more than exporting firms’ energy efficiency. This may be because exporters are more productive and energy efficient than non-exporters, and because exporters are more likely to use advanced foreign technologies to improve their energy efficiency. As a result, advanced exporters are less sensitive than non-exporters to technological improvements brought about by trade liberalization. Thus, trade liberalization promotes energy efficiency more in regard to non-exporters than exporters.

3.4.2. Firm Type Heterogeneity

Firms of a different nature with large differences in regard to business management, strategy, and quality will adjust their energy efficiency intensity to cope with differences in trade liberalization [44]. For this reason, this paper divides the total sample into a subsample of foreign firms and a subsample of domestic firms, according to the weight of each firm’s share of paid-in capital in terms of the total sample. Columns (2) and (4) of

Table 6. Heterogeneity analysis I.

Variable	(1)	(2)	(3)	(4)
	Exporting	Non-Exporting	Foreign	Domestic
tariff	−0.0232 ***	−0.0325 ***	−0.0278 ***	−0.0216 ***
	(−3.47)	(−5.54)	(−6.58)	(−0.80)
loan	0.0041	0.0085 ***	0.0069 ***	0.0014
	(1.50)	(4.50)	(4.52)	(0.15)
regulation	0.1624 ***	0.1942 ***	0.1814 ***	0.1718 ***
	(19.02)	(33.64)	(36.36)	(12.22)
KL	0.0065 ***	−0.0038 ***	0.0014 *	0.0097 ***
	(4.68)	(−3.91)	(1.74)	(3.29)
profit	0.3123 ***	0.2512 ***	0.2796 ***	0.1651 ***
	(41.64)	(49.46)	(64.48)	(11.55)
state	−0.0080 *	−0.0050	−0.0059 **	0.0001
	(−1.78)	(−1.30)	(−2.08)	(1.15)
cons	1.5683 ***	1.5797 ***	1.5592 ***	1.6061 ***
	(89.41)	(105.48)	(143.04)	(24.43)
N	10,813	22,815	15,661	17,967
j. R-sq	0.575	0.590	0.574	0.574

Note: *, **, and *** indicate significance at 10%, 5%, and 1% levels, respectively. t statistic values are reported in parentheses.

show the results of the estimation. The results show that the regression coefficients of trade liberalization pass the 1% significance test and are negative, indicating that the energy efficiency of both foreign and domestic firms will be enhanced by trade liberalization, but based on the size of the estimated coefficients, the contribution of trade liberalization to foreign firms’ energy efficiency is stronger than domestic firms. The reason is that the participation of domestic enterprises in the division of labor in the international market after trade liberalization enables them to have access to more advanced management experience and energy-saving and emission-reducing technologies from abroad. However, there is a problem in terms of optimal technology matching when enterprises use advanced management experience and energy-saving and emission-reducing technologies from abroad. In regard to domestic enterprises, due to their own mastery of technology and foreign technology, where there is a generation gap, they cannot fully absorb foreign advanced management experience and energy-saving emission-reduction technologies; while foreign enterprises themselves have a foreign nature, there is no technology-related generation gap, with foreign advanced management experience and energy-saving emission-reduction technologies benefiting from seamless connections. Therefore, foreign enterprises are more adept at absorbing foreign advanced management experience and energy-saving and emission-reduction technologies to reduce the amount of energy used by such enterprises, so the impact of trade liberalization on foreign enterprises can more effectively promote efficiency than is the case for domestic enterprises.

3.4.3. Growth Stage Heterogeneity

According to the enterprise life cycle theory, there are significant differences in the management and technology absorption of enterprises at different stages of development. The existence of these differences will inevitably affect the way enterprises respond to trade liberalization. In this paper, the stage of development of a firm is measured by whether the firm has survived more than the median number of years in the sample under study. If the firm’s survival is less than the median number of years, it is defined as a development stage firm, otherwise it is defined as a growth stage firm or a declining stage firm. Accordingly, the total sample was split into growing and non-growing firms, and the regressions were run in regard to the subsamples, based on the baseline model. According to the results of the econometric regressions, the coefficient of trade liberalization is significantly negative, indicating that trade liberalization plays a significant role in promoting both growing and non-growing firms, but based on the size of the estimated coefficients, trade liberalization promotes energy efficiency in growing firms more than non-growing firms. Possible explanations are that companies in the growth phase are more motivated to use energy-efficient equipment and processes in response to competitive market pressures and also have a stronger sense of energy conservation. The non-growth stage of an enterprise involves firms taking advantage of their existing brand value and scale advantages, even if the production capacity is relatively backward in the short term excessive profits can be obtained, which results in a lack of incentives to improve energy efficiency. In addition, mature or declining companies often face a variety of problems, such as aging production equipment and innovation inertia, which will inhibit their ability to improve their energy efficiency to some extent.

3.4.4. Region Heterogeneity

Regional differences are an important source of heterogeneity in regard to energy efficiency among enterprises, as China is a huge country with natural resources, regional institutional environments, and levels of development, which vary significantly. Based on the level of economic development, the sample is divided into eastern and midwestern regions, and the regressions are conducted on the subsamples, based on the baseline model. Columns (3) and (4) of

Table 7. Heterogeneity analysis II.

Variable	(1)	(2)	(3)	(4)
	Growing	Non-Growing	Eastern	Midwestern
tariff	−0.0239 ***	−0.0200 **	−0.0279 ***	−0.0275 ***
	(−4.68)	(−2.55)	(−4.78)	(−4.75)
loan	0.0202 ***	−0.0145 *	0.0054 **	0.0069 ***
	(6.45)	(−1.89)	(2.31)	(3.53)
regulation	0.1791 ***	0.1848 ***	0.1736 ***	0.1435 ***
	(33.93)	(17.17)	(22.43)	(23.50)
KL	0.0012	0.0014	0.0025 **	−0.0024 **
	(1.38)	(0.84)	(2.07)	(−2.46)
profit	0.2975 ***	0.2436 ***	0.2318 ***	0.3224 ***
	(51.41)	(38.20)	(39.18)	(55.41)
state	−0.0043	−0.0074	−0.0049	−0.0047
	(−1.13)	(−1.57)	(−1.17)	(−1.21)
cons	1.5568 ***	1.5673 ***	1.5483 ***	1.5977 ***
	(112.30)	(48.17)	(97.95)	(110.45)
j. R-sq	20,453	13,175	23,761	9867
	0.558	0.601	0.575	0.571

Note: *, **, and *** indicate significance at 10%, 5%, and 1% levels, respectively. t statistic values are reported in parentheses.

show the results and the findings suggest that the regression coefficients of trade liberalization both pass the 1% significance test and are negative, implying that a company’s energy efficiency is greatly enhanced by trade liberalization, in both the eastern and midwestern regions. However, from the point of view of the magnitude of the estimated coefficients, trade liberalization has a stronger contributing effect on firms in the eastern rather than the midwestern region. Trade costs are lower in the east and equal tariff cuts have a greater impact on eastern firms; coupled with stricter eastern environmental regulations, the interaction suggests that the inhibitory effect of trade liberalization is greater in the eastern region than the Midwestern region.

4. Mechanism Tests

4.1. Mechanism Model Design

From the above analysis, it can be seen that trade liberalization contributes to an important improvement in a firm’s energy efficiency, but an in-depth analysis of the intrinsic mechanism by which trade freedom affects a firm’s energy efficiency has not yet been carried out. To understand the relationship between corporate energy efficiency and trade liberalization, we analyze this underlying mechanism. For this reason, this part of the paper will construct a mediated effects model to test exactly which pathway of a firm’s energy efficiency is affected by trade liberalization. In the second part of this paper, according to analyzing the influencing mechanism, this paper selects firm size (${size}_{fit}$) and total factor productivity (${tfp}_{fit}$) as mediator variables. To empirically test the mechanism of action, in this paper, the following model of the mediation effect is constructed, based on the existing literature:

$${tfee}_{,fit}={\alpha }_0+{\alpha }_1{tariff}_{it}+{\gamma X}_{fit}+{\beta }_r+{\beta }_f+{\beta }_t+{\epsilon }_{fit}$$

(7)

$${size}_{fit}=b_0+b_1{tariff}_{it}+{\gamma X}_{fit}+{\beta }_r+{\beta }_f+{\beta }_t+{\epsilon }_{fit}$$

(8)

$${tfp}_{fit}=c_0+c_1{tariff}_{it}+{\gamma X}_{fit}+{\beta }_r+{\beta }_f+{\beta }_t+{\epsilon }_{fit}$$

(9)

$${tfee}_{,fit}=d_0+d_1{tariff}_{it}+d_2{size}_{fit}+d_3{tfp}_{fit}+{\gamma X}_{fit}+{\beta }_r+{\beta }_f+{\beta }_t+{\epsilon }_{fit}$$

(10)

In Equations (7)–(10), f denotes firms, i denotes three-quarter industries, and t denotes years. ${Size}_{fit}$ denotes the size of the firm, using the natural logarithm of the firm’s sales +1; ${tfp}_{fit}$ denotes the all-factor productivity of the firm and, here in this paper, we use the Olley–Pakes method (OP) to estimate the all-factor productivity of the firm.

4.2. Mechanism Test Results Analysis

The findings on the mechanism of trade liberalization on firms’ energy efficiency are reported in

Table 8. Mechanism tests.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
	$\mathit{t}\mathit{f}\mathit{e}\mathit{e}$	scale	tfp	$\mathit{t}\mathit{f}\mathit{e}\mathit{e}$	$\mathit{t}\mathit{f}\mathit{e}\mathit{e}$	$\mathit{t}\mathit{f}\mathit{e}\mathit{e}$
tariff	−0.0283 ***	−0.0070 ***	−0.0155 ***	−0.0111 ***	−0.0274 ***	−0.0108 ***
	(−7.02)	(−5.25)	(−5.28)	(−4.69)	(−6.77)	(−4.57)
tfp				0.0198 ***		0.0118 ***
				(6.49)		(6.41)
scale					0.1526 ***	0.0449 ***
					(21.98)	(11.03)
loan	0.0064 ***	−0.0099 ***	0.0109 ***	−0.0058 ***	0.0079 ***	−0.0053 ***
	(4.30)	(−20.18)	(10.08)	(−6.63)	(5.31)	(−6.10)
regulation	0.1799 ***	0.1521 ***	0.1342 ***	0.0296 ***	0.1567 ***	0.0229 ***
	(38.84)	(99.61)	(40.00)	(10.88)	(33.02)	(8.22)
KL	0.0025 ***	0.0030 ***	0.0002	0.0023 ***	0.0020 ***	0.0022 ***
	(3.39)	(12.42)	(0.29)	(5.39)	(2.77)	(5.07)
profit	0.2773 ***	−0.0340 ***	0.1884 ***	0.0664 ***	0.2825 ***	0.0681 ***
	(67.57)	(−25.15)	(63.37)	(27.36)	(68.81)	(28.01)
state	−0.0058 **	−0.0050 ***	0.0014	−0.0074 ***	−0.0051 *	−0.0071 ***
	(−2.07)	(−5.38)	(0.68)	(−4.47)	(−1.80)	(−4.34)
cons	1.5648 ***	0.0580 ***	1.9052 ***	−0.5687 ***	1.5559 ***	−0.5696 ***
	(149.40)	(16.81)	(251.17)	(−80.45)	(148.64)	(−80.59)
N	33,628	33,628	33,628	33,628	33,628	33,628
j. R-sq	0.573	0.578	0.704	0.854	0.574	0.854

Note: *, **, and *** indicate significance at 10%, 5%, and 1% levels, respectively. t statistic values are reported in parentheses.

. The estimation results with the mediating variable firm size as the explanatory variable are reported in column (2). The estimated coefficient on trade liberalization is found to be negative and passes the 1% significance test, implying that trade liberalization contributes to firm size. The estimation results with total factor productivity as the explanatory variable are reported in column (3). The finding shows that the coefficient of trade liberalization is significantly negative at the 1% level, indicating that trade liberalization has an important contribution to firms’ all-factor productivity. One possible explanation is that lower import tariffs lead to the entry of similar products from abroad into the domestic market, which affects domestic products and leads to increased competition in the domestic market. Domestic firms have to invest more and replace obsolete equipment in order to cope with fierce market competition, leading to higher productivity. Columns (4)–(6) report the results of the econometric tests on energy efficiency on the mediating and basic control variables. It is found that the all-factor productivity is significantly positive at the 1% level, implying the improvement in firms’ all-factor productivity has a significant contribution on firms’ energy efficiency. The regression coefficient of the mediating variable firm size is significantly positive at the 1% level, suggesting that increasing the firm size contributes to improving a firm’s energy efficiency. In addition, this is noteworthy compared to the findings in terms of the baseline estimation in column (1). Adding columns (4) and (5) for the all-factor productivity and firm size, respectively, the regression coefficients of trade liberalization show some decrease in both their significance level (t-value) and absolute value. After further estimations, by adding the all-factor productivity and firm size in column (6), compared to column (1), the significance level (t-value) and the absolute coefficient value for trade liberalization continue to decrease. This shows that trade liberalization contributes to corporate energy efficiency mainly through two channels: increasing the TFP and the size of firms. For rigor, this paper also relies on the Sobel test to test the significance of the product term of the regression coefficients on the paths of the two mediating variables, separately [45]. That test is as follows: H₀: $b_1{d}_2$ = 0 and $c_1{d}_3$ = 0. If H₀ is rejected, it means that the mediation effect passes the test of significance. Calculate the standard deviation of the two product terms $b_1{d}_2$ and $c_1{d}_3$, following the Sobel test [45].$S_{b_1{d}_2}=\sqrt{b_1^2{S}_{d_2}^2+d_2^2{S}_{b_1}^2}$, $S_{c_1{d}_3}=\sqrt{c_1^2{S}_{d_3}^2+d_3^2{S}_{c_1}^2}$, where S is the standard deviation of the respective coefficient estimates. Based on the regression findings in Table 8, this paper can easily estimate the standard deviation of 0.00017 and 0.00182 for $b_1{d}_2$ and $c_1{d}_3$, respectively. Furthermore, in this paper, $Z_{b_1{d}_2}$ = −1.52 and $Z_{c_1{d}_3}$ = −7.26 were calculated based on the equations $Z_{b_1{d}_2}=b_1{d}_2/S_{b_1{d}_2}$ and $Z_{c_1{d}_3}=c_1{d}_3/S_{c_1{d}_3}$, both of which pass the significance test. This indicates that the scale effects and technology effects are the two channels, which relate to corporate energy efficiency affected by trade liberalization, thereby increasing the overall level of social welfare.

5. Further Discussion: A Perspective on Resource Reallocation Effects

The previous analysis examined the mechanism through which trade liberalization impacts firms’ energy efficiency, based on micro-firm level data. It was found that trade liberalization is a contributor to all-factor productivity growth and the expansion of the firm size, which create conditions for firms to improve their energy efficiency. However, the previous analysis ignored the impact of resource reallocation effects on energy efficiency. In general, trade liberalization also brings about a structural change in the economy, shifting various production inputs from energy-intensive industries to less energy-intensive industries and, thus, affecting energy efficiency at the industry level. This section examines whether trade liberalization affects energy efficiency at the industry level, by changing the structure of the economy. Based on the methodology by Melitz and Polance, a dynamic decomposition of energy efficiency at the industry level is performed, which is divided into the following four components [22]: the $\mathrm{i}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{a}$-firm effect, which represents the change in total energy efficiency caused by changes in the energy efficiency level of the surviving firms themselves, holding the market share of the surviving firms constant. The inter-firm effect represents the change in total energy efficiency caused by changes in the market share of the surviving firms, holding the energy efficiency of the surviving firms constant. The firm entry effect represents the change in industry energy efficiency due to new firms entering the market. The firm exit effect represents the change in industry energy efficiency due to firms exiting the market. Firm entry and firm exit are the narrow resource reallocation effects. The broad resource reallocation effect consists of the narrow resource reallocation effect and the inter-firm effect.

The formula for measuring energy efficiency at the industry level is as follows:

$${TFEE}_{it}={\sum }_{f\in {\varnothing }_i}{av}_{ft}{tfee}_{ft}$$

(11)

where ${\varnothing }_i$ denotes the set of firms in industry i, ${av}_{ft}$ denotes the ratio of the enterprises’ total industrial output value to the industry’s total industrial output value, and ${tfee}_{it}$ denotes the energy efficiency of firm f in period t. The change (∆) in the industry aggregate ${TFEE}_{it}$ between the two periods can be expressed in Equation (12) as:

$$∆{TFEE}_{it}={\sum }_{f\in (S,EN)}{av}_{ft}{tfee}_{ft}-{\sum }_{f\in (S,EX)}{av}_{f,t-1}{tfee}_{f,t-1}$$

(12)

where S represents the set of surviving firms, EN denotes the set of firms entering the market, and EX denotes the set of firms exiting the market. The dynamic decomposition formula for energy efficiency at the industry level is shown below:

$$\begin{gathered} ∆TF{EE}_{it}=\underset{\mathrm{i}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{a}\text{-}\mathrm{f}\mathrm{i}\mathrm{r}\mathrm{m} \mathrm{e}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{c}\mathrm{t}}{\underbrace{\sum _{f\in S}\widetilde{{av}_f}∆{tfee}_{ft}}}+\underset{\mathrm{i}\mathrm{n}\mathrm{t}\mathrm{e}\mathrm{r}\text{-}\mathrm{f}\mathrm{i}\mathrm{r}\mathrm{m} \mathrm{e}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{c}\mathrm{t}}{\underbrace{\sum _{f\in S}∆{av}_{ft}\left(\widetilde{{tfee}_f}-\widetilde{{TFEE}_i}\right)}} \\ +\underset{\mathrm{f}\mathrm{i}\mathrm{r}\mathrm{m} \mathrm{e}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{y} \mathrm{e}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{c}\mathrm{t}}{\underbrace{\sum _{f\in EN}{av}_{ft}\left({tfee}_{ft}-\widetilde{TF{EE}_i}\right)}}-\underset{\mathrm{f}\mathrm{i}\mathrm{r}\mathrm{m} \mathrm{e}\mathrm{x}\mathrm{i}\mathrm{t} \mathrm{e}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{c}\mathrm{t}}{\underbrace{\sum _{f\in EX}{av}_{f,t-1}\left(tf{ee}_{f,t-1}-\widetilde{{TFEE}_i}\right)}} \end{gathered}$$

(13)

where the wavy line on the variable indicates the adjacent two-period mean: $\widetilde{{av}_f}=\left({av}_{f,t-1}+{av}_{ft}\right)/2$; $\widetilde{{tfee}_f}=\left({tfee}_{f,t-1}+tf{ee}_{ft}\right)/2$; $\widetilde{{TFEE}_i}=\left({TFEE}_{i,t-1}+{TFEE}_{it}\right)/2$.

Increasing the energy efficiency of surviving firms, the intra-firm effect is greater than zero. When firms’ energy efficiency is greater (less) than the mean, their share of the total industrial output increases (decreases), and the between-firm effect is greater (less) than zero. If the exiting company’s energy efficiency is below average, the exit effect is greater than zero.

Table 9. Results of the decomposition of sectoral energy efficiency changes.

Total Change (1)	Intra-Firm Effect (2)	Inter-Firm Effect (3)	Firm Entry Effect (4)	Firm Exit Effect (5)	Entry and Exit (6) = (4) + (5)	Resource Reallocation Effect (7) = (3) + (6)
0.0431	0.0129	0.0051	−0.0152	0.0403	0.0251	0.0302
	29.93	11.83	−35.27	93.50	58.24	70.07

shows the results of the decomposition of the average energy efficiency at each industry level during the sample visit. According to Table 9, the average increase in energy efficiency for each sector is 0.0431. The resource reallocation effect of the exit effect on firms’ energy efficiency is the largest at 93.50%, implying that firms exiting the market each year are less energy efficient than the industry average. One possible explanation is that less productive firms are unable to withstand the pressures of a highly competitive market and withdraw from it. The intra-firm effect contributes 29.93% to the improvement of energy efficiency in the industry, implying that the competitiveness of the surviving incumbents has improved. The inter-firm effect contributes only 11.83% to the improvement in energy efficiency in the industry, indicating that the surviving companies’ market share increased, but only to a limited extent, with new entrants taking up the market share released by exiting firms. The firm entry effect’s negative resource reallocation effect means the energy efficiency of new entrants is lower than the industry average, which lowers the overall energy efficiency of the industry. This is probably because new entrants have to pay some costs to expand their markets. There are not enough funds within new entrants for cleaner technological innovations, which in turn result in lower energy efficiency than surviving firms. The reallocation effect, strictly speaking, accounts for 58.24 percent of the industry’s energy efficiency improvement. Finally, 70.07 percent of the industry’s energy efficiency improvement comes from the resource reallocation effect in the broad sense. In summary, this paper concludes that an important driver of energy efficiency improvements in industry is the resource reallocation effect

Based on the above dynamic decomposition results, this part of the study builds an econometric model to further rigorously test whether the effect of resource reallocation is an effective route for trade liberalization to improve energy efficiency. Equation (14) represents the econometric model:

$$\varnothing {EE}_{it}={\alpha }_0+{\alpha }_1{tariff}_{it}+\gamma X_{it}+{\beta }_i+{\beta }_t+{\epsilon }_{it}$$

(14)

where $\varnothing {EE}_{it}$ denotes the change in energy efficiency in industry i due to resource reallocation effects; $X_{it}$ is an industry-level control variable obtained by adding up the firm-level variables in a weighted manner.

Table 10. Econometric results on the impact of trade liberalization and sectoral changes on energy efficiency.

Variable	(1)	(1)	(2)	(3)
	Intra-Firm Effects	Intra-Firm Effects	Entry–Exit Effects	Resource Reallocation Effect
tariff	−0.1321 ***	−0.1588 ***	−0.1715 ***	−0.2053 ***
	(−5.05)	(−5.21)	(−5.25)	(−6.25)
scale	0.0041 ***	0.0051 ***	0.0056 ***	0.0056 ***
	(2.13)	(2.31)	(2.34)	(2.43)
tfp	0.0742 ***	0.1001 ***	0.1066 ***	0.1245 ***
	(2.70)	(2.81)	(2.86)	(4.08)
loan	−0.1185 ***	−0.1221 ***	−0.1311 ***	−0.1346 ***
	(−4.54)	(−4.42)	(−4.41)	(−4.70)
regulation	0.0164 ***	0.0218 ***	0.0225 ***	0.0221 ***
	(6.43)	(6.58)	(6.51)	(6.56)
KL	−0.0112 **	−0.0112 **	−0.0111 **	−0.0114 **
	(−1.37)	(−1.26)	(−1.11)	(−1.21)
profit	0.0702 **	0.0851 **	0.1041 **	0.1108 ***
	(1.20)	(1.31)	(1.38)	(1.57)
cons	1.3163 ***	1.2753 ***	1.3722 ***	1.2612 ***
	(11.45)	(17.31)	(17.43)	(16.37)
N	3726	3726	3726	3726
j. R-sq	0.161	0.171	0.170	0.175

Note: ** and *** indicate significance at 5% and 1% levels, respectively. t statistic values are reported in parentheses.

presents the results of the regression involving Equation (14). The estimated coefficient on trade liberalization is significantly negative at the 1% level, suggesting the resource reallocation effect is a possible way for trade liberalization to improve energy efficiency, as shown by the estimation results in column (4). One possible explanation is that trade liberalization leads to a shift in the structure of the economy, allowing various inputs to be shifted from inefficient to efficient industries. To provide a comprehensive picture of the impact of trade liberalization on changes in energy efficiency due to different components of the resource reallocation effect, in this paper, trade liberalization is econometrically tested for inter-firm effects, intra-firm effects, and entry–exit effects. From the estimation results, is the following is known: the regression coefficient for trade liberalization in column (1) is significantly negative at the 1% level, suggesting the inter-firm effect is an effective method according to which trade liberalization improves energy efficiency in the industry. This is due to the fact that surviving firms have increased their competitiveness and market share through learning by doing. The estimated coefficient on trade liberalization in column (2) passes the 1% significance test and is negative, implying the intra-firm effect is an effective method for trade liberalization to improve energy efficiency in the industry. The reason may be that the productivity of the surviving firms increases through the accumulation of experience, which contributes to the improvement of their own energy efficiency. The estimated coefficient on trade liberalization in column (3) passes the 1% significance test and is negative, implying the entry–exit effect is an effective method for trade liberalization to increase the amount of energy efficiency in the industry. Combined with the results in Table 9, the findings imply that trade liberalization causes weakly competitive firms to exit the market, freeing up market share for highly competitive firms and, thus, contributing to the improvement of energy efficiency in the industry.

6. Discussion and Implications

In the following important areas, this study makes marginal contributions. First, unlike previous studies that did not involve the testing of impact mechanisms, this paper examines how trade deregulation impacts corporate energy efficiency by constructing a structural equation model, showing that two ways to enhance the energy efficiency of enterprises are the technological and scale effects brought about by trade liberalization. Second, to quantify the contribution of structural adjustments due to trade liberalization to energy efficiency improvements in industry, energy efficiency is disaggregated at the industry level, using the dynamic disaggregation method by Melitz and Polance [22]. Disaggregation shows that 70.07 percent of the energy efficiency gain at the industry level is due to resource reallocation effects. Finally, most of the existing studies use the pollution database on Chinese industrial enterprises, which was compiled a long time ago and may not be a realistic representation of the current situation. This study uses data on listed industrial enterprises in China from 2000 to 2022 to overcome the problem related to the long sample period in the relevant literature, so that the conclusions in this study are more realistic and instructive.

In addition, future research may be enhanced as follows, because the findings presented in this study have some limitations. First, the effect of trade liberalization on the energy efficiency of companies is discussed in this study, taking into account the significant reduction of tariffs after China’s accession to the WTO. However, this study does not consider the impact of other external factors and policies after China’s WTO accession, such as China’s exchange rate reform and the impact of the 2008 financial crisis. The impact of exogenous factors may have far-reaching implications for the effects of trade liberalization, which is a direction for further research in the future. Second, this study used the LP method developed by Levinsohn and Petrin to measure the all-factor energy efficiency of enterprises [46]. However, there are still many shortcomings in regard to the abovementioned methods and better methods for measuring the all-factor energy efficiency of companies could be considered in future studies. Finally, these three channels, in addition to the technology effect, the scale effect, and the resource reallocation effect, influence the all-factor energy efficiency of firms. Future research could also consider other channels of influence external to the firm, such as competition effects and government regulation.

The findings presented in this paper are important to consider: the results of this paper show that trade liberalization is a driver of energy efficiency improvements in firms, suggesting that policies designed to help liberalize trade through the reduction of non-tariff barriers and tariff reductions not only promote the development of China’s foreign trade and economic growth, but also improve the welfare level of Chinese society by improving the energy efficiency of enterprises. China’s rapid economic growth has been driven by the rapid development of foreign trade. As a result, there is an increasing conflict between energy supply and demand, which has become a key constraint on sustainable economic development in China. The key to transforming the flawed development model of high-energy consumption, low output is finding a way to effectively promote economic development, while reducing energy consumption, per unit of GDP. By using the one-size-fits-all approach to shut down high-energy-consuming, high-emission, low-productivity enterprises on a large scale, which can certainly improve the energy effect in the short term, but, in the long-term, based on the development level, will ultimately drag down economic growth, which in turn will undermine the well-being of society as a whole. The conclusions in this paper provide new ideas for the current efforts directed at energy conservation and emissions reduction. That is, the use of the direct closure of high-energy, high-emitting, low-productivity firms will not result in energy-saving and emission-reduction goals that are part of a long-term strategy. Expanding the perspective to become outward looking, the process of trade liberalization can be further promoted by government agencies, to achieve energy savings and emission reductions that are conducive to economic growth. By expanding the opening up of markets to promote economic growth, this process can save energy and reduce emissions. Ultimately, such strategies will realize win–win developments in terms of the economy and energy conservation and emissions reduction.