The Energy-Saving Effect of Tax Rebates: The Impact of Tax Refunds on Corporate Total Factor Energy Productivity

Abstract

This paper investigates whether and how tax and fee support policies at the firm level in China influence the total factor energy productivity of enterprises. Using panel data from Chinese public trading companies for the period 2004–2020, this study employs a panel model for estimation. The findings suggest that tax rebates contribute to the improvement of the total factor energy productivity of enterprises. Specifically, a 1% increase in tax refunds leads to a growth of approximately 0.008% in total factor energy productivity. Robustness tests and endogeneity checks confirm the validity of the results. Heterogeneity analysis reveals that tax rebates have a significant impact on state-owned enterprises, small- and medium-sized enterprises, and non-technology firms in terms of enhancing their total factor energy productivity. Mechanism analysis indicates that tax rebates facilitate firms in alleviating financing constraints and enhancing their innovation capabilities, thereby improving energy efficiency. The research findings of this paper provide empirical support for optimizing policy supply, improving energy usage efficiency, and promoting the development of a globally sustainable economy.

1. Introduction

Industrialization, particularly in emerging economies, has played a critical role in global economic development [1,2], but it has also led to severe environmental pollution, which has gained widespread attention [3,4,5,6,7]. To reduce the damage caused by pollution, one of the key measures is to increase energy efficiency [8,9,10,11]. China, being the largest producer of industrial goods globally, faces substantial challenges concerning environmental pollution [12,13]. For instance, it has the highest level of carbon dioxide emissions globally, and its energy consumption accounts for a significant proportion of the world’s total consumption [14,15]. In response, the Chinese government has implemented crucial measures aimed at improving the environment, including tax reduction policies designed to encourage enterprises to optimize energy use and reduce consumption [16,17]. However, the enterprises’ drive for economic interests often makes them reluctant to optimize their energy usage, leading to additional production costs [18,19]. Therefore, studying the relationship between tax incentives and enterprise energy efficiency is vital for evaluating the practical effects of tax reduction policies and promoting global sustainable development.

Numerous studies have investigated the effect of tax incentives on energy consumption [14,20], with some specifically examining the impact of tax policies promoting the use of new energy sources. For instance, Yang and Tang [21] discovered that subsidy programs for energy-efficient vehicles and privately purchased new energy vehicles contributed to the adoption of energy-efficient and new energy vehicles, resulting in improved fleet fuel efficiency. However, it also led to increased petroleum consumption and carbon dioxide emissions. Hájek, et al. [22] explored the influence of carbon taxes on emissions’ reduction in the energy industry and found that increasing tax rates can effectively reduce greenhouse gas emissions. They observed that each euro increase in carbon taxes corresponded to a reduction of 11.58 kg of per capita annual emissions. Jiménez-Gómez and Acevedo-Prins [23] evaluated tax incentives for wind power plant investments in Colombia and noted that such measures promote investment in wind power plants while diversifying energy consumption through unconventional renewable sources. Sun, Zhan, and Du [18] compared value-added tax preferences for different types of new energy companies and empirically studied their impact on listed new energy companies using the difference-in-differences (DIDs) method. They discovered that value-added tax refunds in the new energy industry negatively affect corporate financial performance. Shahbaz, et al. [24] found that financial development leads to an increased demand for renewable energy, suggesting that governments should implement incentive and tax policies to stimulate business demand for renewable energy.

Although existing research has explored the impact of tax policies on emissions’ reduction from various perspectives, certain limitations remain. Firstly, the current literature primarily focuses on the effects of individual tax incentive policies on energy consumption [18,21,25], neglecting the potential impact of policy combinations on energy use. Secondly, while existing studies analyze the influence of tax policies on the overall quantity and structure of energy consumption, they overlook the assessment of energy consumption efficiency [26,27].

Given the insufficient existing literature, this study investigates the impact of tax refunds on enterprise total factor energy productivity using a sample of Chinese public trading companies from 2004 to 2020. Specifically, this paper employs non-parametric estimation to assess enterprise total factor energy productivity and constructs panel models to empirically test the relationship between tax refunds and such productivity. To address potential endogeneity, instrumental variable and Heckman models are employed, along with substitution variables and additional fixed effects for estimation robustness. The study also explores the heterogeneity of this impact from the perspectives of enterprise and industry characteristics, thereby enhancing the applicability of the estimated results. Finally, the study discusses the internal mechanisms through which tax refunds influence enterprise total factor energy productivity, focusing on the promotion of technological innovation and the alleviation of financing constraints.

This study makes three contributions. First, it comprehensively examines the relationship between tax incentives and enterprise energy efficiency, addressing the limitations of previous studies that focused on individual tax policies such as vehicle subsidy programs [21] and value-added tax incentives [18]. By considering the combined effects of different tax incentives, this study analyzes their impact on enterprise total factor energy productivity. Moreover, it provides a comprehensive analysis of the environmental governance effects of tax incentives, expanding the existing literature in this area. Second, this study employs a novel analytical approach to explore energy efficiency. While existing research primarily focuses on the ecological benefits of energy consumption and consumption structure [8,24,28,29,30], this paper adopts non-parametric estimation methods to analyze the economic and environmental effects of energy usage from the perspective of factor productivity. This approach offers a new avenue for future research on energy efficiency. Third, this study contributes to a deeper understanding of the impact of tax incentives on energy efficiency. While previous research has examined the influence of tax incentives, it often lacks an in-depth analysis of the underlying mechanisms driving this impact [30,31,32]. This paper addresses this gap by investigating how tax incentives promote enterprise innovation and alleviate financing constraints. By verifying the internal mechanisms through which tax incentives affect energy efficiency, it enhances our understanding of the role tax incentives play.

The subsequent sections of this paper are organized as follows: Section 2 presents the research hypotheses and analyzes the internal mechanisms through which tax refunds affect enterprise total factor energy productivity. Section 3 describes the empirical design, including econometric specifications, variable measurements, descriptive statistics of key variables, and correlation analysis. Section 4 presents the empirical results, encompassing baseline regressions, robustness tests, heterogeneity analysis, and mechanism analysis. Finally, Section 5 concludes the study with policy implications.

2. Hypotheses Development

Tax refunds serve as a policy tool widely employed to stimulate economic development, attract investments, and enhance corporate competitiveness [33,34,35]. However, their impact on energy efficiency remains uncertain. By reviewing relevant literature, this study suggests that tax refunds might influence enterprise total factor energy productivity by alleviating financing constraints and promoting enterprise innovation.

2.1. Innovation Effects

Support for innovation is crucial in enhancing the overall energy productivity of enterprises [36,37]. However, due to the externalities associated with innovation, the benefits of research and development (R&D) activities are not fully captured by the entities conducting them, leading to actual R&D investment levels in enterprises falling well below the optimal levels [38,39]. This “market failure” phenomenon, resulting from the positive externalities of R&D activities, often leads to insufficient investment by enterprises [40]. To address this, tax refunds can serve as a means of compensating for the positive externalities associated with enterprise R&D activities, thereby incentivizing enterprises to improve their overall energy productivity [25,41].

First, tax refunds can boost enterprise profits, provide a secure environment for technological innovation, and incentivize enterprises to enhance their overall energy efficiency. A bulk of literature demonstrates that tax intervention policies effectively compensate for the positive externalities resulting from enterprise R&D investments, leading to Pareto improvements [18,20,23,42]. In contrast to indirect incentives such as tax reductions, direct government subsidies can also encourage enterprises to increase their R&D investments by raising marginal income following successful R&D or reducing marginal costs after failed attempts. Yang and Tang [21] revealed that government subsidies offer financial support for enterprise innovation activities, leading to increased R&D investment, risk sharing, and heightened innovation enthusiasm, ultimately enhancing enterprise innovation capabilities.

Second, tax refunds can attract high-quality talent to enterprises and safeguard their innovative efforts aimed at improving energy efficiency. Innovation activities require individuals with substantial knowledge and expertise, and tax refunds can partially alleviate the issue of human capital mismatch. Through German tax policies, Fuest, et al. [43] discovered that enterprises pass on 51% of their tax costs to employees. Thus, when corporate income tax rates decrease, employee wages experience significant increases, thereby attracting more high-quality talent to engage in R&D activities within enterprises.

Finally, tax refunds related to environmental protection convey a crucial message indicating the government’s commitment to green enterprise development. This signal helps attract substantial financial resources that can be utilized to improve energy efficiency in enterprises. As industries expand, competition intensifies within the sector. In order to secure a space for survival and growth amidst fierce competition, enterprises must enhance their competitive edge, pursue innovative R&D efforts, generate excess profits, and further improve their overall energy productivity by fostering innovation capabilities.

2.2. Alleviating Financing Constraints

Tax refunds can alleviate the financing constraints faced by companies. Research conducted by Fox [44] suggests that high tax burdens reduce the internal cash flows of firms, making it challenging for them to secure funds for refinancing [45]. Heider and Ljungqvist [46] also found that excessive tax rates lead firms to rely on debt financing. However, the obligations of regular debt repayment and interest payments are not conducive to high-risk innovation activities undertaken by companies [47]. Specifically, tax reductions can help meet the financing needs of companies by alleviating their financial constraints. Numerous empirical studies have consistently shown that tax reductions indeed stimulate firms to increase their investment in research and development (R&D), as well as the output of new products [42]. Tax refunds provide micro-market entities with a boost to their cash flow through pre-incentives, and the government can specify and guide the use of these funds, further easing the financing constraints faced by companies.

Furthermore, tax refunds can steer the investment of social capital, thus encouraging firms to enhance their overall energy productivity by attracting external funding. Tax refunds directed towards environmental protection initiatives send a crucial signal that the government places importance on the green development of companies. This, in turn, attracts a significant inflow of financial resources into their production and operations. With the infusion of external investments, firms can improve their allocation efficiency and technological efficiency through various means, such as technical support, standardized management practices, and knowledge transfer. Additionally, they can increase their accumulation of production factors, leading to improved energy utilization efficiency. Ljungqvist, et al. [48], as well as Lim [49], argue that lower corporate income tax rates reduce the expected tax burden on corporate risk return, which serves as a risk compensation mechanism and promotes capital inflows. Hanson and Rohlin [50] discovered that federal wage tax exemptions resulted in the attraction of approximately 2.2 new firms for every 1000 existing firms in federal empowerment zones, amounting to a total of 20 new firms in those areas. Empirical studies by Liu and Mao [51] and Yang, et al. [52] have tested the effectiveness of various tax incentives implemented by the Chinese government to attract foreign direct investment. Building on the aforementioned analysis, this study proposes the following research hypothesis:

Hypothesis 1: 

Tax refunds incentivize firms to enhance their overall energy productivity.

Figure 1 depicts the analytical framework of our research hypothesis, where tax refunds serve two purposes. Firstly, they encourage firms to increase their investment in research and development, thereby improving their innovation capabilities. Secondly, tax refunds ease the financing constraints faced by companies in enhancing their overall energy productivity, ultimately resulting in improved energy utilization efficiency.

3. Research Design

3.1. Model Specification

Following Shi, et al. [53], this study constructs specification (1) to examine the impact of tax refunds on firms’ total factor energy productivity.

$$EFF{I}_{it}=\alpha TAXRE{F}_{it}+\beta CONTROLS+\mu +\gamma +{\epsilon }_{it}$$

(1)

where i and t represent firm and year, respectively. EFFI denotes firms’ energy green total factor productivity, TAXREF represents the amount of tax refunds, and α captures the response of firms’ energy green total factor productivity to tax refunds, which is the main focus of this study. CONTROLS represents a series of control variables. β represents the estimated set of fixed variables. μ represents industry fixed effects, γ represents year fixed effects, and ε denotes the random error term.

3.2. Variable Specification

3.2.1. Corporate Total Factor Energy Productivity

This study uses OP, LP, and GMM methods to calculate firms’ total factor energy productivity. In essence, OP, LP, and GMM are non-parametric methods used to calculate total factor energy productivity. They involve controlling for the effects of labor, capital, and intermediate inputs. Specifically, the OP method uses investment as a proxy for unobserved productivity and solves the endogeneity issue between production factors and productivity [54]. Due to the partial lack of investment, the LP method is proposed by Levinsohn and Petrin [55] based on the OP method, with intermediate inputs as a proxy for productivity. Ackerberg, et al. [56] suggested using the GMM method to measure firms’ total factor energy productivity to mitigate identification and endogeneity issues in OP and LP methods.

The LP method’s estimation results are used as benchmark regression, and OP and GMM methods’ results are employed for robustness checks. This is because the micro-level data from listed firms exclude ST companies, resulting in few firm exits, addressing the LP method’s sample exit problem. Furthermore, including OP and GMM methods in robustness checks better addresses the sample selection bias caused by unbalanced panel data and firm exits.

3.2.2. Tax Refunds

This study employs the amount of tax refunds received by firms as a measure of tax refund quantity. Tax policies mainly include large-scale VAT end-of-period deductions and refunds, additional deductions for R&D expenses of technology-based small- and medium-sized enterprises, and VAT exemption for public transportation service income [18,20,57]. Previous studies have primarily focused on the incentive effect of individual tax refunds while neglecting the comprehensive impact of combined tax support policies. By using the amount of tax refunds received by firms as a measure, this study captures the combined incentive effect of various tax reductions and overcomes the measurement limitations of existing literature. Additionally, to mitigate the influence of heteroscedasticity in the model, this study employs the natural logarithm of tax refund amount.

Figure 2 illustrates the trend of tax refund amounts for the sample firms. The left panel displays the overall distribution of tax refund amounts received by firms. It is evident that the distribution of tax refund amounts exhibits a skewed pattern, with most firms concentrated around the value of 15. Furthermore, a significant number of firms did not receive any tax refunds. The right panel presents the distribution of tax refund amounts for different years. It can be observed that the tax refund amounts received by firms generally show an increasing trend over the years, particularly with a more pronounced upward trend in recent years.

3.2.3. Control Variables

Based on the existing literature available [53,58,59], this study controls for the following variables: SIZE, representing the firm’s size, which is the natural logarithm of total assets at the end of the year; LEV, indicating the leverage ratio, calculated as the total liabilities at the end of the year divided by total assets at the end of the year; ROA, representing the return on assets, calculated as the net income at the end of the year divided by total assets at the end of the year; CASHFLOW, representing the cash flow, calculated as the net cash flow from operating activities divided by total assets at the end of the year; FIXED, indicating the fixed asset ratio, calculated as the net value of fixed assets divided by total assets at the end of the year; GROWTH, representing the firm’s growth, measured by the growth rate of operating revenue; TOPONE, indicating the ownership concentration of the largest shareholder, calculated as the ratio of shares held by the largest shareholder to the total number of shares at the end of the period; FIRMAGE, representing the firm’s age, which is the natural logarithm of the number of years since the firm went public.

Table 1. Variable definitions.

Variable	Symbol	Definition
Total factor energy productivity	EFFILP	Calculated using the LP method
	EFFIOP	Calculated using the OP method
	EFFIGMM	Calculated using the GMM method
Tax rebates	TAXREF	Natural logarithm of one plus the amount of tax rebates received by the firm
Firm size	SIZE	Natural logarithm of total year-end assets
Leverage ratio	LEV	Total liabilities divided by total assets
Return on assets	ROA	Net profit of the company divided by total year-end assets
Cash flow	CASHFLOW	Net cash flow generated from operating activities divided by total year-end assets
Fixed asset ratio	FIXED	Net fixed assets divided by total assets
Company growth	GROWTH	Company’s revenue growth rate
Shareholding percentage of the largest shareholder	TOPONE	Number of shares held by the largest shareholder divided by total outstanding shares at the end of the period
Tobin’s Q	TOBINQ	Market value of total assets divided by the book value of total assets
Firm age	FIRMAGE	Natural logarithm of the age of the company

presents the definitions of the main variables used in this work.

3.3. Data Description

Data for this study primarily originate from the China Research Data Services Platform (CNRDS) (https://www.cnrds.com) (accessed on 26 November 2021), which offers comprehensive tax-related information for Chinese enterprises. It includes stock codes, total taxes and fees, basic company details, financial data, and tax payment records. These datasets offer valuable insights into micro-level corporate behavior. Missing data were supplemented using databases from China Securities Market & Accounting Research (CSMAR) (https://data.csmar.com, accessed on 26 November 2021), ResSET (http://www.resset.cn, accessed on 26 November 2021), and Wind (https://www.wind.com.cn, accessed on 26 November 2021).

To ensure data accuracy and validity, we followed the following steps: (1) Excluding listed companies that received special treatment during the sample period (ST, *ST) due to continuous losses over 2–3 years, which may not accurately reflect the tax burden level. (2) Excluding listed companies with negative net assets or debt-to-asset ratios exceeding 1, as they may not effectively represent the tax burden level and operational activities. (3) Excluding listed companies in the financial sector due to unique financial statement preparation. (4) Removing listed companies with significant missing values for the variables used in this study. After data processing, this study obtained 33,792 “company-year” observations, spanning manufacturing, construction, energy, and other sectors. Furthermore, a two-sided 1% Winsorization technique was employed to address potential outliers and their impact on the empirical results. This technique helps mitigate the influence of extreme values in continuous variables.

3.4. Descriptive Statistics

Table 2. Descriptive statistics.

VarName	Obs	Mean	SD	Min	Median	Max
EFFILP	33,792	8.943	1.111	5.854	8.846	12.216
EFFIOP	33,792	8.017	0.998	5.285	7.921	10.864
EFFIGMM	33,792	3.448	0.914	0.356	3.342	9.467
TAXREF	33,792	11.495	7.465	0.000	15.093	23.381
SIZE	33,792	22.059	1.275	19.236	21.881	26.398
LEV	33,792	0.446	0.204	0.027	0.446	0.991
ROA	33,792	0.037	0.066	−0.398	0.036	0.245
CASHFLOW	33,792	0.048	0.072	−0.224	0.047	0.283
FIXED	33,792	0.231	0.168	0.002	0.198	0.806
GROWTH	33,792	0.176	0.437	−0.737	0.111	4.330
TOPONE	33,792	0.351	0.151	0.083	0.329	0.758
TOBINQ	33,792	1.992	1.380	0.802	1.555	17.729
FIRMAGE	33,792	2.784	0.380	0.693	2.833	3.555

Notes: This table presents descriptive statistics for the main variables, with specific variable definitions available in Table 1. The primary data sources are CNRDS and CSMAR.

presents the main descriptive statistics of variables in specification (1), including sample number, variable means, standard deviations, minimum and maximum values, and median. EFFILP has a mean of 8.943, suggesting a relatively low overall level of total factor energy productivity among firms during the research period. The standard deviation for EFFILP is 1.111, ranging from 5.854 (minimum) to 12.216 (maximum), indicating notable variations in total factor energy productivity across firms. Similarly, EFFIOP and EFFIGMM also exhibit low overall levels and significant variations in total factor energy productivity among firms. TAXREF has a standard deviation of 7.465, ranging from 0.000 (minimum) to 23.381 (maximum), signifying substantial differences in the amount of tax refunds received by firms. Additionally, the median of TAXREF is 15.093, surpassing the mean of 11.495, implying that a sizable portion of firms receive relatively small tax refunds, resulting in a concentrated distribution within the sample. The descriptive statistics for other variables align with existing literature, affirming the soundness of the data processing in this study.

3.5. Correlation Analysis

Table 3. Correlation analysis.

VARIABLES	EFFILP	TAXREF	SIZE	LEV	ROA	CASHFLOW	FIXED	GROWTH	TOPONE	TOBINQ	FIRMAGE
EFFILP	1.000
TAXREF	0.209 ***	1.000
SIZE	0.812 ***	0.174 ***	1.000
LEV	0.404 ***	0.040 ***	0.422 ***	1.000
ROA	0.144 ***	−0.005	0.045 ***	−0.352 ***	1.000
CASHFLOW	0.092 ***	0.014 ***	0.045 ***	−0.156 ***	0.361 ***	1.000
FIXED	−0.122 ***	−0.010 *	0.035 ***	0.084 ***	−0.074 ***	0.247 ***	1.000
GROWTH	0.119 ***	−0.009	0.044 ***	0.034 ***	0.241 ***	0.033 ***	−0.051 ***	1.000
TOPONE	0.182 ***	−0.027 ***	0.183 ***	0.058 ***	0.128 ***	0.085 ***	0.094 ***	0.030 ***	1.000
TOBINQ	−0.271 ***	−0.087 ***	−0.339 ***	−0.256 ***	0.144 ***	0.081 ***	−0.114 ***	0.014 **	−0.130 ***	1.000
FIRMAGE	0.176 ***	0.008	0.223 ***	0.083 ***	−0.075 ***	−0.016 ***	−0.106 ***	−0.074 ***	−0.183 ***	0.068 ***	1.000

Notes: *, **, and *** denote significance levels of 10%, 5%, and 1%, respectively. A commonly used criterion for assessing multicollinearity is when the correlation coefficient between multiple variables exceeds 0.5, indicating a severe multicollinearity issue in the model. In this study, only the correlation coefficient between SIZE and EFFILP exceeds 0.5 among the selected control variables, while the correlation coefficients between other variables and EFFILP are all below 0.5. This does not affect the validity of the model specification in this study. Furthermore, Table 1 provides definitions of the main variables, with raw data primarily sourced from CNRDS and CSMAR.

reports the Pearson correlation coefficients of the main variables in Model 1. The correlation coefficient between TAXREF and EFFILP is significantly positive, indicating a strong correlation between tax refunds and the total factor energy productivity of enterprises. The sign of the coefficient suggests that TAXREF promotes the improvement of EFFILP, which provides preliminary support for Hypothesis 1. In addition, the selected control variables in Model 1 are significantly correlated with EFFILP, suggesting that the chosen control variables are appropriate. Furthermore, except for SIZE, the correlation coefficients between the control variables and EFFILP are all less than 0.5, indicating no multicollinearity issues in our model. However, although we have verified the correlation between TAXREF and EFFILP, we cannot determine their marginal effects. Therefore, further regression analysis is needed to examine the specific relationship between tax refunds and total factor energy productivity.

4. Empirical Analysis

4.1. Baseline Regression

Table 4. Baseline regression.

	(1)	(2)	(3)	(4)	(5)
VARIABLES	EFFILP	EFFILP	EFFILP	EFFILP	EFFILP
TAXREF	0.039 ***	0.008 ***	0.008 ***	0.008 ***	0.008 ***
	(47.51)	(15.88)	(17.13)	(16.43)	(16.79)
SIZE		0.699 ***	0.656 ***	0.668 ***	0.647 ***
		(214.40)	(200.14)	(209.07)	(193.18)
LEV		0.353 ***	0.799 ***	0.813 ***	0.801 ***
		(16.78)	(34.81)	(36.06)	(35.47)
ROA			2.459 ***	1.755 ***	1.721 ***
			(32.95)	(24.06)	(23.13)
CASHFLOW			0.611 ***	1.193 ***	1.198 ***
			(10.87)	(21.49)	(21.59)
FIXED				−1.062 ***	−1.105 ***
				(−43.44)	(−45.47)
GROWTH				0.133 ***	0.141 ***
				(14.30)	(15.01)
TOPONE					0.308 ***
					(14.43)
TOBINQ					−0.010 ***
					(−3.85)
FIRMAGE					0.026 ***
					(2.68)
Constant	8.496 ***	−6.729 ***	−6.100 ***	−6.138 ***	−5.866 ***
	(766.58)	(−100.26)	(−92.60)	(−95.78)	(−79.12)
Observations	33,792	33,792	33,792	33,792	33,792
R-squared	0.191	0.722	0.743	0.762	0.766
Industry	YES	YES	YES	YES	YES
Year	YES	YES	YES	YES	YES

Notes: This table presents the key findings of tax refunds on enterprise total factor energy productivity. EFFILP denotes enterprise total factor energy productivity estimated using the non-parametric LP method. TAXREF, the focal variable, represents tax refunds and is measured as the natural logarithm of 1 plus the amount of tax refunds for enterprises. The coefficient of TAXREF indicates its impact on enterprise total factor energy productivity. A positive coefficient suggests a positive impact, while a negative coefficient implies a suppression of improvement. Definitions of other variables are available in Table 1, with primary data sourced from CNRDS and CSMAR. Moreover, *** denote significance levels of 1%, respectively.

shows the estimated impact of tax rebates on firms’ energy efficiency, where the dependent variable is the total factor energy productivity (EFFI) and the variable of interest is the tax rebate (TAXREF). While column (1) examines the basic directional impact of tax rebates on energy efficiency, columns (2)–(5) present estimates that include firm-level control variables.

The benchmark analysis uses column (5), which finds a statistically significant effect of tax rebates on energy efficiency at the 1% level. Specifically, increasing tax rebates can encourage firms to improve their energy efficiency, with a 1% increase in tax rebates associated with a 0.8% increase in total factor energy productivity. These results validate Hypothesis 1.

Among the control variables, column (5) finds that firm size (SIZE) has a significantly positive effect on total factor energy productivity at the p < 0.1 level. This suggests that larger firms may benefit from scale effects, leading to reduced energy consumption per unit of output and improved energy use efficiency. The estimate of leverage (LEV) also has a significantly positive effect at the 1% level, which differs from existing literature. One possible explanation is that the increasing leverage level provides a factor guarantee for firms to improve their total factor energy productivity. In addition, the estimates of ROA, CASHFLOW, and GROWTH are significantly positive, indirectly supporting the idea that enhancing profitability and cash flow can lead to improved energy efficiency. The estimates of other variables are consistent with previous literature.

This study’s findings are consistent with He, et al. [60] research but extend and deepen the understanding of the relationship between tax incentives and energy use efficiency by analyzing a comprehensive indicator of tax rebates. In contrast to the literature on He, Sun, Niu, Long, and Li [60] which focuses on a single energy tax incentive policy at the national level, this study presents a broad analysis of tax incentives’ impact on energy efficiency.

4.2. Robustness

The study confirms the positive impact of tax rebates on firms’ energy efficiency. To enhance the credibility of this benchmark conclusion, the instrumental variable approach and Heckman strategy are employed to address endogeneity concerns. Moreover, additional robustness tests, including core variable substitution and fixed-effect augmentation, are conducted.

4.2.1. Endogeneity

In this study, endogeneity issues may arise from two aspects: (1) Reverse Causality. Tax rebates incentivize firms to enhance their overall factor energy efficiency. Conversely, continuous improvements in energy efficiency reduce environmental damage, thereby increasing the likelihood of government approval and receiving more tax rebates. (2) Sample Selection Bias. It is important to note that our analysis focuses solely on the impact of tax rebates on total factor energy efficiency in listed companies. We acknowledge that there are other entities, such as non-listed companies and small- and medium-sized enterprises (SMEs), that also benefit from various tax incentives. However, due to the scope of this study, these entities have not been included, which may introduce sample selection bias.

(1)

2SLS

To address endogeneity concerns due to reverse causality, this study uses the 2SLS model with a lagged TAXREF as an instrumental variable for estimation. Validation confirms a significant correlation between the instrumental and endogenous variables, meeting the exogeneity requirement. Weak instrument testing also shows no significant correlation between the instrumental variable and total factor energy efficiency. In

Table 5. Endogeneity treatment.

	(1)	(2)
	2SLS	Heckman
VARIABLES	EFFILP	EFFILP
TAXREF	0.007 ***	0.008 ***
	(14.27)	(16.84)
IMR		0.855 *
		(1.82)
SIZE	0.646 ***	0.703 ***
	(164.93)	(22.58)
LEV	0.842 ***	0.684 ***
	(32.29)	(9.97)
ROA	2.026 ***	1.431 ***
	(23.08)	(8.11)
CASHFLOW	0.937 ***	1.297 ***
	(14.67)	(16.80)
FIXED	−0.955 ***	−1.144 ***
	(−35.15)	(−35.60)
GROWTH	0.138 ***	0.126 ***
	(12.57)	(10.30)
TOPONE	0.344 ***	0.104
	(14.16)	(0.91)
TOBINQ	0.010 **	−0.026 ***
	(2.52)	(−3.01)
FIRMAGE	−0.010	−0.044
	(−0.94)	(−1.11)
Constant	−5.759 ***	−7.114 ***
	(−66.68)	(−10.28)
Observations	29,149	33,792
R-squared	0.734	0.766
Industry	YES	YES
Year	YES	YES

Notes: Table 2 displays descriptive statistics for the main variables. Variable definitions can be found in Table 1. Raw data are primarily sourced from CNRDS and CSMAR. IMR represents inverse Mills ratio. Moreover, variable definitions for the main variables are provided in Table 1, with raw data sourced from CNRDS and CSMAR. Additionally, robust t-statistics are presented in parentheses, denoting significance levels as *** (p < 0.01), ** (p < 0.05), and * (p < 0.1). *, **, and *** denote significance levels of 10%, 5%, and 1%, respectively.

, column (1), the instrumental variable method’s estimation results reveal a significantly positive estimate for TAXREF, indicating that tax rebates can improve the overall factor energy productivity of enterprises even after considering reverse causality issues. This supports our baseline conclusion.

(2)

Heckman strategy

To address sample self-selection bias, we employ the Heckman method. First, we construct dummy variables using the mean of TAXREF and estimate them through a Probit model, controlling for covariates in the specification (1). Based on these estimations, we derive the variable IMR and include it in the estimation of specification (1). The Heckman strategy estimation results are shown in Table 5, column (2). In column (2), the positive and significant estimate of IMR indicates the presence of self-selection issues in our research sample, validating the need to consider sample selection bias. Moreover, the estimate of TAXREF remains significantly positive at a 1% statistical level, indicating that tax rebates continue to significantly enhance the overall factor energy efficiency of enterprises even after accounting for sample selection bias.

4.2.2. Additional Robustness Checks

(1)

Alternative Measurement Approaches

This study utilizes OP and GMM methods to measure overall factor energy productivity in enterprises. The estimation results in

Table 6. Additional robustness tests.

	(1)	(2)	(3)	(4)	(5)
	Alternative Measurement of the Variable of Interest		Introducing Fixed Effects		Tobit
VARIABLES	EFFIOP	EFFIGMM	EFFILP	EFFILP	EFFILP
TAXREF	0.006 ***	0.004 ***	0.004 ***	0.004 ***	0.007 ***
	(12.71)	(7.65)	(10.413)	(10.501)	(16.211)
SIZE	0.542 ***	0.155 ***	0.568 ***	0.567 ***	0.661 ***
	(164.39)	(40.13)	(124.277)	(123.048)	(200.620)
LEV	0.638 ***	0.728 ***	0.290 ***	0.275 ***	0.811 ***
	(29.03)	(28.63)	(15.292)	(14.498)	(41.051)
ROA	1.459 ***	1.423 ***	1.065 ***	1.068 ***	1.742 ***
	(20.57)	(16.88)	(26.386)	(26.282)	(30.565)
CASHFLOW	0.767 ***	0.763 ***	0.754 ***	0.745 ***	1.215 ***
	(13.61)	(11.37)	(23.816)	(23.382)	(25.408)
FIXED	−0.612 ***	−2.844 ***	−0.879 ***	−0.867 ***	−1.078 ***
	(−25.55)	(−101.63)	(−38.013)	(−37.311)	(−48.712)
GROWTH	0.155 ***	0.181 ***	0.188 ***	0.188 ***	0.138 ***
	(16.32)	(16.20)	(41.557)	(41.278)	(19.090)
TOPONE	0.266 ***	0.300 ***	0.026	0.046	0.367 ***
	(12.33)	(12.09)	(0.942)	(1.632)	(17.203)
TOBINQ	−0.008 ***	−0.005	0.009 ***	0.009 ***	−0.011 ***
	(−2.90)	(−1.53)	(4.278)	(4.447)	(−4.189)
FIRMAGE	0.042 ***	0.044 ***	0.237 ***	0.183 ***	0.052 ***
	(4.22)	(3.85)	(9.612)	(7.174)	(5.072)
Constant	−4.485 ***	−0.055	−4.352 ***	−4.191 ***	−6.493 ***
	(−61.22)	(−0.65)	(−37.552)	(−35.621)	(−87.898)
Observations	33,792	33,792	33,596	33,587	33,792
R-squared	0.704	0.543	0.926	0.928	——
Industry	YES	YES	YES	YES	YES
Year	YES	YES	YES	YES	YES
Stkcd	NO	NO	YES	YES	NO
Industry × Year	NO	NO	NO	YES	NO

Notes: This table displays the results of robustness tests in this study. TAXREF, the variable of interest, represents tax refunds and is measured as the natural logarithm of 1 plus the amount of tax refunds for enterprises. Definitions of other variables can be found in Table 1. The primary data source is from CNRDS and CSMAR. Additionally, robust t-statistics are reported in parentheses, with *** denoting significance at the 0.01 level.

, specifically columns (1) and (2), demonstrate statistically significant and positive coefficients for TAXREF on EFFIOP and EFFIGMM. These findings indicate that tax rebates effectively enhance energy efficiency, even with alternative measurement approaches for overall factor energy productivity. These results further validate the robustness of the baseline empirical findings.

(2)

Additional Fixed Effects

While conducting the baseline regression, it is important to consider potential heterogeneity across different entities, which may introduce biases in the estimation results. To address this concern, individual differences are controlled for as discussed in section (1). Table 6, column (3), presents the estimates by considering individual variations. Furthermore, given the variations in different industries over time, this study includes interaction terms between industry and year in the model for estimation. Table 6, column (4), displays the estimates by controlling for changes in different industries over time. The estimates of TAXREF remain significantly positive in columns (3) and (4), indicating that the baseline results hold even after accounting for other individual factors and industry-specific effects over time.

(3)

Other Econometric Strategies

Since both the tax rebates and the indicators of overall factor energy productivity used in this study take non-negative values, implying a truncated distribution, it is necessary to employ appropriate econometric models to avoid potential biases. Therefore, this study employs the panel Tobit model for estimation. Table 6, column (5), presents the estimation results using the Tobit model. In column (5), the estimated coefficient of TAXREF is statistically significant at the 1% level, indicating that the fundamental result of tax rebates enhancing the overall factor energy efficiency of enterprises holds even after employing alternative econometric models for estimation.

The preceding empirical results demonstrate that, after considering endogeneity concerns and conducting a series of additional robustness tests, tax rebates continue to significantly enhance the overall factor energy productivity of enterprises. These findings confirm the robustness of the baseline results obtained in this study.

4.3. Heterogeneity Checks

The empirical results above confirm that tax rebates significantly enhance the overall factor energy efficiency of firms. This benchmark finding remains robust even after addressing endogeneity concerns and conducting other robustness checks. However, does the impact of tax rebates on improving overall factor energy efficiency vary across firm and industry characteristics? To answer this question, this section of the paper will examine heterogeneity from three perspectives: firm size, firm attributes, and technological level.

4.3.1. Firm Size

The size of enterprises significantly affects their overall factor energy productivity. Large enterprises benefit from financial resources, human capital, and technological reserves, enhancing their innovative capabilities. However, they may face organizational complexities. On the other hand, SMEs often experience financing constraints that limit their innovation activities. Improving the business environment has two positive outcomes. Firstly, it attracts socially responsible and innovative talents, enhancing the human capital structure of large enterprises and improving innovation efficiency. Secondly, it alleviates the financing constraints faced by SMEs. This study classifies enterprises based on the median size. Those above the median are considered large-scale enterprises, while those below are categorized as small- and medium-sized enterprises.

Table 7. Heterogeneity analysis.

	(1)	(2)	(3)
VARIABLES	EFFILP	EFFILP	EFFILP
TAXREF	0.011 ***	0.007 ***	0.009 ***
	(8.33)	(11.95)	(14.37)
LARGE	0.134 ***
	(10.11)
TAXREF×LARGE	−0.002 ***
	(−2.65)
SOE		0.067 ***
		(5.29)
TAXREF×SOE		0.002 *
		(1.82)
NONTECH			−0.145 ***
			(−9.46)
TAXREF×NONTECH			0.004 ***
			(3.99)
SIZE	0.619 ***	0.647 ***	0.647 ***
	(153.06)	(192.50)	(193.34)
LEV	0.789 ***	0.801 ***	0.805 ***
	(34.98)	(35.47)	(35.69)
ROA	1.679 ***	1.724 ***	1.737 ***
	(22.64)	(23.15)	(23.33)
CASHFLOW	1.186 ***	1.200 ***	1.206 ***
	(21.49)	(21.63)	(21.76)
FIXED	−1.109 ***	−1.103 ***	−1.080 ***
	(−45.63)	(−45.40)	(−43.90)
GROWTH	0.144 ***	0.141 ***	0.140 ***
	(15.46)	(15.00)	(14.98)
TOPONE	0.301 ***	0.307 ***	0.304 ***
	(14.16)	(14.38)	(14.28)
TOBINQ	−0.010 ***	−0.011 ***	−0.012 ***
	(−3.63)	(−3.92)	(−4.37)
FIRMAGE	0.034 ***	0.026 ***	0.028 ***
	(3.51)	(2.60)	(2.82)
Constant	−5.450 ***	−5.846 ***	−5.933 ***
	(−63.76)	(−78.03)	(−79.55)
Observations	33,792	33,792	33,792
R-squared	0.767	0.766	0.766
Industry	YES	YES	YES
Year	YES	YES	YES

Notes: This table displays the heterogeneity analysis results in this study. The coefficients of the interaction terms (TAXREF × LARGE, TAXREF × SOE, TAXREF × NONTECH) are central to our investigation, revealing the moderating effects of various factors in the relationship between tax refunds and the total factor energy productivity of enterprises. Variable definitions for others can be found in Table 1. The primary data source is from CNRDS and CSMAR. Additionally, robust t-statistics are presented in parentheses, with *** denoting significance at the 0.01 level and with * denoting significance at the 0.1 level.

, column (1), presents the estimation of the impact of tax rebates on the overall factor energy productivity of large-scale enterprises. The estimate for TAXREF×LARGE is significantly negative, indicating that tax rebates do not improve the overall factor energy productivity of large-scale enterprises. Conversely, tax rebates incentivize SMEs to enhance their overall factor energy productivity.

4.3.2. Firm Attributes

The ownership structure of enterprises potentially affect the relationship between tax rebates and overall factor energy productivity. Generally, state-owned enterprises (SOEs) are subject to strict government control, bear social responsibilities, and have dual economic−political attributes. Compared to private enterprises, SOEs face stronger regulatory and market pressures and have a common goal of improving energy efficiency and reducing emissions. Moreover, due to their natural association with the government, SOEs have lower information asymmetry, making them more likely to receive tax incentives. Therefore, in practical terms, tax rebates may have a stronger stimulating effect on SOEs’ overall factor energy productivity than on private enterprises.

This study categorized enterprises into SOEs and non-SOEs based on the largest shareholder. Table 7, column (2), presents the estimation of tax rebates’ impact on SOEs’ overall factor energy productivity. The estimate of TAXREF × SOE is statistically significant and positive, indicating that tax rebates contribute to enhancing SOEs’ overall factor energy productivity. Conversely, tax rebates do not significantly impact private enterprises’ overall factor energy productivity. This could be attributed to the tax concealment effect in private enterprises, leading to a relatively weaker influence of tax incentives compared to SOEs.

4.3.3. Industry Heterogeneity

This paper argues that tax rebates have differing effects on the overall factor energy productivity of enterprises based on their technological attributes. Non-tech firms face significant challenges such as financing difficulties and insufficient investment in R&D. These factors impede their ability to improve overall factor energy productivity, particularly in emerging economies. Tax rebates have a more pronounced effect on enhancing the overall factor energy productivity of non-tech firms compared to tech firms.

Using the Chinese government’s categorization criteria for high-tech industries in “Classification of High-Tech Industries (Manufacturing) 2017” and “Classification of High-Tech Industries (Services) 2018,” this study created a dummy variable for non-tech firms (NONTECH). Table 7, column (3), shows the impact of tax rebates on the overall factor energy productivity of non-tech firms. The estimate of TAXREF × NONTECH is positive, indicating a highly significant influence of tax incentives on enhancing the overall factor energy productivity of non-tech firms.

4.4. Mechanism Analysis

The above analysis confirms that tax rebates significantly enhance overall energy efficiency in firms, with variations related to firm size, attributes, and technological advancements. However, the paper has not yet explained the mechanism behind this impact. Therefore, the paper will empirically investigate the underlying factors through which tax rebates affect overall energy efficiency in firms.

4.4.1. Innovation Effects in Enterprises

Theoretical analysis reveals that tax rebate policies can impact corporate innovation, encouraging companies to improve their overall energy efficiency. These policies, such as tax rebates, reduce local governments’ reliance on tax revenue, lightening their fiscal burden and decreasing the tax load on businesses, thereby providing compensation through cash flow. When tax rebate policies support funds to offset positive externalities of corporate R&D, companies are motivated to increase their R&D investments, further enhancing their total factor energy productivity.

To test this theory empirically, we examine the effect of tax rebates on corporate total factor energy productivity using the natural logarithm of a company’s annual patent count as a proxy for innovation. In

Table 8. Mechanism analysis.

	(1)	(2)
VARIABLES	PATENT	FSTRAIN
TAXREF	0.019 ***	−0.000 ***
	(18.54)	(−3.07)
SIZE	0.271 ***	0.022 ***
	(26.84)	(19.16)
LEV	−0.293 ***	−0.022 ***
	(−6.25)	(−5.34)
ROA	2.249 ***	−0.198 ***
	(16.31)	(−16.43)
CASHFLOW	0.813 ***	−0.018 *
	(7.43)	(−1.87)
FIXED	−0.257 ***	0.042 ***
	(−4.63)	(8.74)
GROWTH	−0.138 ***	−0.005 ***
	(−8.77)	(−2.93)
TOPONE	−0.042	0.024 ***
	(−0.78)	(5.40)
TOBINQ	−0.006	0.024 ***
	(−0.92)	(30.43)
SOE	−0.133 ***	−0.012 ***
	(−7.59)	(−8.44)
FIRMAGE	−0.365 ***	−0.597 ***
	(−14.23)	(−220.54)
Constant	−3.493 ***	−2.593 ***
	(−15.64)	(−104.08)
Observations	33,792	33,792
R-squared	0.345	0.819
Industry	YES	YES
Year	YES	YES

Notes: This table presents the empirical results of the mechanism analysis. PATENT represents the capability of firm innovation, measured as the logarithm of 1 plus the number of patents obtained by the firm in the current year. FSTRAIN indicates the degree of financial constraints on firms, measured by the SA index. The variable of interest in this study is TAXREF, representing tax refunds and measured as the natural logarithm of 1 plus the amount of tax refunds for enterprises. Definitions of other variables are available in Table 1. The primary data source is from CNRDS and CSMAR. Additionally, * and *** denote significance levels of 10%, and 1%, respectively.

, column (1), we present estimates for the channel through which tax rebates affect corporate total factor energy productivity. The focus of this paper is the estimate of TAXREF, reflecting the impact of tax rebates on corporate innovation. Importantly, the estimate of TAXREF is statistically significant at the 1% level, indicating that tax rebates stimulate technological innovation in companies, thus improving their overall energy efficiency.

This finding is consistent with existing research. d’Andria and Savin [61] discovered that tax incentives promote innovation, leading to increased corporate environmental sustainability. However, this study delves deeper into how corporate technological innovation drives improvements in total factor energy productivity, thus contributing to the understanding of the relationship between tax incentives and corporate innovation.

4.4.2. Financial Constraints Effects

The implementation of tax rebate policies has an impact on corporate financing constraints. On the one hand, when companies are under pressure from tax burdens, a series of tax support policies, including tax rebates, can effectively alleviate these constraints. When businesses face financing constraints, it becomes challenging to obtain external funding or secure it at a lower cost. Tax rebates can serve as a source of external funding for companies, potentially enabling them to engage in refinancing. Expanding the scale of financing and reducing the cost of financing can improve a company’s liquidity, allowing for the expansion of capital and human resources, ultimately enhancing the firm’s total factor energy productivity.

In this study, we measure corporate financing constraints (FSTRAIN) using the SA index and investigate whether tax rebates alleviate these constraints, consequently enhancing corporate total factor energy productivity. Table 8, column (2), presents the estimates of the impact of tax rebates on corporate financing constraints. In column (2), the estimate of TAXREF is negative, with a p value < 0.01, indicating that tax rebates reduce the financing constraints faced by companies, thereby improving their overall energy efficiency. This finding is consistent with the research of Yang, He, Xia, and Chen [30] and Shi, Yang, and Ji [53].

The empirical findings above demonstrate that tax rebates enhance corporate innovation capabilities, alleviate financing constraints, and thereby improve a firm’s overall total factor energy efficiency.

5. Conclusions and Implications

How does tax refund affect green economic development? Previous research on the effects of tax refunds on green economic development has mainly examined macro and regional perspectives. However, achieving green development requires enhancing the total factor energy productivity of businesses at the micro level. Therefore, a systematic analysis of how tax refunds influence the total factor energy productivity of enterprises is critical for understanding the micro-mechanisms behind how tax policies affect economic growth and enhance the effectiveness of green development.

This paper utilizes tax survey data from Chinese listed companies between 2004 and 2020 to examine the effects and mechanisms of tax refunds on the total factor energy productivity of enterprises at the firm level. The study finds that tax refund growth has a positive impact on total factor energy productivity, verifying Hypothesis 1 of this study. This conclusion holds even after conducting a series of robustness tests and addressing endogeneity concerns. Heterogeneity analysis reveals that tax refunds enhance the total factor energy productivity of state-owned enterprises, small- and medium-sized enterprises, and non-technology intensive enterprises. Mechanism analysis suggests that tax refund promotes the improvement of total factor energy productivity by stimulating enterprise innovation and alleviating financing constraints.

The conclusions of this study provide several implications for evaluating tax policies. First, fiscal policies should be based on socio-economic development and scientific principles. The study found that tax incentives encourage enterprises to enhance their overall energy productivity. Therefore, government authorities should actively monitor and control the total tax burden and its growth rate to prevent excessive burdens. Additionally, local governments should refine tax refunds and other support policies.

Second, optimizing the structure of fiscal revenue and expenditure is crucial. This involves establishing a robust system for fiscal revenue and expenditure with clearly defined responsibilities for government departments. Moreover, managing administrative costs, reducing corporate and individual taxes, cutting public expenditures, and optimizing tax refund policies can effectively alleviate tax burdens. Government authorities should also prudently manage government financial debt and assets.

Third, it is important to increase credit support for private, small- and medium-sized, and non-high tech enterprises to leverage their role in promoting shared prosperity. The study found that excessive tax burdens worsen financing constraints for private enterprises. Local governments should address issues such as difficulties, high costs, and slow processes faced by these enterprises in accessing financing. This can be achieved by strengthening the connection between banks and enterprises, promoting cooperation for mutual benefit, diversifying external financing channels, simplifying financing procedures, and reducing associated costs to lower financing expenses.

This study exhibits certain research limitations. First, it exclusively relies on samples from Chinese companies for estimation, which restricts the generalizability of the empirical findings to other large developing countries. Future research could encompass samples from emerging nations such as India and Brazil to broaden the applicability of the conclusions drawn in this paper. Second, the estimation in this study is conducted solely on samples from publicly traded Chinese companies, neglecting a vast pool of data from non-listed companies. To enhance the reliability of our estimates, future studies can incorporate non-listed company samples through the utilization of survey methodology.