How do vertical fiscal imbalances affect energy efficiency? The role of government spending on science and technology

Abstract

How to improve energy efficiency is a practical issue of common concern around the world. In China, vertical fiscal imbalances (VFIs) generated under the existing fiscal system may have a significant impact on energy efficiency through government spending on science and technology (S&T). However, this potential relationship has not been explored in the literature. In this work, we aim to address this gap. Using Chinese provincial panel data from 2001 to 2017, this study empirically examines the relationship between VFIs and energy efficiency through a mediation model. The results show that VFIs greatly suppress energy efficiency. We further find that VFIs indirectly affect energy efficiency by reducing government spending on S&T. The results show significant regional heterogeneity. The intermediary role of government S&T expenditure is more significant in inland areas than in coastal areas. Therefore, to improve energy efficiency and achieve sustainable development, the Chinese government should focus on innovative reform of the existing fiscal system and reduce VFIs. In addition, the government should focus on fiscal spending in the field of S&T to promote technological innovation to guarantee the improvement of energy efficiency. Inland areas should pay particular attention to this issue.

Introduction

In recent decades, China’s rapid economic development has created the “Chinese miracle,” attributed to China’s continuous implementation of an extensive development model. However, this “pollute first, treat later” character of this model generates high-energy waste and high environmental pollution (Liu et al. 2015; Zhu et al. 2019; Hu et al. 2020), which is not conducive to sustainable economic development. China has become one of the most polluted countries in the world (Huang 2017), leading in CO₂ emissions (Wang and Lin 2017). Despite China’s “3060” dual-carbon target, Climate Action Tracker (CAT)1 recently rated China’s climate commitment for 2030 as “highly inadequate.” According to the Statistical Review of World Energy (2022), China’s energy consumption accounted for 26.5% of global energy consumption in 2021, ranking first globally. With such enormous energy consumption, if China can use its energy efficiently, it will contribute to its sustainable economic growth. However, according to the energy conversion efficiency indicators released by the National Bureau of Statistics, China’s energy conversion efficiency rose from 69.7% in 2001 to 73.3% in 2020, rising by only 3.6 percentage points in 20 years and leaving a large gap remaining with respect to developed countries. Furthermore, according to the China Energy Statistical Yearbook, China’s energy intensity is declining yearly. The continued decline is due to lower growth rates in energy consumption than in GDP (Liao et al. 2007), which has not substantially improved. All the signs show that the practical problems of environmental pollution and energy waste have not been fundamentally solved. Energy inefficiency is at the root of the problem, preventing the economy from moving in the right direction. Improving it has key implications for driving sustainable economic growth (Li and Lin 2017; Wu et al. 2020). China has also made many efforts to fundamentally solve this problem. In 2021, China issued the “Opinions on Complete, Accurate and Comprehensive Implementation of the New Development Concept and Carbon Neutralisation,” setting three phased goals. The first goal is to “significantly improve energy efficiency in key industries by 2025 and reduce energy consumption per unit of GDP by 13.5% compared with 2020.” The second goal is that “by 2030, the energy efficiency of key energy-consumption industries will reach the advanced international level, and the energy consumption per unit of GDP will fall significantly.” The third goal is that “energy efficiency will reach advanced international levels by 2060.” In 2022, China deployed the “Implementation Plan on Promoting High-Quality Development of New Energy in the New Era,” identifying 21 measures to improve energy efficiency. These policies are aimed at improving energy efficiency. In this context, it is of great practical importance to study how to effectively improve energy efficiency and provide recommendations for policy optimization.

Academia has put considerable effort into the study of energy efficiency. Current research focuses on energy efficiency measurements and influencing factors. Hu and Wang (2006) used the data envelopment analysis method for the first time to measure the energy efficiency of 29 provinces in China from 1995 to 2002. In this study, they failed to consider computational bias from undesired outputs, such as environmental pollution. Therefore, subsequent studies have considered a single undesirable output (CO₂ emissions) in calculating energy efficiency (Meng et al. 2016; Song et al. 2016, 2021; Wang and Wang 2020). However, pollution from energy consumption is multifaceted, and considering only a single undesired output can bias the results. Thus, we fully consider carbon dioxide emissions, SO₂ emissions, various accidental dust emissions, and wastewater emissions in our calculations. More extensive studies on the factors affecting energy efficiency are available. Examples of these factors include urbanization (Poumanyvong and Kaneko 2010), technological innovation (Tan and Lin 2018; Wurlod and Noailly 2018; Sun et al. 2019, 2021), China’s Energy Intensity Constraint Policy (Shao et al. 2019), climate change (Yalew et al. 2020), environmental regulation and environmental rights decentralization (Wu et al. 2020), digitization (Ren et al. 2021), and financial development (Ji and Zhang 2019; Song et al. 2021).

However, few studies have focused on the impact of vertical fiscal imbalances (VFIs) on energy efficiency. Improving energy efficiency helps to reduce energy waste and environmental pollution and has positive externalities for the external environment. However, improving energy efficiency is not easy: it often depends on the market and is influenced by intergovernmental relationships. Like many countries in the world, China needs full coordination between the central government and local governments to effectively solve the problem of energy waste. Without good cooperative relations between governments, achieving energy conservation and emission reduction is challenging. However, since the tax-sharing reform in 1994, fiscal relationships between the Chinese governments have been imbalanced (Liu and Zhang 2021). Specifically, given the asymmetry in the distribution ratio of fiscal responsibilities and fiscal authority between the upper and lower governments, local governments bear a heavy income and expenditure gap, leading to high VFIs (Jia et al. 2014) and posing great difficulties for active cooperation among governments. Greater responsibility for spending has troubled local governments with respect to balancing economic development against environmental protection. Local governments prioritize allocations of fiscal resources to extensive development projects, ignoring the consequences of energy waste and environmental pollution (Tang et al. 2021; Lin and Zhou 2021a; Feng et al. 2022). This choice runs counter to the objectives of high-quality economic development. This particular context of intergovernmental fiscal relationships provides an interesting setting for this study, in which we examine the following question: Do VFIs inhibit energy efficiency? This issue has been studied by Lin and Zhou (2021a), who find that VFIs can inhibit energy and environmental performance. They also delve into the impact of the mechanisms at play, such as industrial structure upgrading, technological innovation, and government intervention. However, they overlook the role of government spending on science and technology (S&T). VFIs reduce government S&T spending (Liu and Zhang 2021), which may be an important mediating mechanism by which VFIs affect energy efficiency. The impact of government spending on S&T has theoretical implications, and our study fills this gap. In addition, from a practical point of view, exploring the impact of VFI on energy efficiency from the perspective of government spending on S&T is in line with the government’s approach to achieving the “3060” dual carbon target. Therefore, this study systematically reviews the literature and develops a theoretical framework on VFIs and energy efficiency in an attempt to explain the role of government spending on S&T and to further explore the regional heterogeneity in the above relationship.

The contribution of this study includes the following two main aspects. First, we make a useful addition to existing research from an academic perspective. Lin and Zhou (2021a) explore the relationship of VFIs with energy and environmental performance, but they ignore the intermediary role of government spending on S&T. This study systematically examines the impact of VFI on energy efficiency from the perspective of government spending on S&T. At the same time, different evidence is provided from the perspective of regional development differences. Second, this study makes specific policy recommendations. These research findings can serve as a reference for fiscal system reform and make important contributions to achieving energy conservation and emission reduction at this stage and deciphering the internal logic of the new path of high-quality economic development in China.

The structural arrangement of this paper is as follows. The “Introduction” section is the introduction. The “Literature review and hypothesis development” section presents the literature review and hypothesis development. The “Empirical strategy, variables and data” section discusses the design of the empirical strategy. The “Empirical results” section immediately follows with the empirical results. The “Conclusion and policy implications” section summarizes policy arguments.

Literature review and hypothesis development

Literature review

Many researchers have explored the economic consequences of VFIs from different perspectives. From an economic growth perspective, VFIs can harm economic growth (Mitra and Chymis 2022). Belgium’s constitutional federalism reform of 1993 contributed to an increase in the level of VFIs, thereby undermining economic growth. This negative impact was reduced when the reform was discontinued and VFIs weakened. From the perspective of local government behavior, VFIs can be harmful to local government fiscal behavior and decision-making. VFIs reduce the overall fiscal balance of local governments and have a negative effect on local fiscal performance (Eyraud and Lusinyan 2013). When interacting with horizontal fiscal imbalances, VFIs are more harmful to local finances. This is because VFIs affect local government fiscal spending policies. VFIs lead to a deviation in the structure of local fiscal expenditure, which further affects the efficiency of resource allocation (Bardhan and Mookherjee 2006). Under higher VFIs, local governments increase spending on capital construction but reduce spending on education and administration (Jia et al. 2014) and squeeze spending on S&T (Liu and Zhang 2021). Furthermore, it has been shown that VFIs significantly inhibit local governments’ tax efforts when accompanied by a reduction in their revenue power (Jia et al. 2021). This is detrimental to local government fiscal sustainability (Li and Du 2021). From the perspective of industrial development, VFIs hinder industrial structure upgrading (Lin and Zhou 2021b). First, VFIs encourage local governments to invest more in the industrial sector, which in turn hinders industrial structure upgrading. Second, VFIs create market distortions through resource misallocation, which in turn has a negative impact on industrial structural upgrading. Finally, VFIs inhibit technological innovation and lead to the blocking of industrial development.

The current research on VFIs and energy efficiency is still in its infancy. In the field of fiscal and energy policy, previous studies have focused principally on the influence of VFIs on energy and environmental performance (Lin and Zhou 2021a), energy intensity (Liu and Zhang 2022a), energy consumption (Liu and Zhang 2022b), and CO₂ emissions (Huang and Zhou 2020; Li et al. 2021; Feng et al. 2022). First, the most relevant study is that of Lin and Zhou (2021a). They argue that VFI significantly degrades energy and environmental performance, where the adjustment mechanisms are industrial structure upgrading, technological innovation, and government intervention. However, the study has shortcomings. On the one hand, it does not take into account the fact that local governments’ biased spending choices may also be one of the channels of influence. On the other hand, the study fails to delve into regional heterogeneity in the mechanisms of influence. Second, VFIs have a positive effect on energy intensity because they increase energy consumption. VFIs significantly boost energy consumption by slowing industrial upgrading, which is more evident in the inland sample (Liu and Zhang 2022b). In addition, it is clear that VFIs are very detrimental to technological innovation, which can lead directly to an increase in energy intensity. VFIs have not only a positive effect on local energy intensity but also a positive spillover effect on neighboring regions (Liu and Zhang 2022a). Finally, other studies have looked at the impact of VFI son CO₂ emissions. Huang and Zhou (2020) and Feng et al. (2022) find that VFIs significantly increase CO₂ emissions in China, leading to environmental pollution. Li et al. (2021) use Pakistani provincial panel data to study the influence of VFIs on CO₂ emissions and come to a similar conclusion. In terms of mechanisms of action, the above studies have coincidentally focused on the impact of technological innovation, industrial restructuring, government intervention, and environmental regulation. However, in reality, VFIs directly influence local government fiscal spending decisions (Jia et al. 2014; Liu and Zhang 2021) and may also influence energy efficiency through government spending on S&T. This dimension has been overlooked by these works.

In summary, the above literature on VFIs and energy efficiency–related fields has made corresponding contributions, but the research remains insufficient. First, in terms of the impact mechanism, existing research has not considered whether VFIs affect energy efficiency by affecting the choice of local government spending, and relevant theoretical and empirical research is lacking. Second, existing studies fail to consider regional heterogeneity at the mechanism level. The shortcomings of the above research provide opportunities for further research. Therefore, this study begins with a theoretical dissection of how VFIs can influence energy efficiency through government spending on science and technology. Then, using Chinese provincial panel data from 2001 to 2017, we empirically examine the relationship through a mediation model. Compared to previous literature, our paper offers a new research perspective and makes additions.

Hypothesis development

VFIs and energy efficiency

From the perspective of the historical development process, most countries in the world have experienced institutional reforms of fiscal decentralization (Eyraud and Lusinyan 2013; Lin and Zhou 2021a), as has China. During the transition from a planned economy to a market economy, China’s fiscal system has undergone three important reforms. In the first two reforms, the taxation power of the central government was decentralized. The last and most recent was the tax-sharing reform in 1994. Unlike the previous two reforms, this tax-sharing reform allowed the central government to recentralize taxation power (Liu and Zhang 2021). Given that the focus of previous reforms was mainly on the distribution of revenue power, no coordination of expenditure responsibilities has been imposed. Therefore, the implementation of the tax-sharing reform has directly led to more spending responsibilities being delegated to local governments. Local governments appear to be more autonomous. However, their income power has been weakened as they have assumed unmatched spending responsibilities, resulting in mismatched fiscal resources and serious vertical fiscal imbalances (Jia et al. 2014, 2021).

High VFIs may hinder the improvement of energy efficiency. We analyze this issue along two main aspects. On the one hand, VFIs distort the fiscal spending behavior of local governments, resulting in biased investment. China is a country with “political centralisation and economic decentralisation” (Du et al. 2014). The subordinate relationship between governments is the most basic feature of the Chinese government structure. The traditional view holds that information asymmetry exists between upper and lower levels of government and that decentralizing expenditure responsibility to local governments can solve this problem, thereby improving government efficiency (Oates 1972; Qian and Roland 1998). However, the mismatch between revenue and expenditure leads to a high degree of VFIs, affecting local governments’ choice of spending behavior. To ease fiscal pressure, local governments can only increase investment in productive expenditures (Jia et al. 2014), focusing only on the growth of absolute GDP. The possible direct result is that local governments do not consider energy efficiency issues, resulting in a poor ecological environment. On the other hand, under the political promotion tournament, VFIs cause incentive distortions to the behavior of local governments (Jia et al. 2021), strengthen the use of fiscal means by local governments to intervene in industrial development, and promote the extensive development of local economies. Political centralization and economic decentralization have given local economies enormous dominance in the economy. Given officials’ short tenure in office, in the pursuit of high short-term performance under the political promotion tournament, local officials have an investment preference for “heavy production and less innovation” (Wu 2017). Under a high degree of VFIs, to develop the economy, local governments usually use policies and control factor resources to guide capital investment, resulting in economic growth at the cost of repeated construction (Wu 2017; Lin and Zhou 2021b; Shen and Lin 2021). Long-term blind investment and repeated construction have led to excess capacity in the affected industries (Ma et al. 2020), resulting in wasted energy. As both “referee” and “athlete” in the game of local economic development, local governments lack sufficient incentive to promote energy conservation and emission reduction, which is not conducive to improving energy efficiency. In this context, we propose the following hypothesis:

• Hypothesis 1. Ceteris paribus, the higher the VFIs are, the lower the energy efficiency.

Mechanism analysis

We discuss further potential channels that may exist. Innovation investment in S&T has the specific characteristics of being high cost and high risk while having a long cycle and a low success rate (Hsu et al. 2014; Cui et al. 2021). Excessive local government spending on S&T will not create boost performance much in the short term. In the pursuit of maximizing short-term benefits, VFIs exacerbate the distortions in local government fiscal behavior (Li and Du 2021). Local governments may increase short-term productive spending that achieves high returns, thereby squeezing government spending on S&T (Liu and Zhang 2021). To some extent, governments’ and officials’ spending preference for “focusing on production and ignoring innovation” also guides the biased allocation of social capital. In a government-led institutional environment, local governments can influence the direction of social capital through multiple means, such as exercising government power, formulating industrial policies, and intervening in corporate decision-making (Wu 2017). To achieve the objective of improving energy efficiency, the government urgently needs to increase spending on S&T. The reason is that the energy efficiency problem is essentially a problem between input and output. By increasing spending on S&T and formulating relevant policies, the government is also sending an important signal guiding social capital in the market to flow into high-tech industries. Only by increasing government spending on S&T can the S&T sector have the opportunity to obtain more R&D funds and policy support. However, judging from the current situation, VFIs have led to a reduction in government spending on S&T and delayed technological progress in various aspects, thus hindering the improvement of energy efficiency. In this context, we propose the following hypothesis:

• Hypothesis 2. Ceteris paribus, VFIs hinder energy efficiency improvements by reducing government spending on S&T.

Regional heterogeneity

Coastal areas have a higher level of regional development, a more reasonable institutional structure, and higher resource flexibility and can provide a good development environment for regional economic development (Tian et al. 2021). In contrast, inland areas are at a disadvantage in terms of natural endowments, and the widening gap in VFIs in these areas undoubtedly puts more pressure on local government finances. Weak institutions, such as market and legal protection in inland areas, intensify the fierce competition among local governments for limited fiscal resources (Hong and Hu 2017). In recent years, part of China’s production capacity has gradually shifted from coastal areas to inland areas, requiring the government to increase productive spending. Seeking economic development and political performance, local governments and officials in inland areas are more inclined to increase productive spending, squeezing government spending on S&T to a greater extent. The direct consequence is the lack of sufficient resources to support the development of scientific research. The field of S&T is difficult to develop, which also makes improving energy efficiency difficult. Unfortunately, the relocation of production capacity to inland areas also removed land for local governments’ use under the land finance system. The high-tech industry cannot obtain sufficient investment and land, which limits its long-term development and is extremely detrimental to improving energy efficiency. In general, inland areas have a worse development environment than coastal areas. The mediating effect of government spending on S&T should thus be more significant in inland areas. Thus, we propose the following hypothesis:

• Hypothesis 3. Ceteris paribus, the mediating effect of S&T spending is more significant in inland than in coastal regions.

Empirical strategy, variables, and data

Econometric model specification

To test Hypothesis 1, with reference to Wu et al. (2020) and Lin and Zhou (2021a), we constructed Eq. (1) to examine the effect of VFIs on energy efficiency.

$${\mathrm{EE}}_{\mathrm{i},\mathrm{t}}={\alpha }_0+{\alpha }_1{\mathrm{VFI}}_{\mathrm{i},\mathrm{t}}+{\alpha }_2{\mathrm{Control}}_{\mathrm{i},\mathrm{t}}+{\tau }_{\mathrm{i}}+{\mu }_t+{\epsilon }_{\mathrm{i},\mathrm{t},}$$ 1

where $\mathrm{EE}$ represents energy efficiency, $\mathrm{VFI}$ represents vertical fiscal imbalances, and $\mathrm{Control}$ represents control variables that may affect energy efficiency. ${\alpha }_0$, ${\alpha }_1$, and ${\alpha }_2$ represent the unknown regression coefficients; $\tau$ and $\mu$ represent the province fixed effect (province FE) and the year fixed effect (year FE), respectively; and $\epsilon$ represents the error term. In Eq. (1),$\mathrm{i}$ represents the province, and $\mathrm{t}$ represents the year.

$${\mathrm{GSST}}_{\mathrm{i},\mathrm{t}}={\gamma }_0+{\gamma }_1{\mathrm{VFI}}_{\mathrm{i},\mathrm{t}}+{\gamma }_2{\mathrm{Control}}_{\mathrm{i},\mathrm{t}}+{\tau }_{\mathrm{i}}+{\mu }_{\mathrm{t}}+{\epsilon }_{\mathrm{i},\mathrm{t},}$$ 2

$${\mathrm{EE}}_{\mathrm{i},\mathrm{t}}={\phi }_0+{\phi }_1{\mathrm{VFI}}_{\mathrm{i},\mathrm{t}}+{\phi }_2{{\mathrm{GSST}}_{\mathrm{i},\mathrm{t}}+{\phi }_3\mathrm{Control}}_{\mathrm{i},\mathrm{t}}+{\tau }_{\mathrm{i}}+{\mu }_{\mathrm{t}}+{\epsilon }_{\mathrm{i},\mathrm{t}}.$$ 3

To test Hypotheses 2 and 3, we must introduce a mediator variable into the model for testing. Drawing on the research method of Baron and Kenny (1986), we constructed Eqs. (2) and (3) on the basis of Eq. (1) to form a set of formulas. In Eqs. (2) and (3), $\mathrm{GSST}$ represents the mediating variable: government spending on S&T. ${\gamma }_0$-${\gamma }_2$ and ${\phi }_0$-${\phi }_3$ represent unknown regression coefficients. The other variables are defined as in Eq. (1). We are primarily interested in the significance of the regression coefficients ${\alpha }_1$, ${\gamma }_1$, ${\phi }_1$, and ${\phi }_2$ in Eqs. (1), (2), and (3) to determine whether government spending on S&T is the mediating channel of an effect of VFIs on energy efficiency.

Variable definitions

Measuring energy efficiency

Most current studies do not consider undesired outputs when measuring energy efficiency or consider only a single undesired output (Song et al. 2021). Referring to the measurement methods of Meng et al. (2016) and Song et al. (2021), we calculate the energy efficiency of 30 Chinese provinces through the slack-based model with undesirable outputs. The selection of variables in the model is the same as that in the research of Meng et al. (2016) and Song et al. (2021), but the undesired output variables are modified.

The details are as follows: (1) inputs: input variables include the capital stock, labor, and energy consumption. Among them, the capital stock is calculated with the perpetual inventory method. The labor force is the number of employees in each province at the end of the year. Energy consumption is the total energy consumption of each province (uniformly converted to standard coal). (2) Desirable output: the output variable is the constant price GDP of each province. (3) Undesirable outputs: these variables are CO₂ emissions, SO₂ emissions, soot emissions, and wastewater. The bad output brought about by energy consumption is multifaceted, including air and water pollution. Considering the availability of data, this study collects pollution data covering various aspects to avoid errors in data calculation, ultimately retaining CO₂ emissions, SO₂ emissions, soot emissions, and wastewater as undesired output variables.

Measuring VFIs

Drawing on the practice of Eyraud and Lusinyan (2013) and considering the institutional characteristics of China’s fiscal decentralization, we construct a VFI indicator to measure vertical fiscal imbalances as shown in Eq. (4):

$$\mathrm{VFI}=1-\frac{\mathrm{Fdr}}{\mathrm{Fde}}\left(1,-,\mathrm{Lbd}\right)$$ 4

where $\mathrm{Fde}$ and $\mathrm{Fdr}$ denote fiscal spending decentralization and fiscal revenue decentralization, respectively. This study draws on Lin and Zhou (2021b) and Liu and Zhang (2022b) to measure fiscal spending (revenue) decentralization. $\mathrm{Fde}$ ($\mathrm{Fdr}$) is equal to local per capita fiscal budget expenditure (revenue)/local per capita fiscal budget expenditure (revenue) and central per capita fiscal budget expenditure (revenue).$\mathrm{Lbd}$ is the local fiscal self-sufficiency rate, which is measured by the difference between local fiscal budget expenditure and local fiscal budget revenue/local fiscal budget expenditure. When the degree of asymmetry in $\mathrm{Fde}$ and $\mathrm{Fdr}$ is higher, income dispersion is lower, and expenditure dispersion is higher, and thus, VFIs are higher.

Measuring the mediator variables

For this analysis, the mediating variable is government spending on S&T (GSST). GSST is measured by per capita spending on science and technology in units of 10,000 yuan.

Measuring control variables

To reduce the bias in the estimation of the results, we consider other elements that may influence energy efficiency and select a series of control variables for the model. First, we control for the effect of regional economic development on energy efficiency with (the log of) GDP per capita (lnGDPPC). GDPPC is in yuan. Second, we control for the effect of foreign direct investment on energy efficiency with (the log of) total foreign direct investment (lnFDI). FDI is in 100 million yuan. Third, we control for the impact of the secondary industry on energy efficiency with (the log of) the added value of the secondary industry (lnIDU). IDU is in 100 million yuan. Finally, we control for the impact of local government intervention on energy efficiency with (the log of) local government spending (lnGOV). GOV is in 100 million yuan.

Data sources

The data in some studies are seriously lacking. Therefore, this study selects the panel data of 30 provinces in China (excluding Hong Kong, Macau, Taiwan, and Tibet) from 2001 to 2017 to test our hypotheses. Our panel data come from the Statistical Yearbook of China, Finance Yearbook of China, China Energy Statistical Yearbooks, National Bureau of Statistics of China, and EPS-China Database. Table 1 reports the results of the descriptive statistical analysis for each variable. The maximum, minimum, average, and other data displayed in the report reveal that VFI is relatively high overall but the overall level of EE is low. Likewise, we find that the overall level of GSST is lower. In a simple comparison, a contrast is found between the two, providing a better basis for our research. The descriptive statistical analysis results for the control variables are consistent with those in other studies. We perform a variance inflation factor (VIF) test to avoid biased estimates caused by multicollinearity. Upon inspection, we find that the VIFs have a maximum value of 3.51, indicating no multicollinearity.

Table 1.

Descriptive statistics

Var	Obs	Mean	Std. Dev	Min	Median	Max
EE	510	0.382	0.246	0.176	0.273	1.000
VFI	510	0.736	0.224	0.009	0.813	0.999
GSST	510	0.017	0.025	0.001	0.009	0.167
GDPPC	510	30,479.020	23,125.370	3000.000	25,690.500	136,172.000
FDI	510	371.808	442.786	1.033	193.498	2234.973
IDU	510	6056.862	6698.259	96.090	3833.530	39,124.110
GOV	510	2713.217	2474.579	90.930	1922.685	16,118.430

Empirical results

Baseline regression results

Table 2 shows the baseline regression results. Columns [1] and [2] express the results of the influence of VFIs on energy efficiency. The difference between the two columns is that more control variables are added to column [2]. The explained variable is energy efficiency in both columns [1] and [2]. The regression coefficients for VFI in column [2] are higher than those in column [1], the T values are larger, and the overall goodness of fit and f values are also higher. The results of column [2] can better explain the influence of VFIs on energy efficiency.

Table 2.

Baseline regression results

	[1]	[2]
	EE	EE
VFI	− 0.126** (− 2.25)	− 0.196*** (− 3.31)
lnGDPPC		− 0.331*** (− 4.11)
lnFDI		− 0.024*** (− 2.93)
lnIDU		0.066 (1.14)
lnGOV		0.010*** (3.93)
Province FE	Yes	Yes
Year FE	Yes	Yes
Constant	0.668*** (13.05)	3.242*** (7.75)
R2	0.278	0.381
F test	10.51***	13.47***
Observations	510	510

*, **, and *** denote significance at 10%, 5%, and 1%, respectively; t statistics in parentheses

In the presented results (columns [1] and [2]), the regression coefficient of VFI is significantly negative, which means that VFIs influence energy efficiency in a significantly negative way. This result shows that VFIs significantly inhibit energy efficiency, validating Hypothesis 1. The reason may be that under higher VFIs, the mismatch between revenue and expenditure leads to greater fiscal pressure on local governments. To alleviate this pressure, local governments tend to invest in productive expenditures (Jia et al. 2014), boosting extensive industrial development. In the process, they ignore the importance of energy conservation, thus greatly inhibiting energy efficiency. Lin and Zhou (2021a) argue that when VFIs are high, local governments place more emphasis on economic development than on environmental protection. Therefore, local governments lack incentives to save energy and reduce emissions to the detriment of energy efficiency. This argument is consistent with our findings.

Next, we focus on the results for the control variables. Among the control variables, three are significant. First, the negative influence of regional economic development (lnGDPPC) on energy efficiency is significant. When the economy starts to develop, regional economic development is often accompanied by environmental pollution, that is, a decline in energy efficiency. This result is in line with the first stage of the environmental Kuznets curve, possibly because of the relatively short study period (Wang and Wang 2020). From China’s actual economic development situation, we find that regional economic development depends on extensive industries and that many problems, such as unreasonable energy structure and energy waste, are damaging to energy efficiency. Second, the negative coefficient on lnFDI with respect to energy efficiency is significant, confirming the “pollution paradise” hypothesis. From a global perspective, most industries transferred from developed countries to China are characterized by low added value, high energy consumption, and labor intensiveness. Most foreign capital also invests in and promotes these industries, inevitably causing environmental pollution, which negatively impacts energy efficiency. Finally, the positive influence of local government intervention (lnGOV) on energy efficiency is significant, which may be because local governments have implemented macro-control over related polluting industries through administrative means to control current energy consumption to achieve assessment targets (Shi and Shen, 2013).

Robustness checks

This section examines the robustness of the regression results through a series of methods. Table 3 reports the robustness test results. First, we used GMM to alleviate the endogeneity problem. Column [1] results show that the regression coefficient on VFI is significantly negative, meaning that the results are robust. Second, we use the one-period lag of VFI as the explained variable and conduct the regression again to alleviate endogeneity concerns, and the regression result (column [2]) is still robust. Third, we replace the measure of the explained variable. Referring to the measurement methods of Meng et al. (2016) and Song et al. (2021), we retain only CO₂ emissions as undesired outputs, and energy efficiency is remeasured. The results in column [3] show that the regression coefficient on VFI is significantly negative, meaning that the results are still consistent under the measurement that considers only CO₂ emissions. Fourth, we choose another method to measure VFI. Referring to Feng et al. (2022), we change the calculation of the gap ratio of fiscal revenue and expenditure, using national instead of local standards. The regression coefficient of VFI in column [4] remains significantly negative. This result shows that changing the way VFI is measured does not change the original conclusion. Fifth, to rule out business cycle effects (Baskaran 2010), we average all variables over a 3-year period. After processing, the original balanced panel data become smoother, and the study object is reduced to 180 observations. We perform the regression again; the results are shown in column [5], again supporting the above conclusion. Finally, to rule out the possibility of bias from the choice of estimation methods, we choose ordinary least squares (OLS) and random effects (RE) for regression. We find that the regression results (columns [6]–[7]) remain unchanged. After multiple robustness checks, the results do not change. We argue that our baseline view is sound and that VFIs significantly inhibit improvements to energy efficiency.

Table 3.

Robustness test results

	[1]	[2]	[3]	[4]	[5]	[6]	[7]
	EE	EE	EE	EE	EE	EE	EE
VFI	− 0.259* (− 1.92)	− 0.107** (− 2.21)	− 0.180*** (− 3.04)	− 0.244*** (− 3.24)	− 0.254** (− 2.39)	− 0.302*** (− 4.72)	− 0.307*** (− 5.55)
lnGDPPC	0.302** (2.27)	− 0.418*** (− 6.00)	− 0.324*** (− 4.02)	-0.324*** (− 4.01)	− 0.373*** (− 2.92)	0.167*** (5.85)	0.064 (1.56)
lnFDI	− 0.020 (− 0.78)	− 0.014** (− 1.99)	− 0.031*** (− 3.67)	− 0.025*** (− 3.03)	− 0.021 (− 1.52)	0.029*** (3.64)	− 0.017** (− 2.09)
lnIDU	− 0.226*** (− 4.46)	0.130** (2.60)	0.053 (0.92)	0.056 (0.97)	0.083 (0.92)	− 0.248*** (− 22.20)	− 0.177*** (− 7.28)
lnGOV	0.005* (1.65)	0.005*** (2.20)	0.012*** (4.47)	0.014*** (4.21)	0.012*** (2.78)	0.003 (0.87)	0.012*** (4.57)
Province FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Constant	− 0.433 (− 0.37)	3.982*** (9.35)	3.301*** (7.89)	3.244*** (7.75)	3.430*** (5.17)	0.913*** (2.96)	1.436*** (4.63)
R2	-	0.316	0.394	0.381	0.389	0.651	0.344
F test/Wald chi2	11,532.37***	9.91***	14.23***	13.44***	8.91***	43.33***	294.63***
Observations	510	480	510	510	180	510	510

*, **, and *** denote significance at 10%, 5%, and 1%, respectively; t statistics in parentheses

Mechanism analysis

Next, following Eqs. (1)–(3), we explore the mediating mechanism by which VFIs inhibit the improvement of energy efficiency. Table 4 reports the test results on the mediation mechanism. Column [1] in Table 4 replicates column [2] of Table 2, representing the baseline conclusion, and the result is highly significant. The premise of testing the intermediary mechanism is that VFIs significantly hinder energy efficiency. We find that the regression coefficient of VFI is significantly negative (${\alpha }_1$ = − 0.196***), and thus, the precondition holds. Next, in the results of column [2], the regression coefficient of VFI is significantly negative (${\gamma }_1$ = − 0.015***), indicating that VFI significantly inhibits government spending on S&T, which confirms the results of Liu and Zhang (2021). They conclude that VFIs significantly disincentivize government spending on S&T. With high VFIs, local governments invest their limited fiscal resources in infrastructure development to boost GDP. This situation has weakened local government support for innovative activities. The results of column [3] show that the regression coefficient of VFI is significantly negative (${\phi }_1$ = − 0.136***). By comparison, ${|\phi }_1|$ is less than ${|\alpha }_1|$. In turn, the regression coefficient of government spending on S&T is significantly positive (${\phi }_2$ = 4.025***), which indicates that government spending on S&T can substantially improve energy efficiency. The above test results meet the three conditions proposed by Baron and Kenny (1986). Thus, government spending on S&T (GSST) must be the mediating mechanism whereby VFIs inhibit energy efficiency, in a partial mediating effect. This conclusion provides strong evidence for Hypothesis 2. Under higher VFIs, to compensate for the enormous gap between fiscal revenue and expenditure, local government fiscal expenditure behavior becomes distorted, increasing more than production and constructive expenditure, thereby squeezing government spending on S&T (Liu and Zhang 2021). Especially in the context of China’s political tournament, local government spending on S&T cannot achieve greater economic benefits than productive expenditures in the short term (Tang et al. 2021), giving rise to local governments’ preference for “focusing on production and ignoring innovation” in spending. In the process, local governments have reduced spending on technology, ignoring the importance of improving energy efficiency. Furthermore, referring to Mackinnon et al. (1995), we find that the share of the mediating effect ${\phi }_2{\gamma }_1/({\phi }_2{\gamma }_1+{\phi }_1)$ is 30.745%, showing that government spending on S&T is one of the main channels through which VFIs significantly inhibit improvements to energy efficiency. Lin and Zhou (2021a) also explore how VFIs affect energy efficiency. In their study, they find that industrial structure upgrading, technological innovation, and government intervention are important mechanisms through which VFIs affect energy efficiency. However, in their study, they ignore the important role of government spending on S&T; their research is thus complemented by the findings of this study. Finally, to ensure that the results are robust, we rerun the test as in steps 3 and 4 in the “Robustness checks” section. The results (columns [4]–[9] in Table 4) still support Hypothesis 2.

Table 4.

Mechanism test results

	[1]	[2]	[3]	[4]	[5]	[6]	[7]	[8]	[9]
	EE	GSST	EE	EE	GSST	EE	EE	GSST	EE
VFI	− 0.196*** (− 3.31)	− 0.015** (− 2.46)	− 0.136** (− 2.50)	− 0.180*** (− 3.04)	− 0.015** (− 2.46)	− 0.121** (− 2.22)	− 0.244*** (− 3.24)	− 0.027*** (− 3.60)	− 0.134* (− 1.92)
GSST			4.025*** (9.62)			3.959*** (9.43)			4.010*** (9.49)
lnGDPPC	− 0.331*** (− 4.11)	− 0.051*** (− 6.25)	− 0.124 (− 1.62)	− 0.324*** (− 4.02)	− 0.051*** (− 6.25)	− 0.121 (− 1.57)	− 0.324*** (− 4.01)	− 0.050*** (− 6.07)	− 0.125 (− 1.63)
lnFDI	− 0.024*** (− 2.93)	0.001* (1.66)	− 0.030*** (− 3.95)	− 0.031*** (− 3.67)	0.001* (1.66)	− 0.036*** (− 4.72)	− 0.025*** (− 3.03)	0.001 (1.50)	− 0.030*** (− 3.97)
lnIDU	0.066 (1.14)	0.022*** (3.78)	− 0.024 (− 0.44)	0.053 (0.92)	0.022*** (3.78)	− 0.035 (− 0.65)	0.056 (0.97)	0.021*** (3.56)	− 0.028 (− 0.51)
lnGOV	0.010*** (3.93)	0.001*** (3.39)	0.007*** (2.75)	0.012*** (4.47)	0.001*** (3.39)	0.008*** (3.34)	0.014*** (4.21)	0.001*** (4.37)	0.008*** (2.61)
Province FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Constant	3.242*** (7.75)	0.314*** (7.36)	1.979*** (4.90)	3.301*** (7.89)	0.314*** (7.36)	2.058*** (5.08)	3.244*** (7.75)	0.313*** (7.38)	1.990*** (4.91)
R2	0.381	0.498	0.485	0.394	0.498	0.493	0.381	0.506	0.483
F test	13.47***	21.71***	19.64***	14.23***	21.71***	20.23***	13.44***	22.35***	19.41***
Observations	510	510	510	510	510	510	510	510	510

*, **, and *** denote significance at 10%, 5%, and 1%, respectively; t statistics in parentheses

Heterogeneity analysis

We further capture the regional heterogeneity in the mediation mechanism. Table 5 presents the final results. We analyze regional heterogeneity between inland and coastal regions. Columns [1]–[3] show the estimation results for inland areas. Columns [4]–[6] show the estimation results for coastal areas. The results in columns [1] and [4] show that the regression coefficient of VFI is significantly negative, suggesting that VFIs significantly inhibit energy efficiency in inland regions as in coastal regions and once again verifying Hypothesis 1. From the regression coefficient of VFI, the effects in inland areas are smaller than those in coastal regions. A gap is also found between the T values of the two, suggesting that inland regions are more affected by VFI than coastal regions, which may be due to the poor resource endowment of local governments in inland regions and relatively insufficient momentum in their economic development. VFIs impose greater fiscal pressure at high levels.

Table 5.

Heterogeneity test results

	[1]	[2]	[3]	[4]	[5]	[6]
	Inland area			Coastal area
	EE	GSST	EE	EE	GSST	EE
VFI	− 0.345*** (− 3.71)	− 0.052*** (− 4.57)	− 0.152* (− 1.77)	− 0.225* (− 1.95)	− 0.016 (− 1.54)	− 0.147 (− 1.41)
GSST			3.690*** (8.53)			4.918*** (6.08)
lnGDPPC	− 0.132* (− 1.77)	− 0.063*** (− 6.88)	0.101 (1.40)	− 0.616*** (− 2.90)	− 0.042** (− 2.23)	− 0.407** (− 2.09)
lnFDI	− 0.029*** (− 4.20)	0.001 (0.98)	− 0.033*** (− 5.19)	− 0.035 (− 1.16)	0.002 (0.88)	− 0.046* (− 1.71)
lnIDU	− 0.025 (− 0.51)	0.023*** (3.93)	− 0.111** (− 2.50)	0.179 (1.03)	0.032** (2.04)	0.024 (0.15)
lnGOV	0.012*** (4.51)	0.001*** (3.39)	0.008*** (3.27)	0.011* (1.88)	0.002*** (3.01)	0.003 (0.59)
Province FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Constant	2.158*** (5.05)	0.440*** (8.39)	0.535 (1.26)	5.280*** (4.73)	0.159 (1.59)	4.500*** (4.44)
R2	0.452	0.530	0.564	0.437	0.550	0.546
F test	11.10***	15.20***	16.59***	5.72***	9.01***	8.41***
Observations	323	323	323	187	187	187

*, **, and *** denote significance at 10%, 5%, and 1%, respectively; t statistics in parentheses

We further analyze whether there is regional heterogeneity in the role of government spending on S&T as an intermediary mechanism of the effect of VFIs on energy efficiency. According to the overall regression results, the regression coefficient of VFI in column [2] is significantly negative (${\gamma }_1$ = − 0.052***), indicating that in the inland sample, VFIs significantly inhibit government spending on S&T. The regression coefficient of VFI in column [3] is significantly negative (${\phi }_1$ = − 0.152***), and ${|\phi }_1|$ is smaller than ${|\alpha }_1|$. The regression coefficient of government spending on S&T is significantly positive (${\phi }_2$ = 3.690***), which indicates that government spending on S&T can significantly improve energy efficiency in the sample of inland areas. According to the conditions of Baron and Kenny (1986), government spending on S&T (GSST) remains the mediating mechanism whereby VFIs suppress energy efficiency in the inland sample, and this conclusion still holds. However, in the regression results for coastal areas, the regression coefficients of VFI in columns [5] and [6] are negative but not significant and do not meet the three conditions proposed by Baron and Kenny (1986). Therefore, the mediating role is not obvious in coastal areas, in contrast to inland areas, indicating regional heterogeneity in the role of government spending on S&T as the mediating mechanism in the relationship between VFIs and energy efficiency and thus providing sufficient evidence to establish Hypothesis 3. Under a high degree of VFIs, the mismatch between fiscal revenue and expenditure has caused inland areas to bear greater fiscal pressure. Coupled with factors such as the influx of production capacity and insufficient resource endowments, local governments in inland areas need to integrate most of their local fiscal resources to develop industries that have moved into the local area. These industries are often associated with high energy consumption and high pollution, greatly reducing energy efficiency. However, to mitigate the effects of VFIs and promote local economic development, they must increase infrastructure spending and squeeze government spending on S&T (Liu and Zhang 2021), thereby ignoring the importance of energy efficiency. Coastal areas have a more rationalized institutional structure, favorable resource endowment conditions, and a favorable development environment (Tian et al. 2021). Local governments in coastal areas have sufficient resource mobility to promote balanced development. This is confirmed by the findings of Liu and Zhang (2021). Coastal regions are relatively economically developed and face a relatively small gap between fiscal revenues and expenditures. The level of government spending on S&T may not be related to VFIs in coastal areas, in contrast to in inland areas. Therefore, coastal areas are thus less susceptible to the negative effects of VFIs.

Conclusion and policy implications

Energy is necessary for global economic and social development. Energy is distributed differently around the world and is characterized by scarcity. Therefore, how to achieve efficient use of energy has become a topic of common interest worldwide. For China, improving energy efficiency is a necessary condition for sustainable development. Although China has the advantage of abundant natural resources, its energy utilization efficiency is low. Energy inefficiency also leads to environmental pollution, which needs to be solved urgently. To this end, the Chinese government has also engaged in many endeavors and introduced several measures. However, improving energy efficiency requires cooperation between upper and lower levels of government. Given the implementation of the tax-sharing system reform in 1994, local governments have been subject to increasing gaps between fiscal revenue and expenditure, leading to the phenomenon of VFIs. VFIs distort incentives for local government fiscal behavior (Jia et al. 2021), promote extensive industrial growth, violate sustainable development goals, and lead to serious environmental pollution and reduced energy efficiency. This context provides a good opportunity for this study to examine the impact of VFIs on energy efficiency. We systematically examine this impact from the perspective of government spending on S&T. Furthermore, the regional heterogeneity in the mediation mechanism is explored. This study uses panel data from 30 provinces in China from 2001 to 2017 to test several hypotheses. Several findings are obtained: (1) VFIs have a significant negative influence on energy efficiency. (2) Government spending on S&T exerts a mediating influence in the relationship between VFIs and energy efficiency. Specifically, VFIs negatively affect government spending on S&T, whereas government spending on S&T positively affects energy efficiency. (3) These results show significant regional heterogeneity, being more significant in the inland than in the coastal region sample.

The above research conclusions are instructive. Thus, this study summarizes some policy suggestions. First, the government should consider optimizing the existing fiscal system for energy conservation and emission reduction to improve energy efficiency. The level of revenue can be increased by appropriately expanding the taxation powers of local governments. In terms of fiscal expenditure, the central government should also take on some of the expenditure on public goods and reduce the fiscal expenditure responsibilities of local governments. This will reduce the fiscal pressure on them. In the meantime, the government should guide and weaken the competition pattern of the political tournament and change the GDP-oriented assessment system. In the new assessment system, the government should include energy efficiency and environmental pollution as the key inspection contents, which would be conducive to actively guiding a sustainable economic development model. This initiative would help to reduce the negative impact of VFIs on energy efficiency.

Second, the government should avoid the negative impact of VFIs on fiscal spending in the area of S&T. The government should strengthen the management of the regulation of the use of fiscal resources. It can establish a sound regulatory system to guarantee fiscal expenditure in S&T. With guaranteed funding in the field of S&T, technological innovation activities can be promoted in an orderly manner, which would be conducive to mitigating the loss of energy efficiency caused by VFIs or even correcting the negative impact of VFIs on energy efficiency.

Finally, governments at all levels should be aware of the differences in the impact of VFIs in different regions. The central government should focus on the adverse effects of VFIs in the inland and introduce targeted policies, such as assuming more spending on public goods for local governments in these regions.

This study complements the empirical research on the relationship between VFIs and energy efficiency from a new perspective but has some limitations. First, given data acquisition limitations, we studied relevant issues only at the provincial level and did not deeply consider the prefecture or county levels. Second, given the difficulty of measuring enterprise energy efficiency, this study cannot examine this variable in depth from the perspective of microenterprises. Third, we are unable to obtain data on total social fixed asset investment and total new fixed asset investment for each province after 2017. This results in our inability to calculate energy efficiency beyond 2017. Fourth, given that the data cover only the period up to 2017, we cannot examine the impact of exogenous shock events such as the COVID-19 pandemic, which is a shortcoming that needs to be expanded in follow-up research.

Author contribution

All authors contributed to the study’s conception and design. Data collection and analysis were performed by Tianchu Feng. The first draft of the manuscript was written by Tianchu Feng. Writing, review, and editing were done by Meijuan Liu, and Chaozhu Li. All authors read and approved the final manuscript.

Funding

This work was possible thanks to the National Social Science Foundation of China (17BGL127) and funding from the soft science project of Zhejiang Provincial Department of Science and Technology (2022C35066).

Data availability

Available upon request.

Declarations

Ethical approval

Not applicable.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.