The Bright Side of Uncertainty: The Impact of Climate Policy Uncertainty on Urban Green Total Factor Energy Efficiency

Abstract

Climate change has a significant impact on human economic and social life, and climate issues have rapidly emerged as a global hot topic. Using data from prefecture-level cities in China from 2005 to 2020 as a sample, this study explores the relationship between climate policy uncertainty (CPU) and urban green total factor energy efficiency (GTFEE). The results show that, first, the CPU can significantly improve urban GTFEE, and results are reconfirmed after various robustness tests. Second, the CPU promotes urban GTFEE by improving public environmental concerns and optimizing the energy consumption structure. Third, the promotion role of CPU in urban GTFEE is particularly significant in resource-based cities and economically developed cities. The results of this study provide a theoretical basis and practical enlightenment for the government to formulate forward-looking climate policies and promote the transformation of green development in cities.

1. Introduction

In March 2024, the World Meteorological Organization released the Global Climate in 2023 report. The report pointed out that the levels of greenhouse gases, surface temperature, ocean heat, and acidification on record have been broken again, even showing a sharp change in trend. Frequent climate change could have both physical and transitional impacts on economic activity, the former being direct impacts due to climate risk events, and the latter being indirect impacts due to governance policies adopted by governments [1]. In the context of the complicated evolution of climate risks, the complexity of climate and environmental governance is also increasing [2]. As an important means of coping with climate change and an important factor influencing economic activity, the effects of climate policy uncertainty deserve to be further explored.

In the face of increasing climate risks, many countries have adopted a series of active measures to save energy and reduce emissions, and energy efficiency is one of the important indicators to measure the effectiveness of these environmental protection measures [3]. Moreover, improving energy efficiency not only helps to alleviate the problem of energy shortage but also effectively reduces ecological pressure and climate risks, which is an important measure to achieve sustainable development and address climate change [4]. According to the existing research results, the influencing factors of energy efficiency can be divided into macro and micro categories. The macro level includes environmental legislation [5], urban architectural design [6], import trade [7], and other factors. The micro level includes factors such as the digital transformation of enterprises [8] and investment efficiency [9]. It can be seen that energy efficiency is a hot topic, and its influencing factors are worthy of further exploration.

Although existing studies have conducted multidimensional drivers of energy efficiency, few studies have examined the impact of climate policy uncertainty on urban energy efficiency. Enterprises occupy a dominant position in the urban energy use system, and the consumption of traditional energy, such as coal, is one of the biggest sources of urban air pollution. In the face of climate policy uncertainty, enterprises tend to actively take environmental measures, such as green innovation [10] and increasing renewable energy consumption [11,12], thus improving urban green total factor energy efficiency. At the same time, policy uncertainty also increases public awareness of environmental protection [13], strengthens government supervision and punishment of corporate pollution [14], and forces enterprises to improve energy efficiency. In addition, the CPU also promotes the use of clean energy, optimizes the energy consumption structure [15], and further promotes the improvement of GTFEE in cities.

Therefore, this study explores the impact of climate policy uncertainty (CPU) on urban GTFEE and investigates the influencing mechanism between them. First, we apply a two-way fixed panel model to explore the effect of the CPU index on urban GTFEE. Secondly, this paper uses a mediating effect model to explore the impact mechanism of CPU on urban GTFEE through two possible channels: increasing public environmental concern and optimizing urban energy consumption structure. Finally, we explore the heterogeneity effect of climate policy uncertainty on the improvement of green total factor energy efficiency in cities with different development degrees and different resource types.

The possible marginal contribution of this study is as follows: First, we place CPU and urban GTFEE under the same research framework, expanding the current research on urban energy efficiency and complementing the existing research at the theoretical and empirical levels; Second, we explore the transmission mechanism between public environmental concern and urban energy consumption structure and expand the transmission path analysis of the CPU promoting urban energy efficiency; Finally, we conduct a heterogeneity analysis from two perspectives of urban development degree and resource-based cities and analyze the difference in promotion effect in cities with different economic conditions and economic structures.

The remaining parts of this paper are organized as follows: the second part is the literature review; the third part is the theoretical analysis and research hypothesis; the fourth part is the research design; the fifth part is the empirical results; the sixth part is heterogeneity analysis; and the seventh part is the conclusion and policy implications.

2. Literature Review

The existing research has explored the issues of climate policy uncertainty and energy efficiency extensively, and the literature closely related to this study focuses on two main areas: firstly, exploring the economic impacts of climate policy uncertainty; secondly, studying the factors influencing urban energy efficiency.

Existing studies have not reached a consensus on the economic impact of CPU, which is mainly divided into positive and negative effects. For one thing, the positive effects of CPU are mainly reflected by increasing the use of clean energy [11], promoting green innovation [16], increasing government research and development subsidies [10], encouraging enterprises to invest across regions [17], reducing environmental management costs [18], and reducing carbon emissions [19]. For example, Li et al. (2024) [16] argue that climate policy uncertainty can promote green innovation activities of Chinese agricultural enterprises. Sun et al. (2024) [17] propose that CPU makes it more likely for companies to decentralize their business activities to different regions to diversify their exposure to climate risks.

On the contrary, other scholars argue that the CPU has negative effects. CPU can not only cause dramatic fluctuations in clean energy markets [20], financial markets [21], and carbon markets [22] but also reduce productivity [23] and respond to climate change slowly [24]. For example, Wang et al. (2023) [25] believe that frequent climate policy fluctuations increase instability in carbon markets, green bond markets, and energy price markets, while Ren et al. (2024) [26] propose that climate policy fluctuations inhibit the financialization process of enterprises. In summary, the economic impact of climate uncertainty risk is complex, which may bring both market fluctuations and potential losses and opportunities for sustainable development, and its effects are worthy of in-depth study.

Another type of literature related to this paper is research on the drivers of urban energy efficiency. From the macro level, the factors influencing urban energy efficiency are as follows: industrial structure [27], technological progress [28], low carbon policy [29], For example, Gao et al. (2023) [29] verified that low-carbon policies can alleviate the problem of labor and capital mismatch, improving the GTFEE of cities. At the micro level, digital transformation at the enterprise level [8], import expansion [7], digital technology innovation [30], labor productivity [31], and green innovation elasticity. Wu et al. (2024) [32] also have a significant impact on energy efficiency. At present, the existing research on urban energy efficiency is relatively abundant, but research on urban energy efficiency from the perspective of climate policy needs to be further expanded.

In summary, the current research on CPU and urban energy efficiency has been more extensive, but there are still the following deficiencies: Firstly, most of the existing research on the macro-effects of climate policy uncertainty focuses on the market level, and there is a lack of investigation on the effect of the urban cities; Secondly, few scholars have conducted in-depth research on the factors affecting urban energy efficiency from the perspective of climate policy, and it is particularly important to clarify the effect of CPU on urban GTFEE in the context of climate risk complexity.

3. Theoretical Analysis and Research Hypothesis

This study argues that climate policy uncertainty can significantly improve the energy efficiency of cities as show in Figure 1. The reasons are as follows: it is difficult for enterprises to accurately predict how the government intervenes in economic behavior [2], and such policy uncertainty may improve the enterprises’ sense. Therefore, when climate risk intensifies, enterprises may be threatened from two aspects. On the one hand, direct operating costs resulting from climate risk events [33]. According to the precautionary savings theory, in the face of potential climate risk events, enterprises may manage their funds more carefully to ensure the robustness of their business activities [34]. On the other hand, it comes from the signal released by increasing climate policy uncertainty, which encourages enterprises to increase their R&D investments to cope with potential risks [35]. According to the real options investment theory, innovation investment of enterprises is not only to pursue short-term benefits but also to accumulate long-term value, which can be regarded as a kind of option investment [17]. When enterprise managers feel the increasing uncertainty of climate policy, they are more inclined to choose investment projects with sustainable development potential [36], thus promoting the improvement of the city’s GTFEE. In addition, the government needs to actively respond and compensate for complex climate risks and adjust climate policies, which could increase the construction of urban green infrastructure [37], reduce urban pollution emissions, and improve the GTFEE of the city. Therefore, we propose the following:

Hypothesis 1:

Climate policy uncertainty can significantly improve the urban GTFEE.

Firstly, when governments frequently adjust their climate policies, the media and the public tend to worry about uncertainty, leading them to focus more on issues such as environmental protection and pollution control, and actively urge the government to implement environmental governance through public expression [38]. This trend not only increases the public’s attention to environmental issues but also promotes the government’s positive response in formulating and implementing environmental policies. Secondly, with the continuous improvement of environmental awareness, the public has made higher demands for the environmental behavior of local governments and enterprises. This inhibits the possible collusion between local governments and enterprises to a certain extent and enables regulators to be more strict and effective in supervising the pollution emission behavior of enterprises [39]. This strict regulatory environment has forced enterprises to pay more attention to environmental protection issues, and actively adopt green transformation measures to reduce pollution emissions and improve GTFEE in cities. Therefore, we propose the following:

Hypothesis 2:

Climate policy uncertainty improves the urban GTFEE by increasing public environmental concern.

Uncertainty makes it difficult for enterprises to predict the environmental protection requirements and standards they may face in the future. Therefore, to avoid potential legal risks and market pressures, enterprises have to strengthen internal environmental governance in advance to adapt to the changing policy environment and ensure their sustainable development [40]. Strengthening the internal environmental governance of enterprises promotes the transformation of traditional energy to clean energy, optimizes the energy consumption structure of the city, and then improves the green all-factor energy efficiency of the city. Specifically, on the one hand, strong internal environmental governance can enable the market to eliminate projects with low resource utilization and production efficiency [41], ensure that the remaining investment projects have good environmental performance and efficient production capacity, optimize the urban energy consumption structure, and thus promote the improvement of green total factor energy efficiency. On the other hand, strong internal environmental governance may increase enterprise costs in the short term, but in the medium and long term, green technology innovation promoted by environmental regulation will lead to the improvement of production efficiency and clean energy utilization rate [42], as well as the reduction of pollution emission. Thus, the energy consumption structure of the city can be optimized and the green total factor energy of the city can be improved. Therefore, this paper proposes a third hypothesis:

Hypothesis 3:

Climate policy uncertainty improves the urban GTFEE by optimizing the urban energy consumption structure.

There is considerable heterogeneity in the impact of climate policy uncertainty on the improvement of GTFEE in cities. First, the improvement of energy efficiency involves multiple links, such as energy production, conversion, transmission, and use, all of which require corresponding financial investment and technical support. According to the technological innovation theory, the more severe the financing constraints, the more difficult it is for firms to generate greater technological innovation [43]. Compared with developed cities, firms in developing cities undoubtedly face more severe financing constraints and face greater obstacles to green technological innovation activities [44]. In addition, developed cities tend to have higher levels of industrialization and are more likely to adopt clean energy and environmental technologies. Therefore, we argue that the sensitivity of GTFEE to climate policy uncertainty is heterogeneous across cities at different levels of development.

Second, China has a large number of resource-based cities, a quarter of which are at risk of resource depletion and urgently need to seek opportunities for economic transformation [45]. The economic activities of resource-based cities are often highly concentrated in one or a few resource industries and highly dependent on natural resources, which makes the economic structure of cities relatively fragile and will cause serious economic losses once climate risk events trigger market fluctuations [46]. In the macro context of sustainable development, compared with non-resource-based cities, climate policy uncertainty may make the public and enterprises in resource-based cities pay more attention to environmental issues [47] and show more enthusiasm in addressing climate risks and environmental problems. Therefore, we propose the following:

Hypothesis 4:

The improving effect of climate policy uncertainty on urban GTFEE is more obvious in developed and resource-based cities.

4. Research Design

4.1. Econometric Model

4.1.1. Baseline Model

To explore whether climate policy uncertainty could improve urban green total factor energy efficiency, following Gao et al. (2023) [29], we build a panel model of bidirectional fixed effects as shown below:

$${\mathrm{GTFEE}}_{\mathrm{it}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{CPU}}_{\mathrm{it}}+{\mathsf{\lambda }\mathrm{X}}_{\mathrm{it}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\gamma }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(1)

where ${\mathrm{GTFEE}}_{\mathrm{it}}$ is the explained variable, and subscripts i and t represent the data of the i city in the t period; ${\mathrm{CPU}}_{\mathrm{it}}$ is the core explanatory variable, ${\mathrm{X}}_{\mathrm{it}}$ is the control variable, ${\mathsf{\beta }}_0$ is the constant term, ${\mathsf{\beta }}_{\mathrm{i}}$ is the estimated coefficient of each variable, and ${\mathsf{\epsilon }}_{\mathrm{it}}$ is the random error. At the same time, we also added the terms ${\mathsf{\mu }}_{\mathrm{i}}$ and ${\mathsf{\gamma }}_{\mathrm{t}}$, respectively.

4.1.2. Mediating Effect Model

To further explore the mechanism of climate policy uncertainty on urban green total factor energy efficiency, we adopt a three-step approach to explore the mediating effect of public environmental concern (Pc) and urban energy consumption structure (Es). Following Yi et al. (2023) [45], this paper constructs the mediating effect model as follows:

$${\mathrm{M}}_{\mathrm{it}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{CPU}}_{\mathrm{it}}+{\mathsf{\alpha }}_2{\mathrm{X}}_{\mathrm{it}}+{\mathsf{\mu }}_{2\mathrm{i}}+{\mathsf{\gamma }}_{2\mathrm{t}}+{\mathsf{\epsilon }}_{2\mathrm{it}}$$

(2)

$${\mathrm{GTFEE}}_{\mathrm{it}}={\mathsf{\rho }}_0+{\mathsf{\rho }}_1{\mathrm{CPU}}_{\mathrm{it}}+{\mathsf{\rho }}_2{\mathrm{M}}_{\mathrm{it}}+{\mathsf{\rho }}_3{\mathrm{X}}_{\mathrm{it}}+{\mathsf{\mu }}_{3\mathrm{i}}+{\mathsf{\gamma }}_{3\mathrm{t}}+{\mathsf{\epsilon }}_{3\mathrm{it}}$$

(3)

In the above equation, the explained variable is GTFEE, the explanatory variable is CPU, and M represents the two mediating variables. According to the rules of the mediating effect model, the potential mechanisms are verified when the coefficients ${\mathsf{\beta }}_1, {\mathsf{\alpha }}_1$, ${\mathsf{\rho }}_2$ are significant.

4.2. Variables Description

4.2.1. Core Explanatory Variable

Climate policy uncertainty is the main explanatory variable, which refers to the disordered changes in climate policies and is specifically manifested in the uncertainty of the policymaking body, the time and content of policy promulgations, the role and implementation effects of policies [48]. They can be caused by political factors (such as the United Nations Climate Change Conference), economic factors (such as economic booms or recessions), and major climate events (such as extreme weather) [49]. Referring to Ma et al. (2023) [50], this study uses manual collection, processing, and the deep learning algorithm MacBERT model to construct a climate policy uncertainty index from 253 cities in China from 2005 to 2020, based on 1,755,826 articles from six mainstream newspapers, namely, the People’s Daily, Guangming Daily, the Economic Daily, the Global Times, Science and Technology Daily, and the China News Service. The CPU Index of Chinese cities from 2005 to 2020 is constructed through five steps: data collection and cleaning, manual review, construction of training models, index calculation and standardization, and technical validation.

4.2.2. Explanatory Variable

Referring to Gao et al. (2023) [29], we use the SBM method with non-desired outputs to measure the GTFEE at the city level from 2005 to 2020. The input indicators are capital, labor, and urban energy consumption, and the desired output indicator is urban GDP, while the non-desired output indicator includes three types of pollutants: SO₂, exhaust gas, and wastewater. We apply the linear programming procedure of MATLAB2018a to calculate the urban GTFEE index. The specific linear programming equation is as follows:

$$\min \mathsf{\rho }=\frac{1-\left(1/\mathrm{m}\right){{\displaystyle \sum }}_{\mathrm{i}=1}^{\mathrm{m}}{\mathrm{s}}_{\mathrm{i}}^{-}/{\mathrm{x}}_{\mathrm{i}0}}{1+\frac{1}{{\mathrm{s}}_1+{\mathrm{s}}_2}\left({{\displaystyle \sum }}_{\mathrm{r}=1}^{{\mathrm{s}}_1}{\mathrm{s}}_{\mathrm{r}}^{\mathrm{d}}/{\mathrm{y}}_{\mathrm{r}0}^{\mathrm{d}}+{{\displaystyle \sum }}_{\mathrm{t}=1}^{{\mathrm{s}}_2}{\mathrm{s}}_{\mathrm{t}}^{\mathrm{ud}}/{\mathrm{y}}_{\mathrm{t}0}^{\mathrm{ud}}\right)}$$

(4)

$$\left\{\begin{array}{c}{\mathrm{x}}_0=X\lambda +{\mathrm{s}}^{-}\\ {\mathrm{y}}_0^{\mathrm{d}}={\mathsf{\gamma }}^{\mathrm{d}}\lambda -{\mathrm{s}}^{\mathrm{d}}\\ \begin{array}{c}{\mathrm{y}}_0^{\mathrm{ud}}={\mathsf{\gamma }}^{\mathrm{ud}}\mathsf{\lambda }-{\mathrm{s}}^{\mathrm{ud}}\\ \mathsf{\lambda }\ge 0,{\mathrm{s}}^{-}\ge 0,{\mathrm{s}}^{\mathrm{ud}}\ge 0,{\mathrm{s}}^{\mathrm{d}}\ge 0\end{array}\end{array}\right.$$

(5)

The model assumes that there are n decision units, each of which has m inputs, ${\mathrm{s}}_1$ expected outputs and ${\mathrm{s}}_2$ unexpected outputs. $\mathrm{x}\in {\mathrm{R}}^{\mathrm{m}}$ represents an input vector, ${\mathrm{y}}^{\mathrm{d}}\in {\mathrm{R}}^{{\mathrm{s}}_1}$ represents a desired output vector, and ${\mathrm{y}}^{\mathrm{ud}}\in {\mathrm{R}}^{{\mathrm{s}}_2}$ represents an undesired output vector. The vector ${\mathrm{s}}^{\mathrm{d}}$ represents the deficiency of the desired output and the vectors ${\mathrm{s}}^{-}$ and ${\mathrm{s}}^{\mathrm{ud}}$ represent the redundancy of the input and undesired outputs, respectively. The objective function β is in the interval [0, 1], and the decision unit SBM is valid if and only if $\mathsf{\beta }=1,{\mathrm{s}}^{-}={\mathrm{s}}^{\mathrm{d}}={\mathrm{s}}^{\mathrm{ud}}=0$. If β < 1, it indicates that the non-DMU is effective, and the ratio of input to output needs to be further improved.

4.2.3. Mediating Variables

(1)

(Public Environmental Concern (Pc): In previous studies, there are usually two ways to measure public environmental concern. One is the number of environmental letters and visits, which is calculated by combining relevant data such as the number of proposals from the People’s Congress and the total number of letters from official statistics. However, the petitioning behavior itself has problems of low efficiency and high cost, which makes it difficult to reflect the real attitude of the public towards the environment. Therefore, referring to Wang and Zhao (2018) [51], we adopt the Baidu search index as a proxy variable for the public’s environmental concerns. Specifically, we use the total daily search volume of environment-related terms, such as “smog” and “environmental governance”, as a proxy indicator of public environmental concern, which can briefly and directly reflect the degree of public concern about environmental issues;

(2)

Energy consumption structure (Es): Referring to Zeng et al. (2021) [52], we use the following formula to calculate the urban energy consumption structure:

$$\mathrm{Es}=\frac{\mathrm{D}+\mathrm{T}}{\mathrm{M}}$$

(6)

In the above equation, D is the total electricity consumption, T is the total electricity consumption of natural gas, and M is the total electricity consumption of coal. The energy consumption structure index Es can be obtained by substituting the three. According to the calculation method of the index, the larger the Es index is, the energy consumption type in the city changes from coal consumption to clean natural gas and electricity consumption, the better the energy consumption structure is, and the worse the vice versa is.

4.2.4. Other Control Variables

To control the factors affecting urban GTFEE as much as possible, we refer to the studies of Liu et al. (2020) [53] and Li et al. (2023) [54]. In this paper, the level of economic development (El), financial development (Fl), urbanization (Cl), fiscal decentralization (Cfl), and infrastructure (Il) are used as control variables. Among them, the level of economic development is measured by the logarithm of GDP per capita in a city, financial development is measured by the ratio of the balance of deposits and loans of financial institutions to GDP, urbanization is measured by the ratio of permanent urban population to the total regional population, and the degree of fiscal decentralization is measured by the ratio of public financial revenue to public financial expenditure. Infrastructure is measured as the logarithm of road kilometers per capita in a city.

4.3. Data Source

In this study, the data from Chinese cities from 2005 to 2020 are used as the study sample, and the data are obtained from authoritative databases such as the China Energy Statistical Yearbook, China Urban Statistical Yearbook, and CSMAR. According to Gao et al. (2024) [46], to ensure the data quality, we excluded the samples with serious missing data; to avoid the interference of extreme values on the estimation results, all variables were treated with a 1% shrinking tail. Finally, 253 city samples were collated, with a total of 4048 total data.

Table 1. Descriptive statistics.

VARIABLES	(1)	(2)	(3)	(4)	(5)
	N	Mean	Sd	Min	Max
GTFEE	4048	0.528	0.18	0.175	1.000
CPU	4048	1.250	0.609	0.000	4.057
El	4048	10.165	0.817	4.595	12.456
Fl	4048	2.107	1.023	0.508	12.619
Cl	4048	3.417	1.162	0.772	5.114
Cfl	4048	0.483	0.223	0.026	1.541
Il	4048	3.190	0.565	1.042	5.207

shows the descriptive statistics of the data in this study.

5. Empirical Result

5.1. The Baseline Test

Table 2. The results of baseline regression.

VARIABLES	(1)	(2)	(3)	(4)	(5)	(6)
	GTFEE	GTFEE	GTFEE	GTFEE	GTFEE	GTFEE
CPU	0.122 ***	0.108 ***	0.095 ***	0.091 ***	0.086 **	0.083 **
	(3.27)	(4.10)	(3.25)	(3.01)	(2.25)	(1.99)
El			0.054 **	0.088 ***	0.051 ***	0.053 **
			(2.51)	(9.05)	(2.62)	(2.33)
Fl				0.015 ***	0.053 *	0.013 *
				(3.67)	(1.91)	(1.76)
Cl				−0.002	−0.015 **	0.004
				(−0.46)	(−2.20)	(0.55)
Cfl					0.103 **	0.041 *
					(2.15)	(1.74)
Il						−0.014
						(−0.79)
Constant	2.070 ***	1.150 ***	1.037 ***	1.269 **	1.417 ***	1.208 ***
	(4.17)	(3.87)	(2.95)	(2.77)	(3.25)	(2.66)
Year FE	No	YES	YES	YES	YES	YES
City FE	No	YES	YES	YES	YES	YES
Adj-R2	0.217	0.255	0.354	0.399	0.417	0.438
Num cities	253	253	253	253	253	253

Notes: *** p < 0.01, ** p < 0.05, and * p < 0.10, and the t-value is in parentheses.

shows the results of the baseline regression. Column (1) is the random effects test results, while columns (2)–(6) are the test results of successively adding control variables and controlling for both time and city-fixed effects. As can be seen from the results in column (6), the CPU coefficients are all positively significant, indicating that climate policy uncertainty can indeed promote the improvement of cities’ energy efficiency. Moreover, for each unit increase in CPU, the GTFEE at the city level increases by 0.083 percentage points, confirming hypothesis 1. One possible explanation is that the frequent changes in climate policy undoubtedly send a strong signal to the outside world that the current climate problem in the region has attracted attention. The release of such signals could stimulate firms’ preference for green innovation and encourage them to actively seek solutions to potential climate risk problems [16]. Second, the government’s frequent adjustment of climate policy may make enterprises face higher environmental costs, which can further promote enterprises’ green investment and green innovation behavior [42]. Finally, the government could accelerate the construction of green infrastructure, which could improve the society’s green welfare [55], thereby improving the urban GTFEE.

5.2. Mechanism Analysis

According to the above analysis, climate policy uncertainty can improve the GTFEE of cities through two channels: improving public environmental concern (Pc) and optimizing the energy consumption structure (Es). According to Gao et al. (2024) [12], we use the mediating effect model to test the mechanism analysis proposed in the hypothesis.

First, we argue that increased climate policy uncertainty improves public concern for the environment. Specifically, climate policy uncertainty can improve public concern for the environment, which drives government attention to environmental issues and regulations [56], thereby reducing pollution emissions and improving urban energy efficiency. Following Yu et al. (2023) [57], we use the environment-related Baidu search index as a proxy variable for public environmental concerns. The first three columns of

Table 3. The results of mediating effect models.

VARIABLES	Pc			Es
	GTFEE	Pc	GTFEE	GTFEE	Es	GTFEE
	(1)	(2)	(3)	(4)	(5)	(6)
CPU	0.083 **	0.142 ***	0.071 *	0.083 **	0.216 **	0.069 **
	(1.99)	(3.27)	(1.72)	(1.99)	(2.09)	(1.98)
PC			0.023 **
			(2.08)
Es						0.062 **
						(2.21)
Constant	1.208 ***	3.480 ***	2.073 *	1.208 ***	1.098 **	2.07 **
	(2.66)	(3.71)	(1.79)	(2.66)	(1.98)	(2.33)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes	Yes
Observations	4048	4048	4048	4048	4048	4048
Adj-R2	0.438	0.313	0.595	0.438	0.419	0.647

Notes: *** p < 0.01, ** p < 0.05, and * p < 0.10, and the t-value is in parentheses.

show the results of the mechanism test for public environmental concerns. The regression coefficient of CPU on Pc in column (2) is 0.142 and significant at the 5% confidence level, indicating that CPU and the mediating variable Pc are positively correlated, i.e., an increase in climate policy uncertainty increases public environmental concern. Meanwhile, the regression coefficients of CPU and Pc in column (3) are both positive and at least significant at the 10% confidence level, indicating that public environmental concern is an effective impact channel.

Second, the increase in climate policy uncertainty could improve the energy consumption structure of cities, which may have an impact on the urban GTFEE. Specifically, frequent climate policy adjustments may increase the degree of environmental regulation in cities. Referring to Zeng et al. (2021) [52], this paper calculates the energy consumption structure index based on electricity consumption, natural gas consumption, and coal consumption, and the specific index construction method is shown above. In Table 3, the last three are the test results of the intermediate mechanism of urban energy consumption structure. In column (5), the regression coefficient of CPU on Es is 0.216, which is significant at the 5% confidence level, indicating that the increase in climate policy uncertainty will promote the optimization of urban energy structure. In column (6), the regression coefficients of CPU and Es are 0.069 and 0.062, respectively. At least at the 5% confidence level, it is significant, indicating that the urban energy consumption structure plays a part in a mediating role.

5.3. Robustness Test

This study uses the three ways instrumental variable method, the replacement of explained variables, and the replacement of the estimation model to verify the robustness of the results.

Firstly, because climate policy fluctuates in response to changes in the natural environment, it poses a potential endogenous problem in the previous analyses; we use global mean surface temperature (GMST) data as an instrumental variable to mitigate the effects of endogeneity in the two-stage model regression referring to Ren et al. (2022) [23]. Global mean surface temperature (GMST) is the average temperature of all land and oceans on the Earth’s surface.

Table 4. The results of the IV method.

VARIABLES	2SLS
	CPU	GTFEE
	(1)	(2)
GMST	2.465 ***	0.027 ***
	(6.370)	(3.53)
CPU		0.056 **
		(2.51)
Constant	2.018 ***	2.719 ***
	(2.73)	(2.91)
Controls	YES	YES
City FE	YES	YES
Year FE	YES	YES
Observations	4048	4048
Adj-R2	0.198	0.340

Notes: *** p < 0.01, ** p < 0.05, and the t-value is in parentheses.

below shows the estimation results using the 2SLS model with GMST as the instrumental variable. Both city and year are controlled by regression. It can be seen from the results in the table that the regression coefficient of GMST to GTFEE is still positive at the 1% confidence level, which means that the regression results in this paper are robust.

Secondly, referring to Zhang et al. (2023) [58], we replace the measurement of explained variables to determine the robustness of the baseline regression results. We use three alternative measurement methods: Epsilon-based measure (EBM), Directional Distance Function (DDF), and Super-Slacks Based Measure (Super-SBM). The EBM considers radial and non-radial improvements, the DDF considers directivity factors, and the Super-SBM model can sort Decision Making Units (DMU) according to distance. Columns (1)–(3) in

Table 5. The results of replacing the explained variable and estimation model.

VARIABLES	(1)	(2)	(3)	(4)	(5)
	EBM	DDF	Super-SBM	PCSE	FGLS
CPU	0.017 ***	0.036 **	0.047 ***	0.054 ***	0.041 **
	(3.53)	(2.05)	(3.78)	(2.84)	(2.24)
Constant	1.408 ***	1.277 ***	1.954 ***	2.352 ***	2.227 ***
	(3.72)	(2.68)	(3.21)	(3.34)	(3.61)
Controls	YES	YES	YES	YES	YES
City FE	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES
Observations	4,048	4,048	4048	4,048	4,048
Adj-R2	0.198	0.340	0.381	0.344	0.392

Notes: *** p < 0.01, ** p < 0.05, and the t-value is in parentheses.

below are the test results. The coefficients of CPU in the table are 0.017, 0.036, and 0.047, respectively, and are significant, at least at the 5% confidence level, which means that after controlling for control variables, city, and year factors, the conclusion that climate policy uncertainty contributes to urban GTFEE does not change whether the EBM method, DDF method or Super-SBM method is used to calculate the urban GTFEE.

Finally, the robustness of the main effect regression in this paper is tested by using the alternative econometric model. We use two alternative measurement methods. Panel Corrected Standard Error (PCSE) is a measurement method that corrects the standard error to solve the problem of heterogeneity and sequence correlation in panel data. Feasible Generalized Least Squares (FGLS) is a feasible generalized least squares model, which improves the estimation effect of least squares (OLS) by considering the heteroscedasticity and correlation of error terms. In Table 5 below, columns (4)–(5) are listed as the test results. The coefficients of CPU in the table are 0.054 and 0.041 and are significant at least at the 5% confidence level, which means that after controlling the control variables, city and year factors, the coefficients of CPU are positively significant regardless of whether the PCSE method or FGLS method is adopted. That is, the promoting effect of climate policy uncertainty on urban green total factor energy efficiency does not change with the change of test method.

6. Further Analysis

6.1. Heterogeneity Analysis

To further explore whether there are differences in the promotion effects of climate policy uncertainty on the GTFEE of cities with different characteristics, this study divides cities into the following two categories:

(1)

Resource-based and non-resource-based cities: Resource-based cities take local mineral resources and other natural resources as their economic lifeline, and their economic structure is single and unstable [59]. Compared with non-resource-based cities, the damaging effect of extreme climate events is more obvious in resource-based cities; the government should make more efforts to prevent such risks. The frequent adjustment of climate policies is the government’s prevention against climate risks, which will increase the degree of urban environmental regulation [60], and then force enterprises to make decisions that are conducive to environmental protection. Therefore, sample cities are classified according to the list of resource-based cities issued by The State Council, and regression tests are carried out respectively. Columns (1)–(2) in

Table 6. The results of Heterogeneity analysis.

VARIABLES	(1)	(2)	(3)	(4)
	Resource-Based	Non-Resource-Based	Developed	Developing
CPU	0.126 ***	0.044 **	0.135 ***	0.023
	(2.84)	(2.39)	(3.68)	(1.37)
Constant	1.305 ***	1.756 ***	1.578 ***	1.149 ***
	(3.02)	(3.36)	(2.79)	(2.59)
Controls	YES	YES	YES	YES
City FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Observations	1880	2168	1728	2320
Adj-R2	0.346	0.307	0.483	0.327

Notes: *** p < 0.01, ** p < 0.05, and the t-value is in parentheses.

below show the results of the heterogeneity test: the regression coefficients of CPU in column (1) and column (2) are 0.126 and 0.044, indicating that climate policy uncertainty has a more significant promoting effect on urban green total factor energy efficiency in resource-based cities;

(2)

Developed cities and developing cities: Energy use in cities is often strongly correlated with financial constraints [61]. Developing cities are often faced with more serious financial constraints, which will significantly reduce R&D investments, weaken enterprises’ green innovation and green production activities, and then affect the green total factor energy efficiency of cities. Therefore, 253 sample cities are divided into developed cities and developing cities according to their economic level, and empirical tests are carried out. Columns (3)–(4) in Table 6 below show the results of the heterogeneity test; the regression coefficient of CPU in column (3) is 0.135, significant at the 1% confidence level. The regression coefficient of CPU in column (4) is 0.023, not significant, indicating that climate policy uncertainty has a more obvious promoting effect on urban green total factor energy efficiency in developed cities.

6.2. Discussion

This study discusses the promoting effect of CPU on urban GTFEE, and it finds two impact channels, public environmental concern, and urban energy consumption structure, and this promoting effect is particularly significant in resource-based cities and economically developed cities. First, our findings demonstrate that climate policy uncertainty has a positive impact on the GTFEE in cities, which also confirms similar conclusions in some existing studies. For example, Du et al. (2023) [10] believe that high climate policy uncertainty can positively regulate the relationship between green credit and green innovation, while Li et al. (2024) [16] propose that climate policy uncertainty has a direct positive correlation with corporate green innovation. These results are almost consistent with the idea proposed in the theoretical hypothesis section that climate policy uncertainty can promote enterprises’ green R&D investments, and thus promote the improvement of an urban GTFEE. Cao and Chen (2024) [62] argue that the improvement of public environmental awareness can improve enterprises’ green technology efficiency, and this conclusion also resonated with hypothesis 2 of this paper. In addition, this paper discusses the two aspects of encouraging the government to formulate environmental policies and restraining the possible collusion between the government and enterprises and further analyzes and explains the existing conclusions.

Second, this study also raises several points where it diverges from existing research, particularly about the economic effects of climate policy uncertainty. Some scholars hold a pessimistic attitude towards climate policy uncertainty. For example, Ren et al. (2022) [23] believe that climate policy uncertainty significantly reduces the R&D investment of enterprises, which in turn reduces the total factor productivity of enterprises. On the other hand, Mo and Liu (2023) [60] propose that climate policy uncertainty would increase financial constraints and hinder enterprises’ digitalization process. However, we believe that climate policy uncertainty does increase the production costs for firms in the short term, but speeds up green innovation in the long run, so this inconsistent conclusion is reasonable.

7. Conclusions and Policy Implications

Based on panel data from 253 prefecture-level cities in China from 2005 to 2020, this study explores the contribution of CPU to urban GTFEE. The results of this paper show that, first, climate policy uncertainty is positively correlated with urban green total factor energy efficiency, and this conclusion is still robust after a series of endogenous problems are excluded. Secondly, climate policy uncertainty promotes urban green total factor energy efficiency by increasing public environmental concern and optimizing the urban energy consumption structure. Finally, in developed cities and resource-based cities, climate policy uncertainty has a better effect on promoting urban green total factor energy efficiency.

The limitation and improvement of this study lies in the lack of discussion on the decomposition of the green total factor energy efficiency index. The green total factor energy efficiency index has multiple forms of measurement and can be further divided into the technology progress index and technology efficiency index. Future studies can further investigate the promoting role of climate policy uncertainty from the perspective of index decomposition. In addition, future studies can also consider introducing more mechanism variables to examine the impact of different factors on the relationship between the two. For example, the digital level can be introduced to explore its regulatory effect and to more comprehensively analyze the driving factors and influencing channels of urban green total factor energy efficiency.

Based on the above research conclusions, this paper puts forward the following three suggestions. First, strengthen the forward-looking climate policy formulation: To improve green total factor energy efficiency in cities, governments should strengthen forward-looking climate policy formulation. Forward-looking means that policymaking needs to fully consider future climate change trends and their impact on urban energy systems to make adaptive adjustments in advance, to avoid unnecessary distress and costs to businesses and residents caused by frequent changes. By being more forward-looking, governments can give the public and businesses an environmental policy outlook to more effectively guide the green transformation of urban energy systems, improve energy efficiency, and reduce greenhouse gas emissions.

Second, promote green technology innovation and application: The key to improving urban green total factor energy efficiency is to promote green technology innovation and application. The government should increase investment in the research and development of green technology, encourage enterprises to strengthen technological innovation and promote the market-oriented application of green technology. At the same time, we can also attract more social capital to invest in green technology innovation by setting up green technology innovation funds and providing tax incentives. In addition to strengthening international cooperation and exchanges, the introduction of foreign advanced green technology and experience is also an important way to improve urban energy efficiency. Through the promotion of green technology innovation and application, it can promote the optimization and upgrading of urban energy systems, improving energy efficiency.

Third, improve the energy management and regulatory system: Improving the energy management and regulatory system is an important guarantee for improving the urban green all-factor energy efficiency. The government should establish a sound energy management system, clarify the responsibilities and division of labor of various departments, and strengthen energy data statistics and monitoring. At the same time, the government should strengthen the supervision of the energy market, crack down on illegal energy use, and ensure fair competition and healthy development of the market. By improving the energy management and supervision system, we can ensure that the government’s environmental policies are effectively implemented, and thus promote the green transformation of the urban energy system.