How Does Critical Peak Pricing Boost Urban Green Total Factor Energy Efficiency? Evidence from a Double Machine Learning Model

Abstract

Green and low-carbon development constitutes an essential pathway toward high-quality socioeconomic transformation, with improving urban green total factor energy efficiency (GTFEE) critical to achieving this objective. Based on the sample data of Chinese cities from 2013 to 2022, this study systematically investigated the impact and mechanism of critical peak pricing on urban GTFEE by using the double machine learning method, effectively supplementing the existing literature. This study finds that this policy significantly enhances urban GTFEE. Mechanism analysis indicates that critical peak pricing generates a dual effect by increasing the price difference between peak and off-peak hours and enhancing energy efficiency through two important channels: market expansion and technology-driven innovation. Heterogeneity analysis indicates that the critical peak pricing policy has a more significant promotion effect on non-resource-based, strong government administrative power, as well as central and eastern regions. These findings advance the power marketization reform framework and provide new theoretical support for promoting low-carbon energy transformation.

1. Introduction

Global climate change and environmental problems are becoming increasingly more severe and have become major challenges for the development of human society [1,2]. Against this backdrop, coordinating the relationship between economic growth and environmental protection has become a significant issue that all countries face. As the world’s largest energy consumer, China is under substantial pressure to transform its energy structure. Despite the rapid development of renewable energy, with the installed capacity exceeding 1.5 billion kilowatts by the end of 2023, accounting for 40.2% of the global total capacity, the contribution of clean energy to actual energy consumption is still limited [3]. This structural contradiction not only restricts the process of green and low-carbon development but also highlights the urgency of accelerating the transformation of the energy system. Facing this challenge, green total factor energy efficiency (GTFEE) can effectively measure the real performance of various regions in the context of sustainable development by incorporating economic growth, energy consumption, and environmental pollution into a unified analytical framework, providing a scientific basis for solving the dilemma between economic growth and environmental protection [4].

Existing research has extensively examined the influencing factors of urban GTFEE from multiple perspectives. At the technological innovation level, studies have demonstrated that adopting clean technologies and digital transformation significantly enhances energy efficiency [5,6]. In environmental regulation, moderate policy interventions have been shown to stimulate innovation via the Porter effect [7]. However, the relationship between regulatory intensity and GTFEE exhibits nonlinearity [8]. At the market mechanism level, the price discovery function of green finance and integrated energy market reforms has been identified as a key influencing factor [9,10]. Although existing research has explored policy–market synergies, it focuses disproportionately on conventional determinants while pervasively neglecting peak electricity pricing—a critical yet understudied market-based instrument.

This gap is heightened in significance during energy transitions. Specifically, with the large-scale integration of renewable energy into the grid, the demand for dynamic regulation capabilities in the power system is increasing daily, making the dynamic price mechanism represented by critical peak pricing demonstrate unique value. The critical peak pricing guides demand response by implementing differentiated electricity prices during the peak load period of the system. It can not only effectively reduce peaks and fill valleys based on price elasticity [11], but also enhance the system’s regulation capacity by improving the accuracy of market forecasting [12]. In contexts with high renewable penetration, this policy emerges as a core tool for optimizing the allocation of power resources by guiding electricity consumption behaviors to adapt to power generation [13,14].

This paper argues that critical peak pricing may promote urban GTFEE through multi-dimensional paths. Specifically, at the short-term operational level, this policy leverages critical peak pricing to steer demand-side response. This approach diminishes peak-hour dependence on fossil fuels and enhances the utilization of renewable energy [15]. The expanded peak–valley price difference creates a continuous market incentive at the long-term development level. On the one hand, this policy promotes investment in energy storage systems, thereby expanding the market potential for renewable sources [16]. On the other hand, it drives flexibility transformation on the power generation side and technological innovation in smart grids [11], thereby systematically optimizing the efficiency of the energy structure.

Therefore, this study addresses the following questions: What is the specific impact of China’s critical peak pricing on urban GTFEE? What are its mechanisms? Is there significant heterogeneity among different cities? To systematically answer these questions, we utilize panel data from 273 cities in China from 2013 to 2022 and conduct an empirical analysis, applying a double machine learning (DML) model. The experimental results show that critical peak pricing significantly improves urban GTFEE. Mechanism analysis indicates that market expansion and technological innovation are essential transmission channels. Heterogeneity analysis demonstrates that the critical peak pricing policy has a more significant promoting effect on non-resource-based, strong government administrative power, as well as central and eastern regions.

This study makes several potential contributions. First, it addresses a critical research gap concerning the impact of critical peak pricing on urban GTFEE. While the existing literature has extensively examined influencing factors within traditional dimensions such as technology [5], environmental policy [2], and the market [17], a systematic assessment of critical peak pricing effectiveness has been notably lacking. In response to this limitation, this paper empirically demonstrates the significant positive effect of the critical peak pricing policy on urban GTFEE using DML. Beyond advancing the theoretical foundations of electricity market reform, this discovery offers empirical support for facilitating the low-carbon transition.

Second, it reveals the mechanism by which critical peak pricing enhances urban GTFEE and identifies the heterogeneous characteristics of its impact. Existing studies on critical peak pricing mainly discuss its effects from the perspective of demand-side response [18] or promoting new energy consumption [19] but have not deeply explored its mechanism on urban GTFEE. In response to this deficiency, this study systematically explains its mechanisms from two dimensions: market expansion and technological innovation effects. This multi-dimensional mechanism analysis provides a new perspective for understanding this relationship. Furthermore, this paper identifies the heterogeneous characteristics of resource endowment, administrative power, and regional development dimensions, providing an empirical basis for formulating differentiated regional energy policies.

Third, this study adopts a DML approach to make up for the deficiencies of traditional empirical models. Reducing the risk of omitted variables and model misspecifications enhances the credibility of assessing the critical peak pricing. The existing literature evaluating the impact of environmental policies on energy efficiency typically employs traditional econometric models, such as the difference-in-differences method [20]. These models are highly dependent on prior assumptions for selecting and handling control variables and struggle to capture complex inter-variable relationships flexibly [21]. In contrast, the DML model can handle high-dimensional covariates and enhance robustness against confounding factors. More importantly, its flexible functional form reduces reliance on strong prior assumptions, lowering the risk of model misspecification.

The remainder of this paper is structured as follows: Section 2 presents a literature review; Section 3 provides theoretical analysis; Section 4 details the research design; Section 5 presents empirical results; and Section 6 concludes and proposes policy implications.

2. Literature Review

The relevant literature primarily falls into two strands. The first strand mainly examines the economic effects of critical peak pricing policies. Existing studies can be roughly categorized into two groups. The one group focuses on the demand response associated with critical peak pricing. Empirical studies demonstrate that CPP induces short-term demand response via dynamic pricing signals, motivating load shifting to off-peak hours [18,22,23]. However, its effectiveness is moderated by the penetration rates of smart metering infrastructure and automated control systems [21,24]. Moreover, most studies rely on short-term observational data. This limitation has resulted in limited systematic investigation into potential consumer price adaptation behaviors and the attenuation of demand response intensity.

Another type of literature focuses on the impact of critical peak pricing on the electric power supply side. Studies have shown that critical peak pricing can effectively reshape the system load curve by guiding users, optimizing energy distribution, and ultimately achieving the dual benefits of reducing the power generation cost and carbon emissions [25,26]. From the perspective of the generation side, the load pattern optimization brought by CPP not only significantly reduces the operating time of high-cost peak shaving units, but also improves the utilization rate of base-load units and renewable energy [27,28]. Furthermore, critical peak pricing can significantly enhance power generation and demand-side coordination [29,30]. However, the optimization effect on the operation of the power-generation side highly depends on whether the electricity price signal can accurately reflect the real-time marginal cost and whether the system has the regulatory flexibility to cope with rapid changes in load [31,32]. It is worth noting that the existing research on CPP is still relatively limited, lacking a systematic assessment of the benefits of its improvement in GTFEE.

The second strand mainly explores the synergistic relationship between critical peak pricing and renewable energy integration. Critical peak pricing enhances system operational efficiency by reshaping the spatio-temporal price signals in electricity markets, thereby reducing the curtailment rates of wind and solar power [19]. Additionally, it can promote renewable energy investment by optimizing revenue structures [33]. Nevertheless, the realization of these synergies faces dual challenges. First, the efficacy of critical peak pricing in advancing renewable energy depends on deploying enabling technologies, particularly energy storage and demand-side management systems. However, prohibitive storage costs present a major impediment to such deployments. Second, much of the existing research is based on mature electricity markets (e.g., Australia and the European Union). The potential synergies may be substantially constrained in regions with limited grid flexibility or low renewable energy penetration.

The second strand examines the factors influencing GTFEE. Existing research primarily focuses on determinants across three dimensions. First, in technological innovation, deploying clean technologies and digital transformation have been identified as critical drivers for enhancing urban GTFEE [5,6]. Second, environmental regulation can compel corporate green innovation [34]. Participation in carbon markets has also been shown to enhance urban GTFEE, with the effect strengthening over time [17]. Third, market mechanisms leverage green finance to improve efficiency through reduced financing costs of clean energy initiatives [9]. Although the existing literature widely acknowledges that enhancing GTFEE relies on the synergistic interplay of policy frameworks and market dynamics, current research remains predominantly focused on established factors. Consequently, studies comprehensively investigating critical peak pricing, an integral, market-based price in electricity markets, remain conspicuously scarce.

Overall, the existing research has notable limitations. On the one hand, most of the literature employs fragmented analytical frameworks, either examining only the short-term policy effects of the critical peak pricing policy or analyzing the influencing factors of GTFEE in isolation [17,18]. Systematic path analysis regarding “price signal → energy structure → GTFEE” is lacking. On the other hand, research on urban GTFEE determinants predominantly emphasizes technological advancement [5] and emissions trading systems [34], while overlooking the price leverage effect of critical peak pricing. Although some studies have acknowledged its demand regulation function [21], systematic evidence regarding improvement effects through channels such as market expansion and technological progress remains scarce. Addressing this research gap, this paper systematically examines how critical peak pricing policies affect urban GTFEE and identifies their transmission mechanisms and heterogeneous effects, thereby providing an evidence-based foundation for improving electricity pricing mechanisms.

3. Theoretical Analysis and Research Hypotheses

The signal transduction theory offers a foundational analytical framework for examining critical peak pricing policy in electricity markets. This theory indicates that market participants with an informational advantage under information asymmetry can release observable signals to influence behavior adjustments among less-informed parties [7]. This research logic is shown in the research framework diagram in Figure 1. During the green economy transition, critical peak pricing is a dynamic pricing mechanism that systematically transmits spatio-temporal price signals to enhance urban GTFEE. This mechanism operates through multiple paths. Firstly, at the resource allocation level, critical peak pricing signals spatio-temporal disparities in power supply and demand via dynamic pricing. This optimization of power flows and energy structures subsequently enhances clean energy consumption [21]. Secondly, at the behavioral guidance level, this policy regulates the decisions of both supply and demand sides through price elasticity. With power generation systems, critical peak pricing accelerates the transformation of flexible power sources and renewables investment, leveraging price premium incentives during peak hours [35]. Critical peak pricing incentivizes industrial and commercial consumers to reschedule manufacturing operations and shift loads to off-peak periods for demand-side management. This behavioral adaptation reduces the load peak periods’ dependence [11], enhancing energy utilization efficiency. Finally, critical peak pricing optimizes grid efficiency for system coordination through three mechanisms: reduced reserve capacity requirements, increased generation asset utilization, and mitigated congestion costs. This mechanism optimizes the short-term matching of power supply and demand and guides the energy structure towards green trajectories through long-term price signals. Therefore, we propose the following:

Hypothesis 1.

Critical peak pricing can significantly enhance urban GTFEE.

The impact of critical peak pricing on urban GTFEE is mainly realized through the following transmission paths. First, the market expansion effect manifests as critical peak pricing, enhancing the revenue of energy storage systems through dynamic price signals. In the market environment, where the peak–valley difference of electricity prices is expanding, energy storage systems can obtain stable returns through the arbitrage operation mode of “low storage and high release”, which directly stimulates the large-scale investment and construction of energy storage equipment [36,37]. Simultaneously, the policy’s price leverage mechanism improves the temporal alignment between power system load and intermittent renewable generation. Redirecting electricity demand toward renewable output peaks significantly boosts the return on investment for renewable energy projects [38].

Second, the technological innovation effect operates through dual pathways. On the demand side, the expansion of the peak–valley price difference has formed an effective incentive signal, continuously driving the iterative upgrade of energy storage technology towards high efficiency. Meanwhile, the demand for refined electricity prices drives technological innovation in smart-metering infrastructure. This advancement facilitates the evolution of demand-side management towards intelligence and precision [39,40]. On the supply side, the policy promotes the continuous optimization of renewable energy grid connection technologies and enhances the capacity to consume clean energy. It promotes the traditional thermal power plants’ accelerated flexibility transformation process and improves the system’s regulation capacity. Meanwhile, it promotes the deep integration of digital technology and intelligent dispatching systems and accelerates integrated resource coordination [41,42]. The efficiency gains from technological innovation have further strengthened the promotion effect of CPP. Therefore, we propose the following:

Hypothesis 2.

Critical peak pricing enhances urban GTFEE through the dual paths of market expansion and technological innovation.

Firstly, the endowment of urban resources constitutes a structural constraint. Resource-based cities, where extractive and processing sectors dominate industrial composition, exhibit systemic fossil fuel dependencies in their economic and energy structures [43]. This development path has fostered a distinct technology lock-in effect, leading to high transformation costs. These costs may impede the effective response to price signals and diminish the effectiveness of critical peak pricing. In contrast, non-resource-based cities possess inherent advantages, including greater industrial diversification, enhanced flexibility in energy structures, and robust technological innovation capacity. These attributes enable energy efficiency gains through storage and demand response management [44]. Secondly, the differences in urban administrative power lead to the differentiation of policy effects [45]. With their stricter regulatory capabilities and more complete supporting and effective assessment mechanisms, provincial cities can ensure the full implementation of peak electricity price policies. In contrast, non-provincial cities, constrained by scarce regulatory resources and insufficient enforcement capabilities, often fail to achieve the expected policy implementation results. In addition, regional development effects bring about apparent spatial heterogeneity [46]. The eastern region can fully respond to price signals by relying on its highly market-oriented economic environment, advanced power grid infrastructure, and leading technological innovation capabilities. Despite abundant renewable energy endowments, the western region remains constrained by underdeveloped market mechanisms and inadequate power infrastructures. This region is susceptible to dependence on the low-price electricity development trap, which disrupts price signal transmission and ultimately undermines policy effectiveness [47]. Therefore, we propose the following:

Hypothesis 3.

The promoting effect of the critical peak pricing on the GTFEE of cities is more obvious in non-resource-based, strong government administrative power, and central and eastern regions.

4. Research Design

4.1. Model Design

Compared with traditional econometric methods, double machine learning (DML) has multiple advantages [48,49]. Firstly, DML can flexibly handle high-dimensional control variables and complex functional relationships through machine learning algorithms, effectively alleviating model setting errors and missing variables, thereby enhancing the accuracy of causal recognition. Secondly, DML is robust against complex relationships, such as nonlinearity and interaction effects between control and outcome variables, while traditional linear models tend to fail in such situations. In addition, DML subsequently models the residuals after fitting the control variables with the treatment and outcome variables through the orthogonalization process, which can consistently estimate the policy effects under relatively weak assumptions.

Therefore, this study adopts the DML method to empirically investigate the impact of critical peak pricing on urban GTFEE [48], and constructs the following benchmark model:

$${\mathit{GTFEE}}_{it}={\theta }_0{\mathit{CPP}}_{it}+g\left(X_{it}\right)U_{it}$$

(1)

$${\mathit{CPP}}_{\mathit{it}}=m\left(X_{it}\right)+V_{it}$$

(2)

$$E\left(U_{it}\right.|\left.X_{it},{CPP}_{it}\right)=0$$

(3)

$$E\left(V_{it}|\left.X_{it}\right)\right.=0$$

(4)

In the above formula, i and t represent the city and the year, respectively. ${\mathit{GTFEE}}_{it}$ is the explained variable, which measures GTFEE of city $i$ in year $t$. ${\mathrm{CPP}}_{it}$ is the core explanatory variable. If city $i$ implements the critical peak pricing policy in period $t$, its value is 1; otherwise, it is 0. The coefficient ${\theta }_0$ directly characterizes the average processing effect of the critical peak pricing policy on the urban GTFEE. $X_{it}$ is a set of high-dimensional control variables.

This paper adopts the random forest algorithm, and 5-fold cross-validation is used for parameter estimation [50]. The specific steps are as follows. Firstly, the random forest algorithm is used to fit the control variable function $g\left(X_{it}\right)$ to obtain the estimator $\tilde{g}\left(X_{it}\right)$, and the orthogonalized rescaling of the explained variable is calculated: ${\tilde{Y}}_{it}=GTFE{E}_{it}-\tilde{g}\left(X_{it}\right)={\theta }_{0}CP{P}_{it}+U_{it}$, ${\tilde{Y}}_{it}$ represents the pure $GTFE{E}_{it}$ after removing the influence of control variables. Secondly, the random forest algorithm is similarly employed to fit the function $\tilde{m}\left(X_{it}\right)$, yielding the residual ${\tilde{V}}_{it}=CP{P}_{it}-\tilde{m}\left(X_{it}\right)$. This residual represents the net policy effect after removing the influence of control variables. Finally, ${\tilde{V}}_{it}$ is regarded as an instrumental variable of $CP{P}_{it}$, and the parameter ${\theta }_0$ is estimated using linear regression:

$${\tilde{Y}}_{it}={\theta }_0{\tilde{V}}_{it}+{\epsilon }_{it}$$

(5)

Although partial linear models have advantages in the economic interpretation of parameters, the impact of critical peak pricing on urban GTFEE may have a nonlinear transmission mechanism. We further constructed an interactive model for robustness analysis to test the robustness of the benchmark results and capture potential nonlinear features. The model is set as follows:

$${GTFEE}_{it}=g\left({CPP}_{it},X_{it}\right)+U_{it},E\left(U_{it}|X_{it},{CPP}_{it}\right)=0$$

(6)

$${CPP}_{it}=m\left(X_{it}\right)+V_{it},E\left(V_{it}|X_{it}\right)=0$$

(7)

In the interactive model, the effect of the policy is measured by the conditional average treatment effect:

$${\tilde{\theta }}_0=E\left[g({CPP}_{it}=1,X_{it})-g({GTFEE}_{it}=0,X_{it})\right]$$

(8)

The model’s estimation procedure follows that of partial linear models, employing a DML framework for consistent and comparable results. By comparing the estimation results across both model settings, this approach enables a more robust and comprehensive assessment of critical peak pricing effects.

4.2. Variable Description

4.2.1. Explained Variable

This paper focuses on urban GTFEE as the explained variable. Referring to Gao et al. [51], the super-efficiency slack-based measure (SBM) model, including undesired output, is adopted to calculate urban GTFEE from 2013 to 2022. We suppose there are n decision units (DMUs), and each DMU has a set of production factor inputs, $s_1$ kinds of expected output, and $s_2$ kinds of unexpected output. Define the input matrix of $X\in R^{m\times n}$, the expected output matrix of $Y^g\in R^{s_1\times n}$, and the unexpected output matrix of $Y^b\in R^{s_2\times n}$. Among them, $S^{-}\in R^m,S^g\in R^{s_1},S^b\in R^{s_2}$ represent slack variables with input redundancy, insufficient expected output, and excessive unexpected output, respectively. The linear programming problem is constructed as follows:

$$\mathit{min}\rho ={\displaystyle \frac{1-\frac{1}{m}{\sum }_{i=1}^m\frac{s_i^{-}}{x_{i0}}}{1+\frac{1}{s_1+s_2}({\sum }_{i=1}^{s_1}\frac{s_i^g}{y_{i0}^g}+{\sum }_{i=1}^{s_2}\frac{s_i^b}{y_{i0}^b}) }}$$

(9)

$$s.t. X\lambda +S^{-}=x_0$$

(10)

$$Y^g\lambda -S^g=y_0^g$$

(11)

$$Y^b\lambda +S^b=y_0^b$$

(12)

$x_0,{ y}_0^g,{ y}_0^b$ are the evaluation of the DMU input, the expected output, and the unexpected output, respectively. $s_i^{-},s_i^g,s_i^b$ represent the insufficient input redundancy, the expected output, and the unexpected output excess amount, and ρ is used to assess the efficiency value of the DMU. The input factors mainly include capital stock (K), labor force (L), and energy consumption (EU). The expected output indicators are GDP. In terms of the selection of undesired outputs, this paper comprehensively refers to the studies of Gao et al. [51] and Zhang et al. [52], and selects sulfur dioxide (SO₂), smoke dust emissions, wastewater emissions, and carbon dioxide (CO₂) as environmental negative output indicators. SO₂ and smoke dust emissions are significant sources of air pollution, directly affecting air quality and ecosystem health. Wastewater discharge represents water environmental pollution pressure, while CO₂ emissions are associated with global climate change issues. The above indicators jointly cover the three key ecological dimensions of air pollution, water pollution, and carbon emissions, and can comprehensively reflect the environmental cost of urban economic activities. They are also consistent with China’s environmental statistics system’s focus and dual-carbon policy.

4.2.2. Explanatory Variable

The core explanatory variable is critical peak pricing, denoted as ${CPP}_{it}$. It should be noted that when determining the implementation year of critical peak pricing, considering the timeliness of policy enactment, if a city initiates the policy after October of a given year, the implementation year should be recorded as the following year [21].

4.2.3. Control Variables

Following Zhou and Qi [53], this paper introduces a series of control variables to account for other potential factors affecting urban GTFEE. Specifically, these control variables include the following. Industrial structure (structure), measured by the proportion of secondary industry value-added in regional GDP, captures the overall characteristics of the region’s industrial structure. Economic development level (Pgdp), obtained by dividing the city’s gross domestic product by the total population at the end of the year, represents the degree of economic development of a region. Population density (density), calculated as a city’s population divided by its administrative area, serves as a proxy for spatial heterogeneity in human activity intensity. The level of human capital (education), using educational expenditure at the prefecture level, reflects regional investment in education and human resources. The level of pollutant emissions (Pollu), measured by sulfur dioxide emissions, represents regional environmental pollution intensity.

4.3. Data Sources

This study identified cities that implemented the critical peak pricing policy and the corresponding implementation years by manually organising relevant policy documents and constructing policy variables. The data for the explained and control variables are primarily sourced from authoritative statistical yearbooks published by the National Bureau of Statistics of China, including the “China Urban Statistical Yearbook”, “China Environmental Statistical Yearbook”, “China Energy Statistical Yearbook”, and “China Science and Technology Statistical Yearbook”. These datasets are further supplemented with data from the Wind financial database. This study employed linear interpolation during the data preprocessing stage to address missing values for specific cities or years. Given the relatively small number of missing data points, their impact on the estimation results is considered minimal. The final panel dataset constructed in this study spans 273 Chinese prefecture-level cities from 2013 to 2022, comprising a total of 2111 valid observations.

Table 1. Descriptive statistics.

Variable	N	Mean	S.D	Min	First Quart	Median	Third Quart	Max
$\mathrm{GTFEE}$	2111	0.282	0.084	0.021	0.228	0.274	0.331	1.148
$\mathrm{CPP}$	2111	0.217	0.413	0.000	0.000	0.000	0.000	1.000
$\mathrm{S}\mathrm{t}\mathrm{r}\mathrm{u}\mathrm{c}\mathrm{t}\mathrm{u}\mathrm{r}\mathrm{e}$	2111	50.265	9.970	18.670	44.180	50.600	56.150	90.970
$\mathrm{P}\mathrm{g}\mathrm{d}\mathrm{p}$	2111	10.271	0.688	8.138	9.798	10.259	10.764	13.056
$\mathrm{D}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{y}$	2111	5.795	0.851	1.609	5.278	5.932	6.447	7.882
$\mathrm{E}\mathrm{d}\mathrm{u}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}$	2111	12.499	0.892	7.135	11.926	12.507	13.101	15.853
Pollu	2111	10.671	1.046	0.693	10.175	10.827	11.335	13.434

presents the descriptive statistics for these variables.

5. Empirical Result

5.1. Benchmark Regression Results

This paper employs a DML model to identify the impact of critical peak pricing on urban GTFEE.

Table 2. Benchmark regression.

	(1)	(2)	(3)	(4)
	Partial Linear Model		General Interaction Model
	GTFEE	GTFEE	GTFEE	GTFEE
${\mathit{CPP}}_{\mathit{it}}$	0.035 ***	0.025 ***	0.032 ***	0.029 ***
	(0.004)	(0.005)	(0.002)	(0.002)
Controls	√	√	√	√
Firm FE	✕	√	✕	√
Year FE	✕	√	✕	√
Observation	2111	2111	2111	2111

Note: This table reports benchmark regression results from the DML model. The explained variable is urban GTFEE. The explanatory variable is critical peak pricing. The standard error is shown in parentheses. *** indicates significance at the 1% level. The √ symbol indicates the inclusion of the variable. The ✕ indicates that this variable is not included.

reports the core estimation results. Columns (1) and (2) present the regression outcomes of partial linear models. Column (1) controls for basic control variables. The coefficient of the policy variable ${\mathit{CPP}}_{\mathit{it}}$ is 0.035, which is statistically significant at the 1% level. Column (2) incorporates enterprise and year-fixed effects to address potential omitted variable bias. The coefficient of ${\mathit{CPP}}_{\mathit{it}}$ decreases slightly to 0.025 while remaining significantly positive at the 1% level. Economically, this result indicates that implementing critical peak pricing policies can, on average, increase GTFEE by approximately 2.5%. This increase suggests that the critical peak electricity pricing policy has positive policy value in promoting the achievement of national energy conservation, emission reduction, and green transformation goals. Columns (3) and (4) present the estimation results of the interaction model, with CPP coefficients of 0.032 and 0.029, respectively, both significant at the 1% level and relatively close to the results of the partially linear model. The economic significance is consistent with that of the partially linear model, indicating that CPP has a significant positive effect on GTFEE, further demonstrating that the model setting has a limited impact on the core conclusion.

5.2. Robustness Test

We implement multiple robustness checks to assess the reliability of our regression findings. These include adjusting the DML model configuration, replacing the estimation model, incorporating interaction fixed effects, etc. The specific descriptions and corresponding results are as follows:

5.2.1. Replace the Algorithm Selection

To ensure that the results are not affected by the specific algorithm selection, this paper adopts the support vector machine (SVM) algorithm to replace the random forest algorithm in the benchmark model. The SVM algorithm has advantages in processing high-dimensional data and capturing nonlinear relationships and can alleviate the risk of overfitting [54]. As presented in Column (1) of Table 3, the coefficient estimates for the policy variable ${\mathit{CPP}}_{\mathit{it}}$ remain statistically significant and positive at the 1% level. This confirms that the positive effect of critical peak pricing on urban GTFEE is robust to the choice of machine learning algorithm, thereby strengthening the credibility of our findings.

5.2.2. Adjust the K-Fold Cross-Validation

This paper uses 5-fold cross-validation (K = 5) in the benchmark model. To test the robustness of the benchmark results in selecting cross-validation K values, this study adopted 10-fold cross-validation (K = 10) for re-estimation. The results under 10-fold cross-validation are presented in Column (2) of Table 3. ${\mathit{CPP}}_{\mathit{it}}$ maintains its statistically significant positive coefficient at the 1% regardless of cross-validation folds. This consistency demonstrates the insensitivity of our key findings to the choice of K-value, providing further robustness validation for our core results.

5.2.3. Change the Year Setting for Implementing the Policy

To address the model setting issues that may arise from the division of policy time points, this paper redefines the core explanatory variables, eliminates the original time point rules, and takes the year when the city begins to implement the critical peak electricity price policy as the basis for setting the value of to 1. Column (3) of Table 3 presents the regression results after re-estimation. The policy coefficients of ${\mathit{CPP}}_{\mathit{it}}$ is significantly positive at the 1% level, indicating that the original conclusion is not sensitive to how the policy time is set, further supporting the robustness of the baseline results.

5.2.4. Consider the Interaction Fixed Effect

Our baseline specification incorporates firm-year fixed effects to address potential bias from industry-specific time-varying unobservables. The estimation results are presented in Column (4) of Table 3. Critically, the coefficients for policy variables ${\mathit{CPP}}_{\mathit{it}}$ retain positive and statistically significant at the 1% level, corroborating the robustness of core findings under firm-year fixed effects.

5.2.5. Eliminate Policy Interference from Resource-Based Cities

Resource-based cities usually implement independent energy subsidy policies, which may interfere with the effect of critical peak pricing [55]. This paper excluded resource-based cities from the sample to eliminate this confounding factor and reverted. After excluding resource-based cities, ${\mathit{CPP}}_{\mathit{it}}$ retains statistically significant positive coefficients at the 1% level. This confirms the robustness of our core findings to confounding effects from region-specific energy subsidies.

Table 3. Robustness test.

	(1)	(2)	(3)	(4)	(5)
	Replace the Algorithm Selection	Adjust the K-Fold Cross-Validation	Change the Year Setting for Implementing the Policy	Consider the Interaction Fixed Effect	Eliminate the Resource-Based Cities Policy Interference
${\mathit{CPP}}_{\mathit{it}}$	0.038 ***	0.022 ***	0.024 **	0.033 ***	0.026 ***
	(0.004)	(0.006)	(0.010)	(0.005)	(0.005)
Controls	√	√	√	√	√
Firm FE	√	√	√	√	√
Year FE	√	√	√	√	√
Firm*Year FE	✕	✕	✕	√	✕
Observation	2111	2111	2111	2111	2111

Note: The standard error is shown in parentheses. ***, ** indicates significance at the 1% and 5% levels. The √ symbol indicates the inclusion of the variable. The ✕ indicates that this variable is not included. Firm*Year FE represents the interaction fixed effects between firms and years.

Table 3. Robustness test.

	(1)	(2)	(3)	(4)	(5)
	Replace the Algorithm Selection	Adjust the K-Fold Cross-Validation	Change the Year Setting for Implementing the Policy	Consider the Interaction Fixed Effect	Eliminate the Resource-Based Cities Policy Interference
${\mathit{CPP}}_{\mathit{it}}$	0.038 ***	0.022 ***	0.024 **	0.033 ***	0.026 ***
	(0.004)	(0.006)	(0.010)	(0.005)	(0.005)
Controls	√	√	√	√	√
Firm FE	√	√	√	√	√
Year FE	√	√	√	√	√
Firm*Year FE	✕	✕	✕	√	✕
Observation	2111	2111	2111	2111	2111

Note: The standard error is shown in parentheses. ***, ** indicates significance at the 1% and 5% levels. The √ symbol indicates the inclusion of the variable. The ✕ indicates that this variable is not included. Firm*Year FE represents the interaction fixed effects between firms and years.

5.3. Endogeneity Test

5.3.1. Instrumental Variable Test

This paper constructs a partial linear instrumental variable model to alleviate the endogeneity problem caused by omitted variables and mutual causality. The selected instrumental variable (IV) represents the proportion of words related to low carbon in the annual work reports of the sample city governments. The data are sourced from the official websites of each city government [21]. Under the current Chinese system, the power to formulate and adjust electricity price policies is still concentrated in the relevant government departments. If a local government emphasises energy conservation and other related expressions in its report, it usually indicates that the region is more willing to actively implement various energy conservation policies, including critical peak pricing. Therefore, there is a significant correlation between IV and the implementation of critical peak pricing policies. Meanwhile, the text content in the government work report mainly affects the urban GTFEE by influencing the implementation of policies. It does not directly interfere with the GTFEE, thus satisfying the exclusivity constraint of the IV test. The results are shown in Column (1) of Table 4. After introducing IV to alleviate the endogeneity bias, the coefficient of the policy variable remains significantly positive at the 1% level, indicating that after controlling for potential endogeneity issues, the critical peak pricing policy still has a significant promoting effect on increasing the urban GTFEE.

5.3.2. Replace the DID Estimation Model

This study substitutes the DML approach with a difference-in-differences (DID) specification to address potential model specification concerns. The DID model controls for unobservable time-invariant confounders and mitigates endogeneity concerns by comparing outcome changes between treatment and control groups before and after policy implementation [56]. As shown in Column (2) of Table 4, the DID estimates are positive and statistically significant at the 5% level. This consistency with the baseline estimates underscores the robustness of our core findings to alternative model specifications.

Table 4. Endogeneity test.

	(1)	(2)
	IV	DID
${\mathit{CPP}}_{\mathit{it}}$	0.109 ***	0.019 **
	(0.026)	(0.009)
Controls	√	√
Firm FE	√	√
Year FE	√	√
Firm*Year FE	✕	✕
Observation	2111	2111

Note: The standard error is shown in parentheses. ***, ** indicates significance at the 1% and 5% levels. The √ symbol indicates the inclusion of the variable. The ✕ indicates that this variable is not included. Firm*Year FE represents the interaction fixed effects between firms and years.

Table 4. Endogeneity test.

	(1)	(2)
	IV	DID
${\mathit{CPP}}_{\mathit{it}}$	0.109 ***	0.019 **
	(0.026)	(0.009)
Controls	√	√
Firm FE	√	√
Year FE	√	√
Firm*Year FE	✕	✕
Observation	2111	2111

Note: The standard error is shown in parentheses. ***, ** indicates significance at the 1% and 5% levels. The √ symbol indicates the inclusion of the variable. The ✕ indicates that this variable is not included. Firm*Year FE represents the interaction fixed effects between firms and years.

5.4. Impact Channel: Market Expansion and Technological Innovation Channels

5.4.1. Market Expansion Channel

To verify that critical peak pricing enhances urban GTFEE through the market expansion channel, we examine three dimensions: investment expansion, market penetration, and fiscal revenue increase. In terms of research methods, this paper draws on the methods of Zhao et al. [57], Antoniades et al. [58], and Emmanuel et al. [59]. We operationalize the market expansion channel through three empirically tractable metrics: (1) city-level fixed asset investment (Invest), (2) total consumer goods retail sales (Retail), and (3) municipal government fiscal revenue (Revenue). The corresponding empirical tests examining these transmission mechanisms are reported in

Table 5. Mechanism verification: market expansion channel.

	(1)	(2)	(3)
	Invest	Retail	Revenue
${\mathit{CPP}}_{\mathit{it}}$	0.086 **	0.145 ***	0.112 ***
	(0.031)	(0.027)	(0.033)
Controls	√	√	√
Firm FE	√	√	√
Year FE	√	√	√
Observation	1911	1731	1711

Note: The standard error is shown in parentheses. ***, ** indicates significance at the 1% and 5% levels. The √ symbol indicates the inclusion of the variable.

.

The explained variable in Column (1) of Table 5 is the urban fixed asset investment (Invest). The results show that the coefficient of the policy variable ${\mathit{CPP}}_{it}$ is 0.086 and is significantly positive at the 5% level, indicating that critical peak pricing has significantly stimulated fixed asset investment in energy storage equipment manufacturing and renewable project development. According to investment decision-making theory, critical peak pricing effectively enhances the expected returns of green projects, such as energy storage and renewable energy, by widening the price difference between peak and off-peak hours. It directly encourages enterprises to increase investment in related fixed assets. Meanwhile, clear electricity price signals also prompt very energy-consuming enterprises to shift to self-generated and self-consumed renewable energy during peak electricity consumption periods, further promoting capital formation and capacity expansion in related industries [60].

The explained variable in Column (2) is the total retail sales of consumer goods (Retail). The coefficient on the policy variable $CP{P}_{it}$ is 0.145, which is significant at the 1% level, indicating that the policy significantly increases end-user consumption demand for energy-saving products, particularly energy storage devices. From the perspective of consumption theory, the price signal of rising electricity costs during peak hours, on the one hand, enhances the willingness of industrial and commercial users to purchase energy-saving equipment to reduce electricity expenses through the expected expenditure effect. On the other hand, it also uses the signal transmission effect to guide residents and consumers to increase their preference and purchase intensity for high-efficiency products, thereby effectively expanding the penetration scale and total sales volume of energy-saving related products in the consumer market [61].

The explained variable in Column (3) is urban fiscal revenue (Revenue). The coefficient of the policy variable $CP{P}_{it}$ is 0.112 and significantly positive at the 1% level, confirming that this policy has increased local governments’ fiscal revenue. This increase primarily stems from policy-driven investment expansion and heightened economic activity. These developments boost revenues from related taxes (such as value-added tax and corporate income tax) and land transfers, demonstrating the positive fiscal dividends generated by the economic expansion [62].

Mechanism analysis shows that the critical peak pricing policy has effectively stimulated market vitality by sending clear price signals. Specifically, it has significantly driven investment in green industries, promoted the consumption of energy-saving products, and simultaneously driven the growth of government fiscal revenue. This policy not only guides resources to concentrate on efficient and low-carbon fields through the price difference between peak and off-peak hours, but also creates a continuous demand for energy-saving technology products at the terminal consumption level, thereby simultaneously exerting positive impacts on investment, consumption, and finance. The above findings provide solid empirical support for the hypothesis that market expansion is the core transmission channel and further clarify the role of electricity price policies in stimulating the urban green transformation.

5.4.2. Technological Innovation Channel

To verify that critical peak pricing can enhance the urban GTFEE through technological innovation channels, this paper focuses on two dimensions for analysis: government support and enterprise innovation. Firstly, technological innovation activities, especially in green technology, have significant policy dependence, and government financial support usually plays a key guiding role [1]. Secondly, the direction of technological evolution is deeply influenced by market demand signals, and critical peak pricing guides enterprise innovation through price signals [61]. To operationalize these dimensions, we adopt the following measures. For government support, we construct a proxy variable following Fu et al. [63]: the ratio of government expenditure on R&D to total fiscal expenditure (denoted SE). To measure enterprise innovation output, we employ the count of patents granted, aggregated at the prefecture level (Patent), following the approach of Guo and Zhong [64]. Additionally, for green innovation output, we employ the count of green patents granted (G_patent), following the methodology of Jin et al. [65]. The estimation results for these variables and their roles as channels are presented in

Table 6. Mechanism testing: technological innovation channel.

	(1)	(2)	(3)
	SE	Patent	G_patent
${\mathit{CPP}}_{\mathit{it}}$	0.333 ***	0.096 *	0.094 **
	(0.092)	(0.056)	(0.050)
Controls	√	√	√
Firm FE	√	√	√
Year FE	√	√	√
Observation	1711	1751	1851

Note: The standard error is shown in parentheses. ***, **, * indicates significance at the 1%, 5%, and 10% levels. The √ symbol indicates the inclusion of the variable.

.

The explained variable in Column (1) of Table 6 is SE. The results show that the coefficient of the policy variable ${\mathit{CPP}}_{it}$ is 0.333 and is significantly positive at the 1% level. This suggests that critical peak pricing has enhanced local governments’ financial support for green technologies. From the perspective of public finance theory, the fiscal revenue increase effect brought about by policy implementation has effectively alleviated the budget constraints of local governments and provided necessary financial space for expanding green technology expenditures. Meanwhile, the long-term market signals released by the electricity price policy have also strengthened the government’s recognition of green technologies’ strategic value and positive externalities, thereby guiding it to optimize the fiscal expenditure structure and prioritize supporting related research and development activities [66].

The explained variable in Column (2) is Patent. At this time, the coefficients of ${\mathit{CPP}}_{it}$ is significantly positive at the 10% level, indicating that the policy has dramatically enhanced overall innovation activity. The arbitrage space formed by the peak–valley price difference has enhanced the market expected returns for energy storage and efficiency management, directly motivating enterprises to expand R&D investment [67].

The explained variable in Column (3) is G_patent. The coefficient on ${\mathit{CPP}}_{it}$ is positive and significant at the 5% level, demonstrating that the policy effectively channels innovation resources toward green low-carbon technologies. The critical peak pricing policy transmits a strong demand signal for low-carbon solutions by significantly increasing the cost of peak electricity. These price signals redirect innovative resources toward green low-carbon technologies, creating dual incentives for firms to reduce energy expenses and capture emerging market opportunities [68].

The research finds that critical peak pricing reduces risks in green technology R&D via strengthened government support. Simultaneously, its price signals enhance systemic innovation capacity while channeling innovation resources toward low-carbon technology fields. Therefore, Hypothesis 2 has been verified.

5.5. Heterogeneity Analysis

5.5.1. Heterogeneity Related to the Resource Endowment Effect

To deeply investigate the differentiated impact of critical peak pricing on cities with different resource endowments, this paper divides the samples into two groups based on the “National Sustainable Development Plan for Resource-based Cities (2013–2020)”—resource-based and non-resource-based cities—and conducts group regression. The results in in Columns (1) and (2) in Table 7 show that the policy coefficient of non-resource-based cities is 0.023, which is significantly positive at the 1% level. In contrast, the coefficient estimate for the policy variable in resource-based cities is not statistically significant. This discovery supports Hypothesis 3, indicating substantial resource endowment heterogeneity in the promoting effect of critical peak pricing on urban GTFEE. Specifically, in non-resource-based cities with diversified industrial structures, the price signals formed by the policy can be effectively transmitted. This reduces the uncertainty of enterprises’ energy consumption costs and encourages them to make long-term capital investments, effectively converting price pressure into energy efficiency improvements [69]. Meanwhile, the diversified industrial foundation provides broad application scenarios and risk dispersion channels for new technologies, further strengthening the policy effect [38].

Resource-based cities’ reliance on traditional energy sources has locked production factors onto high-carbon paths. As a result, enterprises are less sensitive to fluctuations in electricity prices, and the marginal cost of innovation and transformation is high. Furthermore, institutional and environmental constraints further suppress the market’s response ability to price signals. Resource endowment dependence and institutional constraints jointly constitute the “Double locking” effect that hinders efficiency improvement [70]. Therefore, based on the above analysis, critical peak pricing significantly improves the GTFEE of non-resource-based cities compared with resource-based cities.

5.5.2. Heterogeneity in Government Administrative Power

To examine the impact of heterogeneity in government administrative power on policy implementation, this paper follows Zhao et al. [45] and classifies the sample into two groups: provincial capital cities and non-provincial capital cities. This classification allows for an analysis of the differential effects of CPP on GTFEE across these two city types. The regression results in columns (3) and (4) of Table 7 indicate that the policy coefficient for provincial capital cities is 0.039 and statistically significant at the 1% level. In contrast, the coefficient for non-provincial capital cities is 0.013 and significant at the 10% level. These findings support this study’s research hypothesis, suggesting significant heterogeneity exists in the influence of administrative power levels on the promotion of urban GTFEE. Specifically, such disparity can be attributed to structural differences in administrative authority and policy implementation capacity among cities at different administrative levels in China. Provincial capital cities typically possess higher administrative ranks, more rigorous environmental supervision systems, and stronger enforcement capabilities. These characteristics enable them to ensure the effective implementation of critical peak electricity price policies through robust regulatory frameworks, supportive mechanisms, and performance evaluation systems, thereby achieving the intended policy outcomes. In contrast, non-provincial capital cities often encounter constraints such as limited regulatory resources, weaker enforcement capacity, and insufficient policy support mechanisms, which hinder the efficient promotion of CPP and prevent the full realization of its policy effects. The empirical findings of this paper thus confirm the heterogeneous role of administrative power levels in the implementation of environmental policies.

Table 7. Resource endowment and government administrative power heterogeneity.

	(1)	(2)	(3)	(4)
	Non-Resource-Based Cities	Resource-Based Cities	Provincial Capital Cities	Non-Provincial Capital Cities
${\mathit{CCP}}_{\mathit{it}}$	0.023 ***	0.016	0.039 ***	0.013 *
	(0.006)	(0.010)	(0.012)	(0.007)
Controls	√	√	√	√
Firm FE	√	√	√	√
Year FE	√	√	√	√
Observation	893	1218	312	1799

Note: The standard error is shown in parentheses. *, *** indicates significance at the 10% and 1% level. The √ symbol indicates the inclusion of the variable.

Table 7. Resource endowment and government administrative power heterogeneity.

	(1)	(2)	(3)	(4)
	Non-Resource-Based Cities	Resource-Based Cities	Provincial Capital Cities	Non-Provincial Capital Cities
${\mathit{CCP}}_{\mathit{it}}$	0.023 ***	0.016	0.039 ***	0.013 *
	(0.006)	(0.010)	(0.012)	(0.007)
Controls	√	√	√	√
Firm FE	√	√	√	√
Year FE	√	√	√	√
Observation	893	1218	312	1799

Note: The standard error is shown in parentheses. *, *** indicates significance at the 10% and 1% level. The √ symbol indicates the inclusion of the variable.

5.5.3. Heterogeneity Related to Regional Development Effects

To deeply explore the regional heterogeneity of critical peak pricing, this paper divides the sample cities into three groups based on geographical location: eastern, central, and western, and conducts group regression.

Table 8. Regional development affects heterogeneity.

	(1)	(2)	(3)
	Western Cities	Central Cities	East Cities
${\mathrm{CPP}}_{\mathrm{it}}$	−0.019	0.029 ***	0.025 ***
	(0.054)	(0.008)	(0.008)
Controls	√	√	√
Firm FE	√	√	√
Year FE	√	√	√
Observation	566	609	936

Note: The standard error is shown in parentheses. *** indicates significance at the 1% level. The √ symbol indicates the inclusion of the variable.

indicates that the policy effect, measured by the coefficient on ${\mathit{CPP}}_{it}$, is strongest in central cities. The coefficient for eastern cities is 0.025 and is significant at the 1% level, while the estimate for Western cities is statistically insignificant. This result provides strong evidence supporting our hypothesis that the effect of critical peak pricing exhibits significant regional heterogeneity.

Specifically, with their mature market, eastern cities ensure that price signals can sensitively and comprehensively reflect supply and demand changes and resource scarcity. Meanwhile, advanced power grid infrastructure has significantly reduced physical friction and loss during the energy transmission and distribution process, ensuring the efficient transmission of price signals in both time and space dimensions. Combining these two factors has significantly enhanced the efficiency of price signals in allocating resources. Consequently, enterprises are guided to optimize their energy consumption behavior and make energy efficiency investments [71,72]. Central cities have a solid traditional industrial foundation and significant potential for renewable development. The abundant endowment of new energy resources and the rapid growth of related industries enable enterprises to access low-cost green energy locally. This enhanced accessibility, in turn, strengthens their responsiveness to price signals and expands potential profit margins. Under the combined effect of these multiple favorable conditions, the policy effects of central cities are the most prominent. Although Western cities possess abundant renewable resources, their power market structure imperfections undermine prices’ capacity to accurately and dynamically reflect supply-demand and resource scarcity. This leads to price distortions that significantly undermine market efficiency. Furthermore, deficient power grid infrastructure (such as grid structure and intelligence level) and limited cross-regional transmission capacity have considerably increased the access cost, transmission loss, and system balancing expense for renewable energy. Therefore, Hypothesis 3 is verified.

6. Conclusions and Policy Implications

Based on a quasi-natural experiment of China’s critical peak pricing policy, this paper uses sample data from 273 cities in China from 2013 to 2022. We employ the double machine learning method to estimate its causal impact on urban green total factor energy efficiency (GTFEE) and the underlying transmission mechanisms. Empirical results show that critical peak pricing has significantly improved urban GTFEE through the dual paths of market expansion and technological innovation. Specifically, after the policy is implemented, the expanded peak–valley price spread has directly stimulated the widening investment scale in energy storage equipment, the increase in installed capacity of renewable projects, and a substantial growth in the number of authorized green technology patents. Heterogeneity analysis indicates that the critical peak pricing policy has a more significant promoting effect on non-resource-based, strong government administrative power, and central and eastern regions.

Based on the above conclusions, this paper provides the following policy suggestions for optimizing the critical peak pricing policy and increasing the urban GTFEE: First, strengthen the design of the critical peak pricing policy to fully leverage its role in enhancing the urban energy efficiency. The empirical results of this paper show that the critical peak pricing policy significantly promotes the GTFEE of cities. To improve the effectiveness of electricity price signals in optimizing energy allocation and guiding green decision-making, it is necessary to set further peak and off-peak periods and price difference levels in a refined and dynamic manner based on regional power supply and demand characteristics. For instance, in regions with a high penetration rate of renewable energy and significant consumption pressure, the peak–valley price difference can be moderately expanded to effectively motivate user-side energy storage and load transfer and enhance the system’s overall regulation capacity and energy utilization efficiency. During periods of tight supply, reasonably increase electricity prices further to optimize the response behavior between supply and demand and promote the continuous improvement of GTFEE. In addition, efforts should be made to enhance the coordinated integration of the electricity market with environmental policy tools such as the carbon market and energy consumption rights trading, promote the formation of an incentive mechanism where electricity price signals and environmental policies reinforce each other, and systematically improve the low-carbon transformation of energy throughout society.

Second, relying on the path of market expansion and technological innovation, enhance the effectiveness of policy transmission. Empirical research shows that the critical peak pricing policy effectively guides the behavioral adjustments of both supply and demand sides through price signals, enhancing short-term operational efficiency and promoting the green transformation of the long-term energy structure. To improve the market expansion effect, it is suggested that a user-side energy storage investment subsidy mechanism be introduced in the implementation area. For energy storage systems that continuously discharge during peak hours, appropriate operation subsidies should be provided based on the discharge volume, further stimulating users’ participation in demand response. To amplify the effect of technological innovation, priority should be given to supporting the research and development of technologies closely related to electricity price response, such as smart grids, virtual power plants, and high-efficiency energy-saving technologies. Tax incentives such as additional deductions for research and development expenses should be implemented for related enterprises, and policy support should be linked to actual energy efficiency improvement, carbon reduction performance, and other indicators to ensure that innovation incentives are precise and effective.

Third, implement regional differentiated policies and enhance the overall implementation effect in a coordinated manner. Heterogeneity analysis indicates that the critical peak pricing policy has more significant effects in non-resource-based cities, provincial capital cities, and central and eastern regions. The leading advantages from such areas should be further consolidated, and they should be supported to take the lead in piloting in deepening the electricity price signal and the design of supporting markets, to form typical experiences that can be replicated and promoted. At the same time, efforts should be made to enhance non-provincial capital cities’ policy reception and implementation capabilities. They should be encouraged to explore differentiated and refined electricity price mechanisms and dispatching models in light of their local industrial and energy structure characteristics. Moreover, technical cooperation and talent exchange mechanisms with provincial capital cities should be established to promote the transfer and diffusion of advanced practices and knowledge. For resource-based cities and the western regions, it is necessary to focus on making up for the shortcomings in systems and infrastructure. It is suggested that a targeted energy transition support fund be established, and part of the increased revenue from the critical peak pricing policy be used to upgrade distribution networks and construct distributed energy and energy storage facilities. At the same time, actively introduce advanced technologies and operation models from the central and eastern regions in areas such as intelligent dispatching and demand response management and enhance the energy system’s flexibility and green transformation capacity through cross-regional collaboration.

This study still has certain limitations and indicates the direction for in-depth research. Firstly, the characterization of policy intensity relies solely on binary variables and fails to reflect the continuous differences among cities. Future research can introduce continuous indicators, such as the actual peak–valley price difference amplitude, to construct a more precise measure of policy intensity, thereby more comprehensively identifying its effect. Secondly, this paper did not conduct robustness tests on non-desired output indicators. In the future, alternative environmental indicators such as industrial solid waste can be introduced to analyze the impact of different output settings on the efficiency measurement results, further enhancing the reliability of the conclusion. Furthermore, although the policy recommendations are based on empirical evidence, they can still be further improved in terms of persuasiveness and operability through quantitative simulation. For instance, power system optimization models or cost-benefit comparison methods can be adopted to compare the net benefits of critical peak pricing policy under different scenarios, thereby providing a more targeted basis for policy design. In addition, enterprise-level microdata can be further integrated, or methods such as questionnaires and in-depth interviews can be adopted to more meticulously depict and verify the specific transmission paths through which the critical peak pricing policy affects market expansion and technological innovation, providing evidence for mechanism analysis. Apart from the core factor of administrative power, the differences among cities in terms of the quality of administrative management and the intensity of regulatory enforcement may also be an important explanatory dimension leading to the heterogeneity of policy effects.