Two-Part Tariff Policy and Total Factor Productivity of Pumped Storage Industry: Stimulation or Failure?

Abstract

The two-part tariff (TPT) policy is implemented as an important initiative to accelerate the marketization of the pumped storage industry and promote its high-quality development. However, it is not clear exactly how the TPT policy affects the productivity of the pumped storage industry. Using the EBM-GML method and the DID model, this paper measures the total factor productivity of the pumped storage industry and explores the impact of the Two-Part Tariff (TPT) policy on its total factor productivity. Based on the samples of 16 provinces in China from 2004 to 2020, we find the following: (1) At present, the total factor productivity of China’s pumped storage industry is still at a low level. (2) TPT policy can promote the improvement of total factor productivity, which was strongly supported by the robustness test. Innovation incentive is one of the main mechanisms. (3) The impact of TPT policy on total factor productivity has obvious regional heterogeneity. By geographic location, the TPT policy has little effect on the pumped storage industry’s TFP in the eastern region, but it exerts a significant positive role in the central region. By energy affluence, TPT policy effect is stronger in provinces with low energy dependence. By environmental governance, the role of this policy is more obvious in provinces with low environmental regulation but developed green financial market. Finally, some corresponding policy implications have been put forward.

1. Introduction

In response to the global climate crisis, the development of the world‘s renewable energy has ushered in a huge leap in the 21st century. According to IEA statistics, global renewable energy generation reached 8349 TWh in 2022, accounting for nearly 30% of total generation [1]. As the largest carbon emitter and consumer of electricity, China has also made great strides in renewable energy generation. Over the past 20 years, China’s share of renewable power generation has risen steadily from 14.8% to 28.7%, basically reaching the global average, which greatly alleviated the pressure of carbon growth. However, as the scale of renewable energy generation continues to expand, some new problems have begun to emerge, such as its volatility, intermittency and unpredictability, leading to a large amount of energy abandonment [2]. Energy storage is an ideal solution to the instability of renewable energy generation [3,4,5]. As a result, the development of the energy storage industry has gradually become a focus of attention.

Pumped storage is currently the dominant form of energy storage [6]. Up to 2022, China has had more than 45 GW of installed pumped storage capacity built, accounting for 26.2% of the global total. However, it is still not enough to meet the huge demand for renewable power consumption. In 2022, China’s installed renewable energy capacity has exceeded 1200 GW and maintained a high growth rate of 14.1%. In order to improve the capacity of the pumped storage industry to absorb new energy power, China has put forward the Medium- and Long-term Development Plan for Pumped Storage, which calls for a more flexible and efficient development of the pumped storage industry. Swiftly increasing the total factor productivity (TFP) of the pumped storage industry has become an urgent task in China’s energy transition.

Price policy is often treated as a powerful tool for adjusting total factor productivity in monopolistic industries. According to the theory of perfectly competitive market, price marketization will improve the production efficiency of monopoly enterprises and increase social welfare [7]. Since 2001, to foster equitable competition and improve the efficiency of SOEs, the Chinese government has presented the overall objective of market-based price reform. Many of the traditional monopolistic industries have undertaken reforms of their price mechanisms, such as the electricity industry [8], the oil industry [9] and the iron and steel industry [10]. Among these reforms, there are not only success stories, but also some failures [9], which have a lot to do with the characteristics of the industry [11]. Price reform for the energy storage industry is relatively late. In 2014, China National Development and Reform Commission (NDRC) issued the Circular on Issues Related to Improving the Price Formation Mechanism for Pumped Storage Power Plants, which proposed a “Two-Part Tariff”(TPT) policy. This price mechanism clarifies the principle of categorized pricing, and encourages localities to adopt market-based pricing, which has a dramatic impact on the development of China’s pumped storage industry. In that case, does the TPT policy really improve the TFP of the pumped storage industry? The question remains unanswered.

In view of this, by using the panel data of 16 provinces in China from 2005 to 2020 and constructing a DID model, this paper empirically examines the impact of TPT policy on the TFP of the pumped storage industry. Our study may contribute to the existing literature in the following two points: (1) In terms of research objectives, we evaluate the productivity effect of price policy in the field of pumped storage industry for the first time, which provides new empirical evidence for the high-quality development of the pumped storage industry, and also expands the scope of industrial research on the relationship between price and productivity. (2) For the research methods, we introduce the EBM-GML method, which effectively overcomes the problems of setting bias and random error under the existing SFA or DEA efficiency measurement methods, and provides a new method choice for the efficiency measurement of other energy storage industries. Nowadays, with a new round of energy technology revolution around the world, the market prospect of pumped storage industry is broad. However, its current market operation mechanism is still relatively immature, especially the pricing mechanism. Pumped storage plants in major countries around the world still use traditional leasing mode or internal accounting mode, e.g., France, Japan and most states in the United States, while just few countries apply the two-part tariff mode. Accurately identifying the impact of TPT policy on total factor productivity is not only conducive to accelerating the high-quality development of China’s pumped storage industry, but also provides an important reference for other countries to formulate and improve their pumped storage price mechanism.

The subsequent structure of the article is arranged as follows. The second part is the literature review. The third part is the methodology and data. The fourth part is the development of China’s pumped storage industry. The fifth and sixth part is the empirical analysis. The last part is the conclusions and suggestions. The technology roadmap is shown in Figure 1.

2. Literature Review

2.1. The Impact of Price Policy on Total Factor Productivity

Existing studies have extensively explored the productivity effects of price policies around different industries, but still have not reached a unified conclusion. Some scholars argued that the implementation of price policy would be conducive to increasing total factor productivity. Taking the grain industry as an example, Jia et al. [12] found that the implementation of the minimum purchase price policy for wheat had a positive and significant effect on both the technical efficiency of wheat production and technological progress, thus contributing to total factor productivity in general. Zhao and Lin [11] conducted a survey of energy-intensive enterprises in China and emphasized that electricity price differentiation policy had increased the production cost of enterprises, which forced enterprises to speed up the scrapping and cleaning of backward production equipment with low technical level, thus improving production efficiency. Based on the data of mineral resource enterprises, Liu et al. [13] supported the market-oriented reform of natural resource prices to promote the total factor productivity of Chinese enterprises. Chen et al. [14], however, put the perspective on the renewable energy sector and found that the implementation of wind power price policy reduced the benchmark price of wind power, which effectively inhibited the blind investment of wind power industry and forced enterprises to carry out technological innovation, thus promoting the improvement of total factor productivity.

In contrast, other studies took the opposite view, thinking that tariff policies could exert a negative effect. Using samples of U.S. and Chinese firms, respectively, Gamtessa and Olani [15] and Ren et al. [16] found that energy price policy shocks had an adverse effect on firms’ total factor productivity. Zheng et al. [17] further examined the cases of energy-intensive industries, but obtained the opposite conclusion from Zhao and Lin [11]. They proposed that differential pricing policy suppressed firms’ total factor productivity, although this policy effect reversed after a few years. Qu et al. [10] obtained similar conclusions, arguing that differential tariff policies hampered firms’ scale effects, which led to a 1.20% per annum decrease in steel firms’ TFP. Moreover, based on cross-country data at the industry level, Gonseth et al. [18] explored the impact of energy tax policies on firms’ productivity, showing that rising energy prices did not have a simple boosting or dampening effect on firms’ TFP, but rather exhibited heterogeneity with the degree of human capital accumulation.

All of the above studies have shown that price policy does affect the total factor productivity of firms, which provides essential support for our study. However, most of these studies focused on traditional energy-intensive industries, and few studies paid attention to the renewable energy industries, especially the rapidly developing pumped storage industry. The question of how important internal factors, such as changes in benchmark tariffs caused by pricing policy, affect the TFP of the pumped storage industry remains unanswered.

2.2. The Efficiency of Energy Storage Industry

Whether the energy storage industry is economically efficient largely determines the enthusiasm of market investment and the effectiveness of price policy [19]. Therefore, some studies have begun to focus on the efficiency of the energy storage industry. According to the research objectives, these studies can be divided into the following two categories: efficiency measurement and efficiency optimization.

As for how to measure the efficiency of the energy storage industry, the cost–benefit analysis method is widely used [20,21]. The energy storage industry is generally considered to be inefficient due to high investment costs and immature market systems. Based on the data of energy storage enterprises in October 2013, Yu et al.’s research showed that the cost of LIB was about 4000 CNY/kWh, which was much higher than the expected market price of 1246 CNY/kWh [19]. Yang et al. [22] proposed the famous YCC index, and found that most electrochemical batteries were not economically profitable, except for long-lived PbAB and SCES. Using the SBM-DEA method, He et al. [23] measured the total factor productivity of 30 energy storage companies in China from 2017 to 2021, and their results also suggested that the overall efficiency of the energy storage industry was still at a relatively low level.

Scholars have also conducted considerable research on how to improve the efficiency of the energy storage industry, including the reform of the price mechanism. Zhang et al. [24] proposed that the optimization of pumped storage capacity price mechanism could effectively improve the utilization rate of renewable energy. Based on the double-layer stochastic optimization model and the assumed samples, Ma et al. [25] posited a competitive equilibrium pricing mechanism for wind power suppliers, pumped storage plants and intermediate vendors. Given the parameters of the storage system, the case study of Li et al. [26] also suggested that adopting a joint pricing mechanism of “wind + pumped storage” would help to improve the efficiency of the storage system. Beyond that, multilevel collaborative planning of energy storage systems [27,28], scheduling optimization [29] and full lifecycle planning [30] have been proven to be viable solutions to improve efficiency.

All of the above studies have made useful attempts to accurately calculate the efficiency and promote the development of the energy storage industry, but there are also some desirable improvements. On the one hand, the calculation method for the efficiency of the energy storage industry is still relatively simple, and there is a scarcity of measures in the pumped storage industry’s TFP. On the other hand, scholars have mostly explored the expected effects of improved policies and lacked an assessment of the effects of existing policies. Furthermore, the predominantly case-study-approach may also make the conclusions less applicable.

3. Methodology and Data

3.1. Model Setting

The Differences-in-Differences (DID) method is often used in policy effect evaluation studies. By separating the ex ante and time differences between the treatment group and the control group at the same time, the ‘treatment effect’ generated by exogenous policy shocks is obtained [31]. Drawing on the studies of Zheng et al. [17] and Qu et al. [10], we treated the TPT policy as a quasi-natural experiment and constructed a DID model to test the impact of TPT on total factor productivity in the pumped storage industry. Meanwhile, to alleviate the endogeneity problem caused by unobservable individual-varying factors and time-varying factors, we introduced both individual fixed-effects terms and time-fixed-effects terms. The baseline model is shown below:

$${TFP}_{it}={\beta }_0+{\mu }_i+{\nu }_t+{\beta }_1{DID}_{it}+{\gamma }^{\prime }{\mathit{X}}_{\mathit{i}\mathit{t}}+{\epsilon }_{it}$$

(1)

where TFP_it denotes the average total factor productivity of the pumped storage industry of region i in year t. β₀ is the intercept term. ${\mu }_i$ is an individual fixed effect term. v_t is a time fixed effect term. DID_it indicates whether the TPT policy is implemented. β₁ represents the impact of TPT policy on total factor productivity. X_it stands for the set of control variables. ${\epsilon }_{it}$ is a random error term.

After that, to ensure the validity of the DID estimation results, we conducted a test for parallel trends. This test aims to explore whether the time or trend effects are consistent between the experimental and control groups before the policy treatment. Referring to the study of Qu et al. [10], the dynamic parallel trend test equation is shown below:

$${TFP}_{it}={\sigma }_0+{\mu }_i+{\nu }_t+{\sum }_{m=-5}^{10}{\sigma }_{-\mathrm{m}}{DID}_{i,t-m}+{\gamma }^{\prime }{\mathit{X}}_{\mathit{i}\mathit{t}}+{\epsilon }_{it}$$

(2)

where m denotes the lag years of TPT policy implementation. When m is greater than 0 and ${\sigma }_{-m}$ is not significant, it indicates that the parallel trend test passes.

Finally, the TPT policy cancels the barriers for social capital to enter the pumped storage industry, and enhances the profitability of pumped storage power stations, which may change the financing quantity and technical level of related enterprises, thus affecting their total factor productivity. To clarify its mechanism, we constructed a two-step mediation effect model with reference to the study of Wang and Wang [32]. Its specific form is shown below:

$${Mid}_{it}={\alpha }_0+{\mu }_i+{\nu }_t+{\alpha }_1{DID}_{it}+{\delta }^{\prime }{\mathit{X}}_{\mathit{i}\mathit{t}}+{\epsilon }_{it}$$

(3)

$${TFP}_{it}={\tau }_0+{\mu }_i+{\nu }_t+{\tau }_1{DID}_{it}+{\tau }_2{Mid}_{it}+{\eta }^{\prime }{\mathit{X}}_{\mathit{i}\mathit{t}}+{\epsilon }_{it}$$

(4)

In the above equation, Mid_it denotes the mediating variable, while other variables have the same meaning as in Equation (1). The mediating effect exists only if both α₁ and ${\tau }_2$ are statistically significant.

3.2. Variables

3.2.1. Dependent Variable: Total Factor Productivity

The early total factor productivity referred to the contribution of factors other than labor and capital to output growth, mainly the rate of technological progress [33]. However, with the increase in the types of factor inputs (e.g., intermediate inputs) and outputs (e.g., non-desired outputs), its conceptual connotation has gradually been expanded to the ratio of comprehensive factor inputs to outputs, which includes a variety of factors, including labor, capital, energy, technology and institutions [34]. With reference to the study of Chen et al. [14], this paper defines total factor productivity in the pumped storage industry as the average output level per unit of various input factors in the pumped storage power generation process, i.e., the overall efficiency of transforming inputs into final outputs. There are currently two main methods for measuring TFP, stochastic frontier analysis (SFA) and data envelopment analysis (DEA) [14,32,35,36]. SFA utilizes the stochastic frontier production function for estimation and then decomposes the residuals to obtain TFP. DEA keeps the inputs or outputs of the decision units unchanged, determines the relatively effective production frontier based on linear programming and statistical data, and then measures the degree (distance) of the decision units’ deviation from the frontier to reflect the TFP. Errors in SFA arise mainly from biased production function settings, whereas those in DEA are due to the failure to account for the effects of random errors [37]. The energy storage industry is still a relatively young industry in China, and its development is in the transition period from the R & D and demonstration stage to the marketization stage [23]. Therefore, the main components and driving factors of its production may change constantly, which leads to the variability of its production function in setting. For this reason, we ultimately chose the DEA method. Furthermore, compared with the traditional DEA model and SBM model, the EBM model contains both radial and SBM distance functions, which provides a more accurate measure of the actual TFP [38]. Since DEA cannot accurately reflect the changes in total factor productivity, referring to the study of Wang and Wang [32], we introduced the GML index and finally constructed the EBM-GML model. Equation (5) is the formula for the EBM model and Equation (5) is the formula for the GML index.

$${\gamma }^{*}=min\frac{\theta -{\epsilon }_x{\sum }_{i=1}^m\frac{{w_i^{-}s}_i^{-}}{x_{ik}}}{\phi +{\epsilon }_y{\sum }_{r=1}^s\frac{{w_r^{+}s}_r^{+}}{y_{rk}}+{\epsilon }_u{\sum }_{p=1}^q\frac{{w_p^{u-}s}_p^{u-}}{u_{pk}}}$$

$$s.t.\left\{\begin{array}{c}{\sum }_{j=1,j\ne k}^n{x}_{ij}{\lambda }_j+s_i^{-}-\theta x_{ik}\le 0,i=1,2,\dots ,m\\ {\sum }_{j=1,j\ne k}^n{y}_{rj}{\lambda }_j-s_r^{+}-\phi y_{rk}\ge 0,r=1,2,\dots ,l\\ {\sum }_{j=1,j\ne k}^n{u}_{pj}{\lambda }_j+s_p^{+}-\phi u_{pk}\le 0,p=1,2,\dots ,q\\ {\lambda }_j\ge 0,s_i^{-},s_r^{+}{,s}_p^{+}\ge 0\end{array}\right.$$

(5)

where ${\gamma }^{\mathrm{*}}$ is the optimal efficiency value, satisfying $0\le {\gamma }^{\mathrm{*}}\le 1$; k denotes the decision-making unit (DMU); m and l are the total number of inputs and outputs respectively; $w_i^{-}$, $w_r^{-}$, $w_p^{u-}$ distributions refer to the weights of each input, expected output and unexpected output; x_ik, y_rk, u_pk stand for the i-th input, r-th expected output, p-th unexpected output of the k-th decision unit, respectively; $s_i^{-},s_r^{+}{,s}_p^{u-}$ are the slack variables; $\theta$ and $\phi$ are the radial components of ${\gamma }^{\mathrm{*}}$; $\epsilon$ represents the core parameter for the transformation between radial and non-radial. When $\epsilon =0,$ it is a radial model; when $\epsilon =0,$ it is an SBM model; and when $0<\epsilon <1,$ it means that the model has both radial and non-radial slack variable components.

GML index is calculated as follows:

$$GM{L}_{t,t+1}={\left[\begin{array}{c}\frac{1+D_0^t\left(x_i^{i,t},y_i^{i,t},u_i^{i,t};y_i^{i,t}-u_i^{i,t}\right)}{1+D_0^t\left(x_i^{i,t+1},y_i^{i,t+1},u_i^{i,t+1};y_i^{i,t+1}-u_i^{i,t+1}\right)}\\ \times \frac{1+D_0^{t+1}\left(x_i^{i,t},y_i^{i,t},u_i^{i,t};y_i^{i,t}-u_i^{i,t}\right)}{1+D_0^{t+1}\left(x_i^{i,t+1},y_i^{i,t+1},u_i^{i,t+1};y_i^{i,t+1}-u_i^{i,t+1}\right)}\end{array}\right]}^{\frac{1}{2}}$$

(6)

where $D_0^t\left(x_i^{i,t},y_i^{i,t},u_i^{i,t};y_i^{i,t}-u_i^{i,t}\right)$ and $D_0^{t+1}\left(x_i^{i,t},y_i^{i,t},u_i^{i,t};y_i^{i,t}-u_i^{i,t}\right)$ represent the current distance function at period t and t + 1, respectively. $D_0^t\left(x_i^{i,t+1},y_i^{i,t+1},u_i^{i,t+1};y_i^{i,t+1}-u_i^{i,t+1}\right)$ denotes the period t + 1 mixing distance function based on period t technology. $D_0^{t+1}\left(x_i^{i,t+1},y_i^{i,t+1},u_i^{i,t+1};y_i^{i,t+1}-u_i^{i,t+1}\right)$ stands for the period t mixing distance function based on the period t+1 technology. GML > 1 indicates total factor productivity growth, and GML < 1 suggests a decline in total factor productivity.

Finally, in terms of the input–output index system, we selected five input indicators and one output indicator. All the input–output indexes of the EBM-GML model are shown in

Table 1. Description of input-output variables.

Index	Variable	Variable Description	Literature Source
Input indexes	Capital input	Net value of fixed assets (CNY 100 million)	[21]
	Labor input	Number of employees on board (people)	[8]
		Remuneration of employees on board (CNY 100 million)	[10]
	Innovation input	The research and test expenses of pumped storage power plants (CNY million)	[21]
	Intermediate input	Pumping power consumption (100 million kWh)	[21]
Expected output indexes	Pumped storage power generation	Pumping power generation (100 million kWh)	[14]

. Notably, non-expected outputs are not considered in this paper for the time being due to data availability constraints.

3.2.2. Independent Variable

In this paper, the TPT dummy variable is selected as a proxy variable for the pumped storage tariff policy. According to the Circular on Issues Related to Improving the Price Formation Mechanism of Pumped Storage Power Plants, China’s TPT policy was promulgated on 31 July 2014 and implemented on 1 August. However, taking into account that policy effects tend to play out with a certain lag, this paper treats the policy time with a half-year lag, i.e., 2015 is taken as the actual time point for policy treatment. Specifically, if a region implemented a TPT policy in 2015 and the year after, this variable (DID) is equal to 1, otherwise it is equal to 0.

3.2.3. Variable Used in Mechanism Analysis

Increased innovation inputs may be an important mechanism for TPT policies to drive TFP improvements in the pumped storage industry. Firstly, the TPT policy pushes for an appropriately proportional increase in electricity prices, which raises the average profitability of the industry and thus stimulates the incentive to innovate and increase innovation inputs in pumped storage power plants. Secondly, innovation inputs are a direct influence on increasing total factor productivity. The more innovation inputs, the higher the innovation level of enterprises, which means the higher TFP. Berchicci’s study shows that innovation inputs are most closely related to R&D expenditure [39]. Therefore, referring to his study, this paper adopts the research and test expenses of pumped storage power plants (inno) to measure the level of innovation input.

3.2.4. Control Variables

Total factor productivity in the pumped storage industry is often influenced by many other factors. In order to eliminate the endogeneity error triggered by industry characteristics and regional characteristics, and to obtain the net effect of the TPT policy on the TFP of the pumped storage industry, the following control variables are added to the model: (1) Renewable energy development [40], represented by wind power generation (wind) and solar power generation (solar). The main objective of the development of the pumped storage industry is to absorb a large-scale and high proportion of excess electricity from renewable energy sources, so the development of renewable energy is closely related to the development of local pumped storage. (2) Equipment capability. Scale economies is one of the important ways to improve efficiency. In general, a larger installed base means that its average cost will be lower, leading to a higher efficiency. In this paper, the number of installed units (num) and the installed volume (vol) of pumped storage power plants are selected to measure the equipment capacity. The equipment capacity of each province’s pumped storage industry is obtained by summing up the data of its power plants. (3) Industrial human capital [41]. According to Schumpeter’s theory of innovation, the more intensive an industry’s human capital factors are, the more productive and innovative that industry will be. In this paper, the investment in employee education at pumped storage power plants (hum) is used to measure the human capital of the industry as a whole. (4) Regional economic development [13,14]. The regional economic level directly determines the environment for industrial development. A higher economic level will be more conducive to industrial upgrading and high-quality development. Consistent with most studies, this paper still uses urbanization rate (urb) and per capita GDP (pgdp) to measure regional economic development. (5) Regional industrial structure (ind), expressed by the ratio of value added in the secondary sector to value added in the tertiary sector [42]. (6) Regional energy structure [14], measured by share of coal consumption (ene) and share of thermal power generation (ele). (7) Regional innovation environment [43], measured by number of university students per 100,000 population (edu) and number of patents granted per 10,000 population (pat). (8) Regional openness (ope), as measured by total import and export trade as a share of GDP [44]. All variables used in this paper are shown in

Table 2. The definitions and explanations of variables.

Types	Variables	Symbols	Definitions	Literature Source
Dependent variable	Total factor productivity in the pumped storage industry	TFP	EBM-GML index	[38]
Independent variable	Whether TPT policy was implemented	DID	Equals to 1 if the region implemented a TPT policy in this year, otherwise equals to 0.	[14]
Mediating variable	Innovation input of pumped storage industry	inno	The research and test expenses of pumped storage power plants (CNY million)	[39]
Control variables	Renewable energy development	wind	Wind power generation (100 million kw·h)	[40]
		solar	Solar power generation (100 million kw·h)	[40]
	Equipment capability	num	The number of installed units of pumped storage power plants (set)	-
		vol	The installed volume of pumped storage power plants (10 thousand kw·h)	-
	Industrial human capital	ham	The investment in employee education at pumped storage power plants (CNY million)	[41]
	Regional economic development	urb	Urbanization rate (%)	[13]
		pgdp	Per capita GDP (CNY/person)	[14]
	Regional industrial structure	ind	The ratio of value added in the secondary sector to value added in the tertiary sector (%)	[42]
	Regional energy structure	ene	Share of coal consumption (%)	[14]
		ele	Share of thermal power generation (%)	[14]
	Regional innovation environment	edu	Number of university students per 100,000 population (person)	[43]
		pat	Number of patents granted per 10,000 population (item)	[43]
	Regional openness	ope	Total import and export trade as a share of GDP (%)	[44]

.

3.3. Description of Sample

Based on the availability and authority of pumped storage industry and power plants data, this paper finally selects 16 provinces in China from 2004 to 2020 as research samples. The data of pumped storage industry in each province are derived from China Water Power Generation Yearbook and China Electric Power Yearbook. The data of wind and solar power generation are sourced from China Energy Statistics Yearbook. The data of control variables come from China Regional Economic Statistics Yearbook and China Urban Statistics Yearbook. Missing data are filled in by interpolation. In order to eliminate the interference of outliers, we tailed all variables except DID by 1%. The descriptive statistics of all variables are shown in

Table 3. Descriptive statistics o variables.

Variable	Obs	Unit	Mean	Std. Dev	Min	Max
TFP	188	-	1.047	0.248	0.442	2.481
DID	188	-	0.484	0.501	0.000	1.000
inno	188	CNY 106	2.332	1.951	0.017	8.767
wind	188	108 kwh	57.443	70.130	0.000	317.660
solar	188	108 kwh	24.201	44.528	0.000	206.450
num	188	set	5.968	4.485	2.000	24.000
vol	188	104 kwh	4.691	0.849	2.079	6.590
ham	188	CNY 106	0.774	0.438	0.290	2.381
urb	188	%	59.100	11.800	36.00	86.600
pgdp	188	CNY 104/person	5.530	2.752	1.005	16.422
ind	188	%	105.600	38.500	19.400	199.000
ene	188	%	92.000	41.700	2.500	246.100
ele	188	%	82.600	16.000	39.800	99.700
edu	188	%	2.800	1.100	1.400	6.800
pat	188	piece	13.296	14.487	0.366	61.161
ope	188	%	39.800	40.700	5.100	158.400

.

4. Development of China’s Pumped Storage Industry

4.1. Price Mode

Looking at the development of China’s pumped storage industry over the past two decades, its price mechanism or mode has gone through roughly two phases, the leasing mode and the two-part tariff mode. Understanding two patterns will help us to find out how the TPT policy role plays out.

(1)

Leasing mode (2004–2014). In 2004, NDRC issued the Notice on Issues Related to the Construction and Management of Pumped Storage Power Stations, stipulating that pumped storage power plants are mainly constructed and managed by power grid operating enterprises. In this mode, the pumped storage power plant is built by an independent power generation enterprise, and the grid operator pays a fixed annul lease fee to the pumped storage power plant, which is approved by the national price department. Correspondingly, electricity prices are no longer implemented, and social capital is not allowed to enter.

(2)

TPT mode (2014–present): In 2014, NDRC issued the Circular on Issues Related to Improving the Price Formation Mechanism for Pumped Storage Power Plants, which proposed a price mechanism for pumped storage power plants, a Two-Part Tariff. The TPT consists of two components, the capacity tariff and the electricity tariff. Specifically, capacity tariffs primarily reflect the value of ancillary services such as standby, frequency regulation, phase regulation and black start provided by pumped storage power plants, and are intended to compensate for the fixed costs of the plant, including construction investment, loan repayments and depreciation. It is usually calculated by the transformer capacity or maximum demand, independent of actual generation. The electricity price is mainly reflected in the benefit of peak shaving and valley filling realized by pumped storage power station, so as to make up for the variable cost such as pumped power loss. Its on-grid price equals to the benchmark on-grid price of local coal-fired units, and the pumping price is 75%. More importantly, for some economically developed regions, it is encouraged to use market-based approaches to price formation. In 2021, the launch of the Opinions on Further Improving the Price Formation Mechanism for Pumped Storage puts higher emphasis on the price marketization.

4.2. Plants Distribution and TFP

By the end of 2020, China had a total of 26 built pumped storage power stations, widely distributed in 16 provinces and municipalities across the country. For a clearer presentation, we used ArcGIS 10.7 to map the spatial distribution of its number of sites and total factor productivity, as shown in Figure 2.

From these maps, some important findings can be obtained. (1) There is a serious imbalance in the distribution of pumped storage power stations in China. As shown in Figure 2a, All the pumped storage power stations were built in the eastern and central regions, but none in the western region. As the most abundant area of wind and solar energy resources in China, the western region should become the core area of pumped storage development, but it is far from the reality. Why does this happen? This can be explained by the geo-climatic conditions and demand for electricity in the western provinces. Firstly, the site selection of pumped storage power stations usually requires sufficient water resources and appropriate terrain gaps. However, the western region of China is either arid and semi-arid areas or high-altitude areas, which brings great difficulties in technical conditions, so it is not conducive to the construction of pumped storage power stations. Secondly, compared with the eastern and central regions, the manufacturing industry in the western region is relatively backward and the population density is relatively low, so the local demand for electricity and power transfer is not high, which also reduces the local government’s willingness to build expensive pumped storage power stations. (2) The total factor productivity of China’s pumped storage industry is still at a low level. In 2020, the average TFP of pumped storage industry in China’s provinces is only 1.0282, slightly higher than 1. Among them, half of the provinces’ TFP is less than 1, which means that the operational efficiency is declining in a large area. The COVID-19 epidemic may be one of the influencing factors. (3) Total factor productivity has a certain relationship with regional economic level. As shown in Figure 2b, Beijing, Jiangsu, Zhejiang and Shandong have higher TFP, while other regions are relatively low. Generally speaking, the higher the economic level of the region, the richer the innovation resources and capital elements, the more conducive to the generation or introduction of more advanced technology and management experience. Furthermore, it is even more noteworthy that all these provinces are the ones that are encouraged to marketize electricity prices in the TPT policy, so does the TPT policy also contribute to higher TFP? This question is subject to further empirical inquiry.

5. Empirical Analyses

5.1. Paralleled Trend

A basic precondition for DID identification is the parallel trend assumption, i.e., that the treatment and control groups have followed a similar trend prior to the start of the TPT policy. If this condition is not met, the coefficients derived from the two differentials do not stand for the net effect of the policy. Drawing on Liu et al. [45], we performed a dynamic parallel trend test. As shown in Figure 3, all estimated coefficients are not significant in the nine periods before the implementation of the TPT policy, but turn positive in the first year after the policy is implemented. This shows no obvious difference in trends between the treatment and control groups over the period 2006–2015, passing the parallel trend test.

5.2. Results and Analysis of Baseline Regression

Since we have selected a total of 13 control variables, this may lead to multicollinearity problems, resulting in the OLS estimator no longer being valid and the estimation results being distorted. To exclude this influence, we conducted a multicollinearity test in advance, as shown in the

Table 4. The effect of TPT policy on TFP.

Variable	Dependent Variable: TFP
	OLS		FE		Dynamic FE
	(1)	(2)	(3)	(4)	(5)	(6)
L.TFP					−0.0695 **	−0.1140 **
					(0.0278)	(0.0448)
DID	0.0025	0.0901 *	−0.1150	0.5620 **	0.0681	0.5540 *
	(0.0391)	(0.0487)	(0.1500)	(0.2330)	(0.0431)	(0.2990)
Constant	1.0460 ***	1.2340 ***	1.1430 ***	1.1360	1.0370 ***	1.2160
	(0.0276)	(0.3180)	(0.1330)	(0.8610)	(0.0569)	(0.8820)
Control	NO	YES	NO	YES	NO	YES
vif	-	8.00	-	8.00	-	8.04
Province FE	NO	NO	YES	YES	YES	YES
Year FE	NO	NO	YES	YES	YES	YES
N	188	188	188	188	172	172
R2	-	-	0.124	0.200	0.153	0.221

Notes: Robustness standard error is presented in parentheses. *, ** and *** represent significances at 10%, 5% and 1% levels, respectively.

. The results show that the mean VIFs of the three models are 8.00, 8.00 and 8.04, respectively, all of which are significantly less than 10, indicating that there is no serious multicollinearity problem.

Table 4 is the estimation result of Equation (1). According to whether the control term and TFP lag term are introduced, we divide them into three categories: OLS, FE and dynamic FE. The estimation results without and with control variables are presented in the odd and even models, respectively. The results show that in the odd models, the estimated coefficients of DID are not significant. However, in the even models, the estimated coefficients of DID are significantly positive at the 5% level. This suggests that the TPT policy has a positive effect on the total factor productivity of the pumped storage industry, and this effect needs to be observed when the price policy is considered together with other factors, which justifies the introduction of extensive control variables.

Welfare economics theory holds that price distortion will lead to a decrease in welfare, resulting in a loss of efficiency [46]. The main reason why the TPT policy can increase total factor productivity in the pumped storage industry is that it facilitates tariff marketization and alleviates tariff distortions. For a long time in the past, the price mechanism implemented by China’s pumped storage industry has been a leasing mode. In this mode, the electricity price is strictly controlled by the superior department or the power grid company, and does not change with the market law of supply and demand. For example, in the peak and trough periods of electricity consumption, prices are still the same. This directly leads to the loss of profits of the plant enterprises, resulting in the mismatch of power resources, so that the overall resource allocation level of the industry is in a low state. According to our calculation results, from 2004 to 2014, the average total TFP of China’s pumped storage industry was only 1.0244, with a cumulative growth rate of −5.06%. The implementation of TPT policy improves this situation. The policy encourages plant enterprises to price in a market-oriented way and allows social capital to enter, which greatly stimulates the enthusiasm of plant power generation and the competition between plants, thus effectively improving the utilization level of industrial resources. The actual data also support this conclusion. From 2015 to 2020, the total TFP of China’s pumped storage industry achieved significant growth, from 0.9893 to 1.1856, an increase of nearly 20%.

5.3. Robustness Test

The empirical results in the benchmark regressions reveal that the PTP policy significantly increases total factor productivity. To verify whether the results are robust, the paper conducts a series of robustness tests.

5.3.1. Replace the Dependent Variable

For the total factor productivity measure, drawing on Chen et al.’s [14] study, we used pumped storage generation as the output variable, which reflects the core function of supplying electricity from pumped storage power stations. However, in other studies [10,13,17], firm revenue or industry output is more frequently used, which better reflects the general function of the firm. In this regard, we remeasured the new TFP using the main business income of the pumped storage plant as the output variable and used it as a proxy variable for robustness test. As shown in column 1 of

Table 5. The results of robustness test.

Variables	Dependent Variable: TFP
	Replace Dependent Variable	Exclude Special Samples	Use Cluster Robust Standard Errors
	(1)	(2)	(3)
DID	0.1990 *	0.2660 *	0.3550 **
	(0.1060)	(0.1460)	(0.1450)
Constant	1.0820 **	1.3810 *	1.4720 *
	(0.3850)	(0.7740)	(0.7450)
Control	YES	YES	YES
Province FE	YES	YES	YES
Year FE	YES	YES	YES
N	188	172	188
R2	0.158	0.203	0.191

Notes: Robustness standard error is presented in parentheses. *, ** represent significances at 10%, 5% levels, respectively.

, the estimated coefficient of DID is 0.1990 and is significantly positive at the 10% level, which suggests that the TPT policy has a positive effect on both revenue-oriented TFP and generation-oriented TFP, proving the robustness of the baseline regression result.

5.3.2. Exclude the Influence of Special Samples

As we all know, a COVID-19 epidemic broke out in December 2019 in China. For reducing the spread of the epidemic, the Chinese government enforced strict personnel control, which had a significant impact on the national economy and also seriously interfered with the normal construction and operation of the pumped storage industry. The addition of sample data from epidemic period may make the baseline regression results deviate from the normal situation, thus generating some errors. To eliminate this influence, this paper excludes all sample data during the epidemic, mainly for the 2020 sample data. As shown in column 2 of Table 5, the estimated coefficient on DID remains significantly positive, suggesting that the impact of TPT policies on TFP is robust.

5.3.3. Use Cluster Robust Standard Errors

Robust standard errors are included in the benchmark regressions to correct for heteroskedasticity at the individual province level [47]. However, there may be some heteroskedasticity issues at the broader regional level as provinces develop in a regional setting. In response, this paper introduces clustered robust standard errors in the benchmark regression to obtain more robust results. As can be seen in column 3 of Table 5, the dependent variable is still significantly positive, demonstrating that the positive effect of the TPT policy on pumped storage TFP is robust.

5.3.4. Placebo Test

Finally, to avoid the TPT treatment effect coming from other random factors, referring to Cao and Chen [48], we conducted two types of placebo tests. (1) In-time placebo test, i.e., the policy time is advanced and lagged by a few periods, respectively, and then DID estimation is performed. (2) In-individual placebo test, i.e., randomly selecting some individuals from the sample as “ fake treated unit” without any put-back, and then conducting DID estimation. This process is repeated 500 times to obtain the distribution of the estimated coefficients.

As shown in Figure 4, after performing 500 random sample regressions, the individual obeys a clear normal distribution and most of the points are concentrated near the estimated value 0. The p values of the two-sided test and the right-hand test are both less than 0.05, rejecting the null hypothesis that the treatment effect is 0, which indicates that the baseline regression results are robust. The estimation results of

Table 6. In-time placebo test results.

Variables	Dependent Variable: TFP
	Two Years Ahead	One Year Ahead	One Year Lag	Two Years Lag
	(1)	(2)	(3)	(4)
DID	0.0447	0.0223	0.1100 **	0.1580 ***
	(0.0457)	(0.0517)	(0.0463)	(0.0479)
Constant	3.5610 *	3.4950 *	3.1400 *	3.8270 *
	(1.6500)	(1.7290)	(1.6580)	(1.8560)
Control	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
Year FE	NO	NO	NO	NO
N	125	125	125	125
R2	0.249	0.248	0.259	0.265

Notes: Robustness standard error is presented in parentheses. *, ** and *** represent significances at 10%, 5% and 1% levels, respectively.

show that the estimated coefficients of DID in the one-year and two-year ahead are not significant, but both the lag term are significantly positive, indicating that the effect of TPT policy is not caused by random factors, and also supports the robustness of the conclusion.

5.4. Endogeneity Test

The selection of sites for pumped storage power plants in China is not random and generally requires an ex ante project performance assessment. Only sites that are expected to perform well are built, which may lead to endogenous problems. To mitigate this problem, we performed an instrumental variable estimation. Referring to Bellemare et al. [49], the lagged one-period term of the dependent variable is selected as the instrumental variable in this paper and 2SLS regression is conducted, as shown in

Table 7. Endogeneity test results.

Variables	1-Stage Regression	2-Stage Regression
	Dependent Variable: DID	Dependent Variable: TFP
	(1)	(2)
IV	0.4055 ***
	(0.0622)
DID		0.1393 ***
		(0.0687)
Constant	0.2847	2.2937 ***
	(1.3563)	(0.5252)
Control	YES	YES
N	172	172
R2	0.795	0.305
1-stage regression F statistic	93.51
Cragg–Donald Wald F statistic	34.57
Kleibergen–Paap Wald F statistic	43.89
Kleibergen–Paap rk LM statistic	8.93

Notes: Robustness standard error is presented in parentheses. *** represent significances at 1% level.

.

In order to ensure the validity of the instrumental variable, this paper conducted a series of tests. Firstly, the Cragg–Donald Wald F statistic and Kleibergen–Paap Wald F statistic are 34.57 and 43.89, respectively, which are much larger than 10, which excludes the ‘weak instrumental variable’ problem at a significant level of 10%. Secondly, the unrecognizable test Kleibergen–Paap rk LM statistic is 8.93, and passes the significance test at the 1% level, so the ‘unrecognizable problem’ is excluded. Furthermore, the Hansen J statistic passes the significance test at the 1% level, meaning that the model is just identified, so the instrumental variable meets the exogenous conditions. All above test results strongly support that the instrumental variable is valid. From the regression results of the second stage, after using IV to solve the endogenous problem, the impact of TPT policy on total factor productivity is still significantly positive, which is consistent with the benchmark regression results.

5.5. Mechanism Analyses

Based on the results in 5.2, this paper investigates how the TPT policy affects TFP.

Table 8. Mechanism analysis: how does TPT policy affect TFP.

Variables	Mediating Variable: Innovation Input
	(1)	(2)	(3)	(4)
	Inno	TFP	Inno	TFP
DID	1.2980 ***	0.0757 ***	4.5180 *	0.1100
	(0.3450)	(0.0258)	(2.2370)	(0.1230)
inno		0.0033 **		0.0110 *
		(0.0055)		(0.0056)
Constant	−6.8010 **	1.2480 ***	−3.2380	1.7030 *
	(3.1360)	(0.2210)	(4.3020)	(0.8480)
Control	YES	YES	YES	YES
Province FE	NO	NO	YES	YES
Year FE	NO	NO	YES	YES
N	188	188	188	188
R2	-	-	0.416	0.150

Notes: Robustness standard error is presented in parentheses. *, ** and *** represent significances at 10%, 5% and 1% levels, respectively.

shows the estimation results for Equations (2) and (3). Columns 1 and 2 are mixed OLS regression results, while columns 3 and 4 are two-way fixed effects model regression results.

Firstly, in column 1 and column 3, the estimated coefficients of DID are 1.2980 and 4.5180, respectively, and both are significantly positive at the 10% level, indicating that TPT policy has a positive effect on the increase of innovation investment. Secondly, in column 2 and column 4, the estimated coefficients of inno are 0.0033 and 0.0110, respectively, and both are significantly positive at the 10% level, suggesting that innovation investment can contribute to the total factor productivity of pumped storage industry. According to the results of columns 3 and 4, the implementation of TPT policy in a province will increase the R&D expenses of pumped storage power plants by an average of CNY 4.518 million, and for every CNY 1 million increase in these expenses, TFP will increase by 0.0110. Combining the two results, we can see that TPT policy can improve total factor productivity by stimulating innovation investment.

It is not difficult to understand why innovation can improve total factor productivity. According to Schumpeter’s innovation theory, innovation refers to a new production function, which aims to obtain more profits by recombining production factors. Therefore, from the perspective of input and output, innovation means the improvement of efficiency [50]. However, why does TPT policy stimulate innovation? It can be explained from two aspects: the benefits and costs of innovation. For innovation benefits, under the TPT mode, the government’s control over electricity prices is weakened, and some site enterprises are allowed to have certain independent pricing power. Enterprises can reduce power generation costs and their own power supply prices through technological or management innovation to obtain more market benefits. Therefore, it is profitable for site enterprises to innovate, which makes more enterprises have the willingness to innovate. For the cost of innovation, enterprises often have to pay high investment costs to innovate, which requires sufficient financial support. In the leasing mode, the operating funds of the site enterprises are all derived from the state-owned capital, with a single channel and a limited number. Therefore, most of the site enterprises do not have the actual innovation ability. However, the TPT policy has broken this restriction. The policy has set a precedent for social capital to enter the pumped storage industry in China, which opens up a new important channel for the financing of site enterprises, greatly relieves the cost pressure of innovation, and ultimately improves the possibility of innovation realization. The improvement of innovation willingness and the decline of innovation pressure jointly stimulate the innovation activities of enterprises, which leads to increased investment. This interpretation is well supported by the actual data.

Table 9. Innovation inputs for pumped storage plants from 2005 to 2020.

Year	2005	2006	2007	2008	2009	2010	2011	2012
Innovation input (CNY million/station)	1.16	1.18	1.20	1.13	1.08	1.14	1.20	1.31
growth rate of innovation input (%)	3.47	1.30	2.14	−5.94	−4.86	6.08	5.29	8.67
Year	2013	2014	2015	2016	2017	2018	2019	2020
Innovation input (CNY million/station)	1.32	1.43	1.56	1.64	1.80	1.69	1.87	1.97
growth rate of innovation input (%)	1.30	8.05	9.52	4.98	9.89	−6.12	10.59	5.18

shows the average innovation input per station over the years. From the data, we can see that in 2014, there was a remarkable increase in innovation input, with a growth rate of 8.05%, much higher than the 1.30% in 2013.

6. Further Analyses

The pumped storage industry in Chinese provinces has extensive disparity in the development level and environment, which may lead to significant differences in the effects of TPT policy application. In this regard, this paper further explores the regional heterogeneity of TPT policy affecting total factor productivity in the pumped storage industry.

6.1. The Heterogeneity of Geographic Location

As shown in 4.2, there are distinct differences in the development of the pumped storage industry in the eastern, central and western regions of China, so it is necessary to take into account the heterogeneity of geographic location. As there are no built pumped storage plants in the western region so far, all samples are only divided into two regions, the east and the central. The results are reported in column 1 and 2 of

Table 10. The results of heterogeneity analysis.

Variables	Dependent Variable: TFP
	Geographic Location		Energy Dependence		Environmental Regulation		Green Finance
	East	Central	High	Low	High	Low	High	Low
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
DID	−0.2390	0.5710 **	−0.2660	0.4930 *	0.1750	0.2740 *	0.5840 *	0.0018
	(0.1690)	(0.2090)	(0.9360)	(0.2120)	(0.4640)	(0.1320)	(0.2890)	(0.3090)
Constant	2.2850	1.4800	−6.9780	0.2170	1.5880	−0.4610	−0.7240	1.4320
	(1.9830)	(1.3420)	(5.8520)	(1.8450)	(1.5350)	(1.8130)	(2.4310)	(1.0690)
Control	YES	YES	YES	YES	YES	YES	YES	YES
Province FE	YES	YES	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES	YES	YES
N	125	63	106	82	87	101	111	77
R2	0.253	0.755	0.239	0.620	0.5470	0.3060	0.2990	0.6290

Notes: Robustness standard error is presented in parentheses. *, ** represent significances at 10%, 5% levels, respectively.

. In Column 1, the estimated coefficient of DID is −0.2390, but not significant, suggesting that TPT policy works weakly in the eastern provinces. The estimated coefficient of DID in Column 2 is 0.5710, which is significant at the 5% level, indicating that the implementation of the TPT policy is effective in the central region. Comparing the results, we can conclude that the TPT policy is more applicable to the TFP improvement of pumped storage industry in the central region.

The underlying reasons may be related to the stage of development of the pumped storage plants in different regions. Compared with the central region, the pumped storage industry in the eastern region started earlier, with a more mature and stable operating system, so it is not easily stimulated by external policy shocks. For example, Guangzhou station, Thirteen Tombs station, Tianhuangping station and Taishan station all belongs to the first pumped storage sites in China, built in the 1980s. After more than 30 years of development, the operational efficiency of these sites has reached a high level, and the growth space is relatively limited. Under the new electricity price model, it may lead to a gradual decline in total factor productivity due to technological backwardness and lack of market competitiveness. Our measured data show that from 2014 to 2020, the TFP of these sites does present an overall downward trend. Among them, the Guangzhou station fell by 20%, Thirteen Tombs station fell by 12%. Different from these sites, Huilong station, Xilongchi station, Bailianhe station, Heimufeng station and Hongping station, which are located in the central region, were all built after 2007 or 2010. Most of them are still in their infancy, with a stronger demand for social capital, so the promotion effect of TPT policy is more obvious.

6.2. The Heterogeneity of Energy Dependence

Regional energy dependence may also have an important impact on the performance of TPT policy. In general, the higher the level of (fossil) energy dependence in a region, the more exclusionary it is to the development of renewable energy sources such as pumped storage [51], leading to a weaker stimulus effect of the relevant policies. For this, we conducted a heterogeneity test for energy dependence. Specifically, we use the energy consumption to GDP ratio as a proxy variable of energy dependence and classify all provinces into high energy-dependent group and low energy-dependent group. The regression results are shown in columns 3 and 4. From the results, we can see that the estimation coefficient of DID is not significant in the high energy-dependent group, but it is significantly positive in the low energy-dependent group, indicating that higher energy dependence will weaken the positive effect of TPT policy on TFP, which is consistent with our expectations. Specifically, in these provinces, the implementation of the TPT policy will increase TFP by an average of 0.4930.

The above phenomenon can be attributed to the “resource curse” effect; that is, more abundant natural resources will limit the development of other local industries and weaken the implementation effect of relevant policies [52]. In provinces with high energy dependence, the proportion of fossil energy production and consumption is higher, and the energy factor is more abundant. Taking raw coal as an example, the proportion of raw coal production in high energy-dependent regions is as high as 38.5%, but less than 1% in energy-dependent regions. This constitutes a comparative advantage of factor endowment in these provinces, where most enterprises depend on fossil energy supply for their survival and development. Taking Shanxi as an example, among the 41 listed companies in the province, more than half of them have strong industrial connection with coal or other minerals, and the market value of the energy industry accounts for 28.02%. The exploitation of renewable energy sources such as pumped storage is contrary to their development needs. Therefore, the actual effect of the TPT policy is not obvious.

6.3. The Heterogeneity of Environmental Regulation

The extent of regional environmental regulation is closely related to the development of the energy industry. According to Porter’s hypothesis, environmental regulations can effectively stimulate technological innovation. Applied to the energy sector, this implies that the more stringent the environmental regulations in a region, the stronger the demand for technological innovation in renewable energy, and the easier it is for the innovation incentivizing effect of TPT policy to be realized [53]. To explore this heterogeneity, we use the ratio of industrial pollution control investment to industrial added value as the proxy variable of environmental regulation strength [54], and divide all provinces into high-regulation group and low-regulation group. As shown in columns 5 and 6 of Table 10, in the high-regulation group, the estimated coefficient of DID is 0.1750, which is not significant. However, in the low-regulation group, the estimated coefficient of DID is 0.2740 and has passed the significance test, indicating that environmental regulation has reduced the positive effect of TPT policy on TFP. In the low-regulation area, the TPT policy would raise TFP by 0.2740 on average. This conclusion conflicts with the research of Zhang et al. [53].

As for the reason, we believe that this may be related to the cost of environmental regulation. Excessive environmental costs will squeeze the investment space of enterprises in other production and operation, limit innovation activities and core competitiveness, and thus reduce the overall operational efficiency of enterprises. The pumped storage industry is a typical example. It has high environmental costs, which broadly include the costs of water monitoring, environmental monitoring and water ecological restoration, as well as the costs of biodiversity conservation. In contrast, as a power supply with certain public goods attributes, the pumped storage industry itself is not very profitable. According to the TPT policy requirements, the profit ratio of the capacity tariff is only around 3%. Therefore, in areas with stronger environmental regulation, the pressure on the environmental costs of site enterprises is greater, which weakens the positive role of TPT policy.

6.4. The Heterogeneity of Green Finance

Green finance also plays an important role in promoting the development of renewable energy [55,56]. It can effectively increase corporate environmental protection investment to alleviate the economic pressure of enterprises, so as to avoid enterprises falling into the dilemma of ‘profit or environment’ [57], which is conducive to improving their TFP. Referring to the research results of Lee and Lee [58], we measured the green finance development index of each province and also segmented them into high and low green finance groups. As shown in columns 7 and 8 of Table 10, the estimated coefficients of DID are significantly positive in the provinces with high green finance indices, but not significant in another group, suggesting that green finance is favorable to the positive effect of TPT policy on TPT. In the high green financial area, the TPT policy would raise TFP by 0.5840 on average. This value is the highest of all heterogeneity analyses, implying that the categorization criteria for the level of green finance are the most effective and that regional differences are the most pronounced.

This finding is in line with the expectation that green finance markets have indeed acted as a “catalyst”. In addition to reducing the environmental costs of pumped storage sites, the most critical function of green finance is to facilitate the price marketization. On the one hand, green finance provides a demand-side impetus for electricity price marketization. A higher level of green finance means a higher social demand for clean energy, which creates an upward market push for fluctuating electricity prices in the pumped storage industry. On the other hand, the green finance market will also assess the environmental responsibility and risk of different pumped storage projects, which could guide the flow of capital to sites with higher quality, thus strengthening market competition and promoting price marketization.

7. Conclusions and Suggestions

The purpose of this paper is to investigate the development of China’s pumped storage industry and the role of price policy in it. Accordingly, we first measured the total factor productivity of the pumped storage industry in each province by the EBM-GML method. Then, using DID model and unbalanced panel data of 16 provinces in China from 2004 to 2020, we innovatively tested the impact of ‘Two-Part Tariff’ reform on the total factor productivity of the pumped storage industry and its mechanism. Finally, the influence of some regional own characteristics on the price policy effect was explored as a heterogeneity analysis. Our study found that: (1) Both the construction and operation of China’s pumped storage industry are very unbalanced, and the overall TFP is at a low level. (2) The TPT policy has a significant positive effect on the total factor productivity of the pumped storage industry. This conclusion still holds under multiple robustness tests. Further mechanism analyses show that the TPT policy can increase TFP by promoting an increase in innovation inputs. (3) The impact of the TPT policy on the TFP of the pumped storage industry has obvious regional heterogeneity. The policy effect will be stronger in the provinces which are located in the central region or with low energy-dependence, lax environmental-regulation and developed green-finance, vice versa in other provinces.

The question of how price policy affects firm efficiency, particularly total factor productivity, is an ongoing topic of research. It has been widely explored in many industrial studies [8,10,12,14,16]. However, due to the different types of industries and the different innovation speed of enterprises, the actual effect of price policy is not consistent, which is roughly divided into the following three categories: long-term negative, short-term negative but long-term positive, long-term positive. The conclusion obtained in this paper is only one of them. Throughout these studies, we can find that almost all price policy orientation is marketization. This means that competition intensifies, and the cost of the enterprise will often increase. If enterprises cannot rely on innovation to quickly obtain more benefits, then TFP will decline for a long time, such as Qu et al.’s research on China’s iron and steel industry [10]. If the speed of innovation is too slow, it will lead to a short-term decline in TFP, but long-term growth, such as Zheng et al.’s research on the entire energy-intensive industry [17]. If rapid innovation can be achieved, TFP will also grow rapidly, such as Chen et al.’s research on emerging industries such as wind power industry [14]. For the pumped storage industry studied in this paper, as an emerging strategic industry, it has strong innovation vitality at this stage, which is why we observe the long-term positive effects of the two-part tariff policy. Our study provides more empirical evidence on the relationship between price policy and productivity. In addition to the academic value, some corresponding recommendations are proposed in this paper, as shown in follows:

Firstly, the western region is a new direction for the development of pumped storage industry. This region has the richest wind and light resources across the country, with a large scale and high proportion of renewable energy generation, which urgently needs to be consumed by pumped storage sites. However, our research shows that the current development of the pumped storage industry in the western region is clearly lagging behind, with no completed sites yet. Therefore, the construction of pumped storage plants in this region needs to be accelerated. Secondly, the TPT policy has proved to be favorable to increasing total factor productivity. Consequently, the government should continue to promote the price marketization. However, at present, the operation of pumped storage power stations is still under the strict control of grid companies, which makes it difficult for them to participate in competition as independent market players, and the process of marketization is seriously hindered. Therefore, in the future, the pumped storage industry should also consider the implementation of the “separation of electricity and grid” to strengthen the independent status of pumped storage sites. Thirdly, local governments should adopt relevant complementary policies or initiatives in accordance with local conditions, so as to give full play to the positive effects of the TPT policy. The eastern provinces should play the leading role of social capital and actively encourage social capital to participate in the construction and development of pumped storage industry, while the central provinces need to be supported by more financial subsidies from local governments. Energy-dependent provinces should accelerate the structural transformation of industries in order to drive the transformation of energy structure, whereas non-energy-dependent provinces should vigorously promote the joint development of water, wind and solar integrated renewable energy. Moreover, governments should actively promote the development of green financial markets and rationalize the level of environmental regulation throughout the country.