The Impact of Market Integration on Renewable Energy Technology Innovation: Evidence from China

Abstract

China is vigorously building a unified domestic market, with priority given to regional market integration while maintaining a national unified market. Limited research has been conducted on whether market integration affects renewable energy technology innovation (RETI). This paper empirically studies this topic based on panel data of 30 Chinese provinces from 2004 to 2020 using FMOLS (Fully Modified Ordinary Least Squares), DOLS (Dynamic Least Squares method), and FGLS (Feasible Generalized Least Squares). Research results have been verified by robustness tests. The main conclusions are as follows: (1) Market integration has an important positive impact on RETI, namely, boosting it. This conclusion remains robust when different indicators of innovation and market segmentation are included. (2) The regional impacts of market integration on RETI are heterogeneous, being the greatest in east China, followed by west China and central China. (3) Market integration affects RETI through energy structure and technological innovation. It can optimize energy structure, improve technological innovation, and thus enhance RETI. Based on the above conclusions, in order to improve RETI in China and expand its contribution to carbon neutrality and carbon peaking, China needs to strengthen the construction of a national unified market and implement differentiated market integration policies in east, central, and west China. Furthermore, it is necessary to give full play to the role of energy structure and technological innovation in market integration by optimizing energy structure and improving technological innovation.

1. Introduction

In view of global climate change and energy security, many countries are vigorously developing renewable energy and reducing the proportion of non-fossil energy consumed. China is a responsible big country committed to peaking its carbon emissions around 2030 and achieving 20% non-fossil energy as a proportion of primary energy supply by 2030.-In 2021, the cumulative installed capacity of renewable energy in China was 1.06 TW, which took up 44.8% of installed power capacity, a year-on-year rise of 14.4%, and accounted for 29.7% of total power generation, up 12.1%. Total utilization of renewable energy amounted to 750 million tons of standard coal, which accounted for 14.2% of total primary energy consumed. The installed capacity of solar energy rose by 54.93 GW, of which 54.88 GW was contributed by photovoltaic generation. The cumulative installed capacity of photovoltaic generation reached 306.56 GW, a year-on-year growth of 21.8%. Power generation amounted to 327 billion kWh, up 25.5%, and accounting for 3.9% of total power generation. The cumulative capacity of distributed photovoltaic generation exceeded 100 GW, which constituted one-third of total photovoltaic power generation. China is implementing RETI to address high cost and difficulties in large-scale renewable energy consumption. RETI helps reduce the cost and increase the share of renewable energy, thereby facilitating the transformation from fossil fuels to renewable energy [1]. This is an important path for China to achieve carbon peaking and carbon neutrality.

Some researchers have studied renewable energy consumption and nonrenewable energy, Pao and Fu (2013) employed Brazil’s yearly statistics from 1980 to 2010 to explore the causal relationships between the real GDP and four types of energy consumption: non-hydroelectric renewable energy consumption (NHREC), total renewable energy consumption (TREC), non-renewable energy consumption (NREC), and the total primary energy consumption (TEC). They found that expanding renewable energy would not only enhance Brazil’s economic growth and curb the deterioration of the environment, but also create an opportunity for a leadership role in the international system and improve Brazil’s competition with more developed countries [2]. Tang et al. (2016) found that the tourism-led-growth hypothesis is valid, but the energy-led-growth hypothesis is invalid in India. Therefore, policymakers should promote and expand tourism industry in order to sustain the process of economic growth and development in India [3]. Nathaniel et al. (2019) explored the relationship between ecological footprint, urbanization, and energy consumption by applying the ARDL estimation technique on data spanning 1965–2014 for South Africa. While urbanization and energy use promote environmental quality in the long run, financial development and economic growth degrade it further [4]. Adami et al. (2017) presents the case of Rio Grande do Sul (RS), in Brazil, a state which implemented a regional industrial policy (IP/RS) in the wind energy sector. The IP/RS promoted the introduction of public goods and horizontal market interventions for the wind sector [5]. Strielkowski et al. (2021) assessed energy efficiency in Russia on its path towards the modernization of its energy systems. They presented the perspectives of developing the renewable energy technologies in Russia [6]. Wang and Li (2016) aimed to calculate the P (population)–A (affluence)–T (technology) effects of energy use in China and India—the world’s two most populous and largest developing countries. Findings suggest that market-oriented economic and energy reforms need to send the correct price signal to promote energy-efficient technologies, thus improving energy efficiency, which is the key to a sustainable energy future in China and India [7]. Lin and Presley (2014) investigated claims concerning a Granger causality relationship from energy consumption to economic growth in South Africa. This suggests that energy use would have a long-run effect of raising the country’s CO₂ emission levels. Hence, there might be a need to develop a more balanced energy structure, which will include a higher share of renewable energy [8].

China is also promoting market integration. It issued the Opinions of the CPC Central Committee and the State Council on Accelerating the Construction of a National Unified Market on 10 April 2022, which proposes to break down local protectionism and market segmentation, achieve breakthroughs in key points that restrain business cycles, promote the smooth flow of commodity factors in a wider range, and speed up the construction of an efficient, regulated, and fully open national unified market featuring fair competition. Market integration refers to the free flow of commodities and factors among regions based on market signals. When the integration among regions is strengthened, the cost of commodity transportation and trade will decrease, and the price gap among regions will narrow [9]. Market integration and market segmentation are opposite concepts with inverse proportion [10,11]. The greater the degree of market segmentation, the lower the degree of market integration [12,13,14].

The former is considered an important factor affecting RETI in China [15,16]. As different regions have varied renewable energy endowments (for example, west China has rich renewable energy resources, while east China possesses more advanced renewable energy resources), market integration can optimize resource allocation across the country and foster improvement of RETI in west China.

The existing literature on market integration places emphasis on its impact on carbon emissions. For example, Li and Lin (2017) evaluated the CO₂ performance of 28 Chinese provinces from 1995 to 2012 and investigated the impact of market integration on CO₂ performance [17]. They found that market integration can effectively improve CO₂ emission performance. He et al. (2018) analyzed data about 30 Chinese provinces from 2002 to 2011 and found market integration exerts a significant positive impact on reducing marginal CO₂ costs. The effect of RETI, a new type of technological innovation, is affected by market factors [18]. However, most studies ignore the impact of market integration on RETI. Based on this, the research purpose of this paper is to take China as an example to analyze the impact of market integration on RETI, which is helpful in complementing the existing literature and enriching the theory of RETI.

This paper makes the following contributions. First, it provides a new perspective for studying RETI by calculating the market integration index and RETI of 30 provinces. Second, it offers new insights into factors influencing RETI by analyzing the impact of market integration on RETI. Third, it takes into consideration regional heterogeneity and provides key evidence for policymakers by analyzing the impact of market integration in different regions on RETI. Fourth, it looks into the mechanisms of the impact of market integration on RETI in terms of energy structure and technological innovation.

The rest of this paper is organized as follows. Part 2 presents a literature review. Part 3 introduces the research model, variables, and related data. Part 4 is an overview of empirical results. Part 5 examines regional heterogeneity and the mechanisms of impact. Part 6 elaborates on conclusions and policy recommendations.

2. Literature Review

Previous research on RETI focuses mostly on its impact on CO₂ emissions and factors influencing it. Lin and Zhu (2019) analyzed data about Chinese provinces from 2000 to 2015 through panel data models and revealed CO₂ emissions and innovation have a bidirectional relationship, namely, CO₂ emissions boost RETI [19]. RETI helps reduce CO₂ emissions, while innovation in fossil energy technologies is ineffective at reducing CO₂ emissions [20]. He et al. (2021) discussed the impact and mechanism of impact of RETI on the total-factor CO₂ emission performance index (TCPI). They found that RETI can effectively improve TCPI, but this is affected by market factors [21]. Su and Fan (2022) revealed that RETI exerts a significant positive impact on green development [22]. Some studies uphold that RETI significantly alleviates household energy poverty [23].

In terms of influencing factors of RETI, some studies suggest that political pressure also affects RETI [24]. Raza et al. (2020) made it clear that only electricity production and investment in renewable energy has a significant impact on RETI [25]. Li and Shao (2021) examined the determinants of renewable energy innovation by applying a negative binomial model. They found that renewable electricity installed capacity, share of expenditure on research and development (R&D) of GDP, and implementation of the Kyoto Protocol promoted RETI [26]. In a few words, there is a lack of research on the impact of market integration on RETI.

Market integration has an impact on technological innovation [27]. Technological innovation is driven by market incentives, and strengthened market integration can significantly raise innovation efficiency and accelerate the commercialization of innovation results [28]. Market integration increases the efficiency of factor allocation, fosters the establishment of consistent environmental standards, and enhances technological innovation [29].

In contrast, market segmentation counts against technological innovation. Sun et al. (2020) analyzed the impact of market segmentation on technological innovation with panel data about 29 Chinese provinces from 2004 to 2015, and concluded that market segmentation hinders technological innovation [30]. Market segmentation severely restricts the free flow and efficient aggregation of production factors, which not only leads to information asymmetry, but also hinders enterprises’ technological innovation due to unreasonable regional allocation of resources [31,32]. It also curbs technological innovation through distorted factor prices, resource allocation, technological diffusion, and local protectionism [33,34].

In terms of research methods, some existing research adopted the threshold regression method and others adopted the FMOLS method. There are also studies that use spatial econometric methods for analysis. Moreover, Cross-sectional dependence testing was not performed in most relevant studies. Ignoring cross-sectional dependencies before performing panel model estimation may result in inconsistent and invalid estimation results. Therefore, the cross-sectional dependence test was used in this paper, and three methods of FMOLS, DOLS, and FGLS were used for regression to make the estimation results more robust.

To sum up, the existing literature focuses on factors influencing renewable energy or the impact of market integration on technological innovation. There is a lack of literature that works on the impact of market integration on RETI. On the basis of measuring market integration and RETI, this paper explores the impact of market integration on RETI, China’s market integration, and completion of carbon neutrality and carbon peaking.

3. Model Specification and Data Description

3.1. Model Specification

The following economic model (1) is built to analyze the impact of market integration on RETI:

$$\ln RET{I}_{it}=c_0+{\alpha }_1\ln M{I}_{it}+{\displaystyle \sum {\gamma }_j}{x}_{ijt}+{\epsilon }_{it}$$

(1)

where i indicates the province, t the year, RETI the explained variable, MI the core explanatory variable (market integration), x_ijt control variables (per capita energy-industry investment, trade openness, industrial structure, and real per capita GDP), and ${\epsilon }_{it}$ the random error term, which obeys a normal distribution.

A mediation model is established on the basis of the above benchmark model to examine the mechanism of impact of market integration on RETI [35,36]. Mediation means the transmission of the effect of an explanatory variable (X) on an explained variable (Y) through one or more other variables through a third-party variable called a mediator [37]. Mediation models are often used in fields such as environment studies [38]. Mediation effects are usually tested with the causal-step approach [35,36].

$$Y=cX+e_i$$

(2)

$$M=aX+e_i$$

(3)

$$Y=c^{\prime }X+bM+e_i$$

(4)

The first step is to test whether coefficient c in Equation (2) is significant, the second step tests whether coefficient a in Equation (3) is significant, and the third step tests whether coefficients $c^{\prime }$ and b in Equation (4) are significant. If c, a, and b are all significant, but $c^{\prime }$ is not significant or significant but with a decreased value, M is a mediating variable.

To test the mechanism of impact of market integration on RETI, the following Equations (5)–(7) are constructed.

$$\ln RET{I}_{it}={\alpha }_0+{\alpha }_1\ln MI+{\displaystyle \sum {\gamma }_j}{x}_{ijt}+{\epsilon }_{it}$$

(5)

$$M_{it}={\beta }_0+{\beta }_1\ln MI+{\displaystyle \sum {\gamma }_j}{x}_{ijt}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(6)

$$\ln RET{I}_{it}={\lambda }_0+{\lambda }_1\ln MI+{\lambda }_2\ln M_{it}+{\displaystyle \sum {\gamma }_j}{x}_{ijt}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(7)

M in formula (6) refers to mediating variable energy structure and technological innovation, whose meanings are interpreted as below.

(1) Energy structure

Market integration enhances information exchanges, raises the efficiency of information search and circulation, and alleviates information asymmetry between energy supply side and demand side. Measures can be taken to allocate energy based on the needs in different regions to optimize energy structure. Regions with large natural resource endowments are more likely to form an energy consumption structure dominated by the resources they own, thereby inducing more related R & D and innovation activities [39]. Therefore, market integration may affect RETI through energy structure, namely, the share of coal in total energy consumed.

(2) Technological innovation

Market integration facilitates the flow of technological factors and technological spillovers [40]. The research and application of renewable energy technologies across regions promote studies on renewable energy [41]. The mechanism by which market integration affects RETI through TI is studied in this paper. R & D intensity is used herein to refer to technological innovation.

3.2. Dependent Variable

The dependent variable in this study is RETI, but official statistics in China do not include it. Most of the literature uses the knowledge stock of renewable energy patents (RPATS) as the representation of RETI [42,43].

Therefore, this approach is also used in this paper. RETI is calculated via perpetual inventory, as shown in Equation (8).

$$RPAT{S}_t=RPAT{C}_t+(1-\delta )\cdot RPAT{S}_{t-1}$$

(8)

RPATC_t is the number of new RPATS at time t, the source of which is the Patent Search and Analysis System (PSA) of the China National Intellectual Property Administration (CNIPA). RPATS cover wind energy, solar energy, marine energy, biomass energy, and energy storage. The classification codes of granted renewable energy patents are obtained from the International Patent Classification (IPC) Green Inventory on the website of World Intellectual Property Organization (WIPO). The patents are categorized by the date and province of publication. RPATS_t–1 is the knowledge stock of renewable energy patents in the previous period, which should be depreciated when performing calculations. Referring to the existing literature [44,45,46], a depreciation rate (δ) of 10% is adopted. China’s patents began to be recorded in 1985, so RPATC₀ is the number of RPATS in 1985 [47].

3.3. Independent Variable

The key independent variable in this study is market integration, which is usually measured by production, price, trade, business cycle, and questionnaire survey. Among them, price is the most widely used. Drawing on Parsley and Wei (2001) and Zhang et al. (2020), this paper employs price index to calculate relative price variance based on three-dimensional (t × i × k) panel data, where t is the year, i the province, and k the commodity [48,49]. The prices of 16 categories of commodities are included, including: food; beverages, tobacco, and alcohol; clothing, shoes, and hats; textiles; household appliances and audio–visual equipment; cultural and office supplies; daily necessities; sports and entertainment supplies; transportation and communication supplies; furniture; cosmetics; gold, silver, and jewelry; Chinese and Western medicines and health care products; books, newspapers, magazines, and electronic publications; fuels; building materials and hardware. The commodity retail price index used in this paper is chain data with the previous period as the base period, so relative price is the logarithmic first-order difference of relative price ratio.

The formula is:

$$\Delta Q_{i,j,t}^k=\ln (p_{i,t}^k/p_{j,t}^k)-\ln (p_{i,t-1}^k/p_{j,t-1}^k)=\ln (p_{i,t}^k/p_{i,t-1}^k)-\ln (p_{j,t}^k/p_{j,t-1}^k)$$

(9)

In Equation (9), $\Delta Q_{i,j,t}^k$ is the first-order difference of relative price, $p_{i,t}^k$ and $p_{i,t-1}^k$ the retail price indexes of commodities of type k in period t and t − 1 in province i, and $p_{j,t}^k$ and $p_{j,t-1}^k$ the retail price indexes of commodities of type k in period t and t − 1 in province j. This paper chooses the absolute value of the first-order difference of relative price $\left|\Delta Q_{i,j,t}^k\right|$ to eliminate the influence of differential code on the variance of relative price. With the above calculations, we obtain the absolute values of first-order differences of 118,320 relative prices $\left|\Delta Q_{i,j,t}^k\right|$ from 2004 to 2020 in 435 pairs of provincial combinations of 16 types of commodities.

To eliminate price changes unrelated to differences in market environment between regions, this study deals with relative prices by de-averaging in reference to Bian et al. (2019). Suppose $\left|\Delta Q_{i,j,t}^k\right|={\alpha }^k+{\epsilon }_{i,j,t}^k$, where ${\alpha }^k$ indicates price change caused by the nature of commodities of type k, while ${\epsilon }_{i,j,t}^k$ refers to price change caused by different market environments in regions i and j. To eliminate fixed effect ${\alpha }^k$, $\overline{\left|\Delta Q_t^k\right|}$ is obtained by averaging the relative prices of 435 pairs of provincial combinations of commodities of type k in a year t $\left|\Delta Q_{i,j,t}^k\right|$, the difference between which, $q_{i,j,t}^k$, is the basis for calculating relative price variance. The above calculation process is expressed as Equation (10):

$$\left|\Delta Q_{i,j,t}^k\right|-\overline{\left|\Delta Q_t^k\right|}=({\alpha }^k-\overline{{\alpha }^k)}+({\epsilon }_{i,j,t}^k-\overline{{\epsilon }_{i,j,t}^k})={\epsilon }_{i,j,t}^k-\overline{{\epsilon }_{i,j,t}^k}=q_{i,j,t}^k$$

(10)

The market integration index of the 30 provinces is obtained by combining the relative price variances of 435 pairs of combinations and then taking the reciprocal of their squares. A total of 510 observations are gained.

The control variables are per capita energy-industry investment [50], trade openness [51], industrial structure [52], and real per capita GDP [53]. Per capita energy-industry investment is calculated by dividing energy-industry investment by the population of each province. Trade openness is the proportion of total import and export in regional GDP. Industrial structure is the share of secondary industry in GDP. Real GDP per capita is deflated by per capita GDP in 2004.

4. Results and Discussion

4.1. Data

Samples in this study are data about 30 Chinese provinces from 2004 to 2020. Except for RETI and market integration, other raw data come from statistical yearbooks of related provinces from 2005 to 2021. Missing data have been filled by interpolation. Descriptive statistics for all variables are presented in

Table 1. The descriptive statistics of variables.

Variable	Unit	Obs.	Mean	Std. Dev.	Min	Max
RPATS	-	510	161.178	222.701	1.023	1283.583
MI	-	510	61.821	13.807	23.088	96.985
TO	%	510	30.751	37.839	0.764	184.289
PEII	Ten thousand/person	510	0.221	0.217	0.022	1.296
SE	%	510	45.139	8.621	15.800	61.500
GDP	Yuan/person	510	17,692.67	11,364.25	4215.00	67,003.97

.

4.2. Correlation Test and Multicollinearity Test

The correlation and multicollinearity between variables are examined with the Pearson test and variance inflation factor (VIF), the results of which are shown in

Table 2. Matrices of correlations and variance inflation factor statistics.

Variables	lnRETI	lnMI	lnPEII	lnTO	lnSE	lnGDP
lnRETI	1.000
lnMI	0.285 ***	1.000
lnPEII	0.048	0.139 ***	1.000
lnTO	0.425 ***	−0.106 **	−0.332 ***	1.000
lnSE	−0.072 *	−0.060	0.099 **	−0.155 ***	1.000
lnGDP	0.570 ***	−0.036	0.043	0.775 ***	−0.201 ***	1.000
VIF	-	1.03	1.56	3.82	1.07	3.48
1/VIF	-	0.969	0.643	0.262	0.935	0.287
Mean VIF	2.19

Notes: *** < 0.01, ** < 0.05, * < 0.1.

. Correlation tests indicate there is a high correlation between real GDP per capita and trade openness. The correlation coefficients between other variables are all below 0.58. VIF tests reveal there is a low collinearity between variables. The mean of VIF values is 2.19, being lower than benchmark value 5. Therefore, there is no apparent multicollinearity between variables.

4.3. Cross-Sectional Dependence Test

Ignoring cross-sectional dependence before performing panel model estimation may lead to inconsistent and thus invalid estimation results [54]. Therefore, cross-sectional dependence (CD) test is conducted before performing baseline model estimation to verify cross-sectional dependence. The results are presented in

Table 3. Results of cross-sectional dependence test.

Variables	Pesaran (2004) CD-Test [41]	p-Value	Corr	Abs (Corr)
lnRETI	74.01	0.000	0.861	0.861
lnMI	60.04	0.000	0.698	0.698
lnTO	21.19	0.000	0.246	0.564
lnPEII	58.43	0.000	0.679	0.720
lnSE	61.58	0.000	0.716	0.716
lnGDP	56.56	0.000	0.658	0.700

Notes: Under the null hypothesis of cross-sectional independence CD~N (0,1).

. All variables are significant at the 1% level, so the null hypothesis is rejected. That is, each sequence has a significant cross-sectional dependence. Further testing of panel data stability is required.

4.4. Stationary Test

To prevent spurious regression, a unit root test is necessary to verify data stability. LLC (Levin–Lin–Chu), IPS (Im, Pesaran, and Shin), ADF (Augmented Dickey–Fuller), and Fisher Chi-square tests are carried out to verify the stationarity of all variables, the results of which are displayed in

Table 4. Unit roots tests.

Variables	Levels
	LLC		IPS		ADF–Fisher
	Intercept	Intercept and Trend	Intercept	Intercept and Trend	Intercept	Intercept and Trend
lnRETI	−11.611 ***	−3.471 ***	−5.642 ***	1.433	138.529 ***	44.933
lnMI	−1.264	−4.798 ***	−0.537	−2.119 **	54.522	76.004 *
lnPEII	−4.910 ***	−4.685 ***	−1.301 *	−2.0310 **	77.602 *	80.981 **
lnTO	−1.196	−2.965 ***	1.420	−0.887	46.049	67.912
lnSE	5.427	−1.551 *	7.844	3.239	12.098	35.191
lnGDP	−6.088 ***	−0.065	−1.163	5.297	69.045	18.783
Variables	First-order difference
	LLC		IPS		ADF−Fisher
	Intercept	Intercept and Trend	Intercept	Intercept and Trend	Intercept	Intercept and Trend
lnRETI	−5.394 ***	−5.469 ***	−7.048 ***	−8.238 ***	157.885 ***	181.610 ***
lnMI	−9.966 ***	−7.234 ***	−8.964 ***	−5.389 ***	191.129 ***	128.356 ***
lnPEII	−8.888 ***	−7.794 ***	−6.816 ***	−4.460 ***	149.568 ***	114.670 ***
lnTO	−10.590 ***	−10.005 ***	−8.999 ***	−6.049 ***	191.152 ***	137.917 ***
lnSE	−2.342 ***	−2.303 **	−3.090 ***	−4.172 ***	92.960 ***	110.604 ***
lnGDP	−4.729 ***	−7.392 ***	−3.603 ***	−4.791 ***	99.392 ***	119.338 ***

Notes: *** < 0.01, ** < 0.05, *<0.1.

. Some variables are horizontally non-stationary, so the null hypothesis of the panel unit root cannot be rejected, and a first-order difference is needed. The results say that the first-order differences of all variables are stationary, so the null hypothesis is rejected, and the alternative hypothesis is accepted for the cointegration test.

After confirming that the first-order differences of variables are stationary, the authors studied whether there is long-term cointegration between variables, with the method used by Pedroni (1999) and Kao (2019) [55,56]. The results are exhibited in

Table 5. Pedroni’s (2004) residual cointegration test results.

Statistics	Within-Dimension		Statistics	Between-Dimension
	Value	p-Value		Value	p-Value
Panel v-stat	0.621	0.267
Panel rho-stat	2.892	0.998	Group rho-stat	4.643	1.000
Panel PP-stat	−11.839 ***	0.000	Group PP-stat	−19.915 ***	0.000
Panel ADF-stat	−9.189 ***	0.000	Group ADF-stat	−13.530 ***	0.000

Notes: The null hypothesis is that the variables are not cointegrated. Lag selection is based on SIC. *** < 0.01.

and

Table 6. Kao’s (1999) residual cointegration test results.

	t-Stat	p-Value
	−6.405 ***	0.0000
Residual variance	0.194
HAC variance	0.157

Notes: The null hypothesis is that the variables are not cointegrated. Lag selection is based on SIC. *** < 0.01.

. Pedroni (1999) residual cointegration tests reveal that most of the statistics are statistically significant. Therefore, the null hypothesis of no cointegration relationship is rejected, as there is a long-term stable cointegration between variables in the model. Kao’s test results further show that panel variables in the model are cointegrated at the 1% significance level. Regression analysis is then carried out based on above analysis.

4.5. Benchmark Regression

The impact of market integration on RETI is estimated through regression, as shown in Equation (1). Columns (1) to (9) of

Table 7. The effect of lnMI on RETI (Dependent variable: lnRETI).

Variables	FMOLS			DOLS			FGLS
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
lnMI	1.706 *** (0.224)	0.542 *** (0.171)	0.836 *** (0.191)	1.255 *** (0.267)	0.556 ** (0.261)	0.618 ** (0.278)	1.047 *** (0.010)	1.348 *** (0.024)	0.769 *** (0.147)
lnPEII		1.202 *** (0.087)	0.170 ** (0.072)		0.972 *** (0.107)	0.186 * (0.108)		0.700 *** (0.050)	0.435 *** (0.051)
lnTO			0.600 *** (0.082)			0.302 ** (0.130)			0.349 *** (0.067)
lnSE			−0.408 ** (0.191)			−0.719 ** (0.299)			−1.322 *** (0.131)
lnGDP			0.113 (0.112)			0.409 ** (0.173)			0.615 *** (0.084)
Constant Term	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	510	510	510	510	510	510	510	510	510
R-squared	0.704	0.825	0.3064	0.7143	0.9331	0.7123

Note: Standard errors in parentheses; *** < 0.01, ** < 0.05, * < 0.1.

list regression results based on FMOLS, DOLS, and FGLS. Control variable per capita energy-industry investment is introduced in columns (2), (5), and (8), and all control variables are included in columns (3), (6), and (9). All of the coefficients of market integration in regressions (1) to (6) are positive and significant at 5% or less. This means promoting market integration helps improve RETI in China, which is consistent with the research result of Ren et al. (2021). Market integration prevents local governments from restricting the flow of resources and products between regions in order to protect the interests of local enterprises. This enables companies to form economies of scale and have greater motive for technological innovation. Market integration also facilitates the dissemination of innovative knowledge and low-carbon technologies within a region, and promotes the application of renewable energy technologies [57]. This conclusion based on empirical research provides another path for enhancing RETI, namely, the government can strengthen RETI by breaking market segmentation and promoting regional market integration. This is consistent with the hypothesis in the paper.

4.6. Robustness Analysis

Key variables are transformed when conducting robustness tests, the results of which are shown in

Table 8. Robustness tests.

Variables	(1)	(2)	(3)	(4)
	lnRPATC	lnRPATC	lnRETI	lnRETI
lnMI	1.019 *** (0.011)	0.680 *** (0.150)
lnMS			−0.520 *** (0.005)	−0.215 *** (0.058)
lnPEII		0.435 *** (0.053)		0.703 *** (0.039)
lnTO		0.360 *** (0.068)		0.100 * (0.056)
lnSE		−1.243 *** (0.133)		−1.356 *** (0.099)
lnGDP		0.609 *** (0.086)		0.910 *** (0.069)
Constant	Yes	Yes	Yes	Yes
Control	NO	YES	NO	YES
Observations	509	509	510	510
Method	FGLS	FGLS	FGLS	FGLS

Note: Standard errors in parentheses; *** < 0.01, * < 0.1.

. In this paper, two methods are used to test the robustness. Firstly, the renewable energy patents (lnRPATS) variable, instead of RETI (lnRETI), is adopted to check robustness. Columns (1) and (2) show that the coefficient of market integration is significantly positive at the 1% level. In other words, market integration can significantly improve renewable energy patents.

Second, market segmentation, instead of market integration, is adopted to further test robustness. Columns (3) and (4) indicate the coefficient of market segmentation is significantly negative at the 1% level. That is to say, market segmentation can significantly reduce RETI. This means the main estimation results in this study are robust.

5. Further Discussion

5.1. Regional Analysis

Based on the above benchmark results, the authors explore the impact of market integration on RETI in different regions. Compared with DOLS and FMOLS, FGLS can solve cross-sectional dependence and heteroscedasticity problems [58]. Therefore, FGLS estimation is used in

Table 9. Results of spatial heterogeneity.

Variables	Eastern			Middle			Western
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
lnMI	0.794 *** (0.263)	1.307 *** (0.062)	1.386 *** (0.286)	0.994 *** (0.020)	1.250 *** (0.058)	0.814 *** (0.257)	0.924 *** (0.017)	1.074 *** (0.025)	0.877 *** (0.280)
lnPEII		0.140 (0.124)	0.119 (0.136)		0.508 *** (0.108)	1.461 *** (0.102)		0.455 *** (0.066)	0.355 *** (0.088)
lnTO			−0.832 *** (0.158)			−0.905 *** (0.159)			0.425 *** (0.114)
lnSE			1.163 *** (0.242)			1.194 *** (0.397)			−2.688 *** (0.481)
lnGDP			3.381 *** (3.237)			0.186 (0.323)			1.065 *** (0.270)
Constant	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Control	NO	YES	YES	NO	YES	YES	NO	YES	YES
Observations	187	187	187	136	136	136	187	187	187
Method	FGLS	FGLS	FGLS	FGLS	FGLS	Method	FGLS	FGLS	FGLS

Note: Standard errors in parentheses; *** < 0.01.

. In columns (1) to (9), the coefficients of market integration are all positive and significant at the 1% level, that is, market integration promotes RETI in all regions. This further verifies the conclusion that market integration helps advance RETI. Specifically, when all control variables are included, the coefficients of columns (3), (6), and (9) are 1.386, 0.814, and 0.877. That is, market integration has the greatest impact on RETI in east China, followed by west China and central China. This may be because, among the three regions, east China had the most developed economy, the highest level of technology, and the most mature soft power and hardware for market integration. West China had the lowest level of economic development and insufficient infrastructure. By developing a national unified market, the renewable energy potential of west China can be tapped into and its RETI will be strengthened.

5.2. Mechanism Analysis

Table 10. Results of the mediating-effect estimation.

	(1)	(2)	(3)	(4)
Variables	lnES	lnRETI	lnTI	lnRETI
lnMI	−0.092 *** (0.019)	0.455 *** (0.116)	0.102 *** (0.034)	0.724 *** (0.137)
lnES		−0.172 * (0.089)
lnTI				1.394 *** (0.071)
lnPEII	−0.001 (0.011)	0.711 *** (0.039)	0.077 *** (0.013)	0.349 *** (0.046)
lnTO	0.028 ** (0.013)	0.1127 ** (0.055)	0.185 *** (0.016)	0.082 (0.057)
lnSE	0.470 *** (0.042)	−1.090 *** (0.1575)	−0.914 *** (0.029)	−0.027 (0.144)
lnGDP	−0.319 *** (0.044)	0.866 *** (0.070)	0.311 *** (0.020)	0.160 * (0.082)
Constant	YES	YES	YES	YES
Control	YES	YES	YES	YES
Observations	510	510	510	510
Method	FGLS	FGLS	FGLS	FGLS

Note: Standard errors in parentheses; *** < 0.01, ** < 0.05, * < 0.1.

shows the mediating effect of market integration on RETI. According to columns (1) and (2), market integration affects RETI by affecting energy structure. In column (1), the coefficient of market integration is significantly negative, that is, market integration can reduce the proportion of coal consumed and improve energy structure. In column (2), energy structure is included in model (7). The coefficient of market integration is still significantly positive, but the value decreases from 0.769 to 0.455. This indicates that energy structure has a partial mediating effect. In column (3), the coefficient of market integration is significantly positive, that is, market integration can increase energy efficiency. In column (4), technological innovation is taken into account in model (7). The coefficient of market integration is still significantly positive but the value decreases, which means technological innovation also has a partial mediating effect.

6. Conclusions and Implications

China is vigorously building a unified domestic market, with priority given to regional market integration while maintaining a national unified market. Limited research has been conducted on whether market integration affects renewable energy technology innovation (RETI). This paper makes the following contributions. First, it provides a new perspective for studying RETI by calculating the market integration index and RETI of 30 provinces. Second, it offers new insights into factors influencing RETI by analyzing the impact of market integration on RETI. Third, it takes into consideration regional heterogeneity and provides key evidence for policymakers by analyzing the impact of market integration in different regions on RETI. Fourth, it looks into the mechanisms of the impact of market integration on RETI in terms of energy structure and technological innovation. This paper empirically studies this topic based on panel data of 30 Chinese provinces from 2004 to 2020 using FMOLS, DOLS, and FGLS. Research results have been verified by robustness tests. The main conclusions are as follows: (1) Market integration has an important positive impact on RETI, namely, boosting it. This conclusion remains robust when different indicators of innovation and market segmentation are included. (2) The regional impacts of market integration on RETI are heterogeneous, being the greatest in east China, followed by west China and central China. (3) Market integration affects RETI through energy structure and technological innovation. It can optimize energy structure, improve technological innovation, and thus enhance RETI.

This paper proposes the following policy recommendations in order to improve RETI in China and expand its contribution to carbon neutrality and carbon peaking.

(1) Strengthen the construction of a national unified market. This study emphasizes the importance of market integration to RETI. Promoting market integration should be regarded as a policy goal for enhancing RETI. This can not only boost economic development through the free flow of factors and optimal allocation of resources, but also strengthen RETI to a certain extent. Therefore, it is necessary to strengthen the construction of a national unified market. On the one hand, the central government should supervise and punish some local governments for market segmentation. On the other hand, as China’s national unified market is in the early stage of development, the central government can link the degree of local market integration with the performance of local governments in assessment to effectively promote market integration.

(2) Implement differentiated market integration policies in east, central, and west China. Heterogeneity tests herein reveal that market integration has the greatest impact on RETI in east China, followed by west China and central China. Therefore, the government should advance market integration in line with the characteristics of each region instead of imposing uniformity in all regions. East China has leading technical strength, which is a great advantage in market integration, so it should take the lead in breaking protective barriers between provinces, accelerate market integration, and play a demonstration role by giving full play to its technological advantages. Vast and sparsely populated west China has advantages in resource endowment. Its reserves of renewable energy such as wind energy and solar energy are national-leading. To promote market integration in this region, the government should strengthen support in this regard, improve its infrastructure, and create conditions for a national unified market.

(3) Give full play to the role of energy structure and technological innovation in market integration by optimizing energy structure and improving technological innovation. It is necessary to rationalize coal prices so that they can reflect the scarcity of coal and reduce the amount of coal consumption, implement measures such as coal clean technology to improve coal efficiency, and strengthen investment in the energy industry to enhance technology, thus improving technologies for renewable energy such as solar energy, wind energy, geothermal energy, and water energy.