Can Innovation-Driven Policy Reduce China’s Carbon Emission Intensity?—A Quasi-Natural Experiment Based on the National Innovative City Pilot Policy

Abstract

Innovation and carbon neutrality are hot topics nowadays and are important issues related to development quality, efficiency, and long-term sustainability. The purpose of this paper was to analyze the impact of innovation-driven policy on reducing carbon emission intensity. Using urban panel data from 2003 to 2018, this paper constructed a time-varying difference-in-difference model based on the natural experiment of national innovative city pilot policy and systematically examined the specific effects and mechanisms of innovation-driven policy on reducing carbon emission intensity. It was found that the innovation-driven policy represented by the national innovative city pilot policy can significantly reduce the carbon emission intensity in China, and this result was still valid after various robustness tests. The mechanism analysis showed that a reduction in carbon emission intensity was mainly achieved by innovation-driven promotion of green production technology progress and improvements in energy use efficiency. Further heterogeneity analysis showed that the policy had a more significant carbon reduction effect in cities with a strong innovation capability, large size, and low level of industrial structure. These research conclusions provide useful references for further understanding of the economic and social benefits of innovation-driven policy and effectively unleashing the carbon emission reduction effects of innovation-driven policy and promoting sustainable development.

1. Introduction

China’s economy in transition faces the trade-off and coordination between environmental protection and economic growth. Promoting the implementation of innovation-driven policy and strengthening carbon emission management are strategic initiatives for China to promote high-quality development under the new stage of development. Since the reform and opening up, China has achieved a miracle of economic growth, but its high growth rate mainly stems from leveraging its low-cost advantages in labor and environmental resources. As China enters a new stage of development, its low-cost advantage gradually disappears, but technological innovation can provide unique competitive advantages such as allowing companies to produce more high value-added products in a cleaner way. At the same time, the extensive growth of the past few decades has brought increasing environmental problems, one of which is rising greenhouse gas emissions. Greenhouse gases have high transmittance to visible light from solar radiation and are highly absorbent of long-wave radiation reflected from the Earth, resulting in the “greenhouse effect” that leads to global warming, destroys the ecological balance, and threatens the food supply and living environment of human beings. China’s total carbon emissions climbed from 1.419 billion tons to 9.429 billion tons from 1978 to 2018, an increase of 5.64 times [1]. As the world’s largest carbon emitter, it is particularly important for China to reduce its carbon emission and achieve the goal of carbon peak and carbon neutrality at an early date.

An important question that is closely related to the aforementioned reality and policy context but has not been fully verified is: what is the relation between innovation-driven policy and carbon emission in China? In other words, do innovation-driven policy have a significant impact on China’s carbon emission? If so, is the effect positive or negative? What are the mechanisms that generate such effects? Systematically exploring these problems and clarifying the casual relationships are of great significant for deeply implementing innovation-driven policy and reducing the carbon emission intensity, which are the problems that this paper aimed to focus on. Specifically, based on detailed and rich urban panel data, this paper took the national innovative city pilot policy as a quasi-natural experiment and used a normative casual identification strategy to analyze these problems.

The economic and social effects of the national innovative city pilot policy have been well verified in the existing literature. On the one hand, the construction of national innovative cities can promote green technology innovation through the guidance of government strategy, the support of innovative talents, and the optimization of industrial structure [2,3,4], while green technology innovation has a significant inhibitory effect on carbon emission [5]. On the other hand, science and technology innovation driven by the national innovative city pilot policy can promote the overall optimization and upgrading of industrial structure by enhancing industry productivity [6,7]; at the same time, there can be a positive promotion effect between industrial structure upgrading and carbon emission reduction [8,9,10,11,12]. In addition, some scholars believe that science and technology innovation can effectively promote the replacement of traditional energy with clean energy, thus reducing the cost and the quantity of carbon emission [13,14]. In addition, studies on the economic and social effects generated by innovation-driven policy have also concentrated on their impact on reducing environmental pollution: Gao and Yuan (2021) investigated the pollution-control effects of innovation-driven policy using industrial sulfur dioxide, industrial soot, and industrial wastewater as indicators [15]; Meng et al. (2021) started from the four latitudes of human capital, physical capital, urban function, and government affairs [16] and found that the land use transformation promoted by innovation-driven policy during rapid urbanization was conducive to reducing industrial pollution in the region.

When reviewing the above literature, it was not difficult to find that there were some existing studies on the economic and social effects of innovation-driven policy, but the following important aspects need to be expanded. Firstly, different studies can support the positive effect of innovation-driven policy on reducing carbon emission intensity in terms of the coherence of theoretical logic, but only a few studies have directly examined the casual relationships between the two, and the question of whether the national innovative city pilot policy can reduce carbon emission intensity has not been well answered. Secondly, among the few existing studies that examined the effect of innovation-driven policy on reducing carbon emission, the mechanisms of the impact of innovation-driven policy have not been effectively identified.

Based on the above analysis, it can be seen that innovation-driven policy may have great potential in reducing carbon emission intensity, but it has not been fully verified, and the mechanism of action also needs to be explored. In view of this, this paper attempted to further investigate whether innovation-driven policy reduced carbon emission intensity based on the existing literature. The possible marginal contributions are mainly as follows: (1) In contrast to the existing literature that focused on the pollution reduction effect of innovation pilot policy, this paper treated the national innovative city pilot policy as a quasi-natural experiment and applied the “Difference-in-difference” model to identify these effects and mechanisms and creatively assesses the impact of the policy on carbon emission intensity, which enriches the research on the effect of innovation-driven policy to a certain extent, is of great significance to the carbon emission reduction work in China and the whole world, and provides policy references for global carbon emission reduction; (2) while existing research mainly focused on the impact of innovation-driven policy on entrepreneurship, innovation, economic development and other aspects, this paper theoretically and empirically analyzed the impact of innovation-driven policy on reducing carbon emission intensity and expanded the beneficial impact of innovation-driven policy on ecological environmental protection, which is an important supplement to the existing literature; (3) this paper investigated the transmission mechanism of innovation-driven policy to reduce carbon emission intensity in detail and identified the mechanism pathway to reduce carbon emission intensity by promoting the level of green technology innovation and improving energy efficiency, which enriches the mechanism of innovation-driven policy to reduce carbon emission; and (4) compared with the existing literature, this paper further distinguished the asymmetric impact of innovation-driven policy on carbon emission intensity from the perspectives of innovation capacity, city size, and industrial structure, which is conducive to further understanding the heterogeneous carbon emission reduction effects of innovation-driven policy in different urban characteristics.

2. Policy Background and Mechanism Analysis

2.1. Policy Background

As an important development strategy in China, the innovation-driven policy is not a policy design limited to a single area such as education and enterprise technology innovation. The policy scope covers all areas in a city that can help to form an innovation-driven development model, maintain innovation advantages for a long time, maintain strong competitiveness and promote innovation development. It is a comprehensive and multi-field policy system based on stimulating innovation vitality, promoting innovation advancement, and transforming the power of development. To deal with the insufficient science and technology innovation capacity since the reform and opening up, in 2005 the China State Council promulgated the “National Medium and Long-Term Science and Technology Development Plan (2006–2020)” (https://www.most.gov.cn/xxgk/xinxifenlei/fdzdgknr/fgzc/gfxwj/gfxwj2016/201612/t20161213_129574.html, accessed on 2 December 2022), which proposed to make building an innovative country a major strategic choice in China’s future development plan. The national innovative city pilot policy is an important progressive reform policy to implement the innovation-driven development strategy in China, and Shenzhen became the first approved national innovative pilot city in 2008 when the practical experiment of “piloting first, accumulating experience, and gradually rolling out” began. In 2010, the number of approved national innovative pilot cities reached the peak of the entire reform process, with 16 cities (districts) (including Dalian) approved by the China Development and Reform Commission, 20 cities (districts) (including Haidian District of Beijing) approved by China’s Ministry of Science and Technology in the first batch, and 18 cities (districts) (including Shijiazhuang of Hebei Province) approved in the second batch as national innovative pilot cities (districts). In 2016, China’s Ministry of Science and Technology and the China Development and Reform Commission jointly issued the “Guidelines on the Construction of Innovative City” (https://www.most.gov.cn/xxgk/xinxifenlei/fdzdgknr/qtwj/qtwj2018/201804/t20180404_139016.html, accessed on 2 December 2022), which further revised the relevant guidelines and index system, integrated the national innovative pilot cities approved in the early stage, and announced the specific list of 61 national innovative pilot cities. In 2017, all 61 innovative pilot cities passed the examination and accepted. In 2018, China’s Ministry of Science and Technology and the China Development and Reform Commission jointly issued the “Letter on Supporting the New Round of Cities to Construct Innovative Cities” (https://www.most.gov.cn/xxgk/xinxifenlei/fdzdgknr/qtwj/qtwj2018/201804/t20180404_139016.html, accessed on 2 December 2022), which approved a new round of pilot projects in 17 cities (including Xuzhou city in Jiangsu Province). To date, the national innovative city pilot policy has covered 78 cities in 31 provinces.

It is clearly required in various documents of the policy that the construction of a national innovative city should be guided by the realization of innovation-driven development, take the improvement of independent innovation ability as the main idea, take the innovation of system and mechanism as the driving force, improve the innovation system, gather innovation resources, highlight benefits and efficiency, and use “green and low carbon” as the guiding ideology and construction principles of urban construction by relying on scientific and technological innovation to solve the problem of green development, accelerate the construction of a resource-saving and environment-friendly society, and promote harmonious development between humans and nature. This can improve the regional innovation system, increase investments in education and science and technology, enhance the effective supply capacity of regional innovative talents and technology, promote the upgrading of urban industries, and optimize the regional industrial structure. A series of innovative index systems have also been constructed; these include highlighting the forest coverage rate and carbon dioxide emission intensity per unit of GDP for green and ecological innovation cities. Since the implementation of the policy, each pilot city has formulated detailed development plans according to the guidelines and requirements of the policy combined with the local reality and explored the formation of innovative development models and development patterns with their own characteristics. Therefore, this paper took the “National Innovative City Pilot Policy” as the entry point and used the “green and low carbon” feature of this policy to study whether it has the effect of reducing carbon emission intensity and the action mechanism of reducing carbon emission intensity.

2.2. Mechanism Analysis

Carbon emission is the emission of greenhouse gases, and the burning of fossil fuels is the main source of carbon emission. In today’s era, human production and life cannot be separated from fossil fuels, carbon emissions are generated every moment. The influencing factors of carbon emission are diverse, and factors such as economy, population, industrial structure and energy intensity will all affect carbon emission [17,18,19]. As an important institutional innovation achievement, the national innovative city pilot policy—a policy implemented in multiple fields that promotes innovation and development at all levels of the city—will have an impact on the factors affecting carbon emission, which will indirectly lead to changes in the carbon emission intensity. Therefore, while considering the complexity of the factors affecting carbon emission and the importance of the government policy, this paper indirectly analyzed the action mechanism of the innovation-driven policy on reducing carbon emission intensity from the perspective of the factors that affect carbon emission. Finally, considering that different urban characteristics have different impacts on carbon emission, this paper investigated in depth the heterogeneity of the impact of innovation-driven policy on carbon emission intensity from three perspectives: innovation capacity, city size, and industrial structure.

Firstly, the implementation of innovation-driven policy will have a direct impact on carbon emission. The national innovation-driven city pilot policy issued by China’s Ministry of Science and Technology and the China Development and Reform Commission calls for green and low-carbon construction principles to create an excellent living environment and urban ecology and strives to build the city into a beautiful home where humans and nature live in harmony. Municipal governments at the local level used the guidelines of the policy combined with their own local characteristics to formulate detailed implementation plans for the pilot project. For example, Jinan is to become an innovative regional center city for low-carbon economic development, and Nanjing is required to promote the application of green and low-carbon technology innovation. At present, the impact of environmental regulation on carbon emission is mainly manifested as the “pushback effect” [20], which refers to the government constraining the production behavior of enterprises (especially those with high pollution and energy consumption) through mandatory orders, setting energy conservation and emission reduction targets, and forcing enterprises to conduct green technology innovation or change their management mode to reduce carbon emission [21]. Therefore, this paper proposes the following hypothesis:

Hypothesis 1.

Innovation-driven policy can directly reduce carbon emission intensity.

Secondly, innovation-driven policy may reduce carbon emission intensity through green technology innovation and energy-efficiency improvement. Green technology innovation is a technological innovation that contributes to improvements in environmental quality [22]. In the long term, technological progress is an effective means to solve environmental pollution problems. In particular, green technology-oriented innovation [23] and green technology innovation can significantly reduce carbon emission [24,25]. In the national innovative city pilot policy document, it is required to pay close attention to the incentive of innovative talents; implement major talent projects; innovate the mode of talent training, use, and introduction; improve the flow and service guarantee mode of innovative talents; increase the investments in education and science and technology, build and introduce high-level education and research institutions; and enhance the effective supply capacity of innovative talents and technology. This provides high-quality human capital for green technology innovation and lays a good foundation for innovation factors. At the same time, the policy also requires the establishment and improvement of a modern industrial technology innovation system combining industry–university–research, which can make the cooperation and connections among enterprises and universities and research institutes closer, increase the collaborative innovation ability of enterprises, and thus promote an improvement in the green technology innovation level. With regard to improving energy efficiency, the current industrial production activities in China rely heavily on the excessive input of energy and other resources, which reflects a low energy efficiency. Energy efficiency is an important indicator of a country’s energy utilization capacity, and its emission reduction effect has been confirmed by extensive research [26,27,28,29,30]. Improving energy efficiency has become a key way to reduce emission while maintaining economic growth [31,32]. Energy conservation generates emission reduction benefits in two main ways: first, energy efficiency improvements not only directly reduce energy consumption in the industrial production process, but also promote energy conservation throughout the product life cycle [29,33], which significantly reduces energy consumption and thus reducing carbon emission; second, energy efficiency improvements imply greater economic benefits per unit of energy use. The increased benefits will promote more investment in technology research, which will promote energy conservation and emission reduction projects, forming a system of mutual promotion and a virtuous cycle [34,35]. The national innovative city pilot policy mentions that it is necessary to strengthen the planning, guidance, and overall layout; innovate management models and operation mechanisms; and promote the gathering of innovation themes, the aggregation of innovation resources, the clustering of innovation services, and the fusion of emerging industries. This policy can guide and promote industrial agglomeration and the optimization and upgrading of industrial structure. Some existing studies argue that industrial agglomeration can improve energy efficiency through the positive externality of agglomeration [36,37,38], the improvement of energy utilization efficiency is influenced by the industrial structure [39], and the quality improvement of industrial structure adjustment has a significant contribution to the improvement in energy efficiency [40]. Therefore, this paper proposes the following hypothesis:

Hypothesis 2.

Innovation-driven policy can reduce carbon emission intensity through green technology innovation and energy efficiency improvement.

3. Research Design

3.1. Model Design: Time-Varying DID Benchmark Regression Model

China proposed to build an innovative country as a national strategic choice in 2005 and began a gradual attempt to implement the national innovative city pilot policy in 2008. Several innovative cities were established in different batches in 2008, 2010, and 2011, which was an exogenous policy shock to energy conservation and emission reduction. In this paper, the national innovative city pilot policy is considered as a quasi-natural experiment, and the following regression model was constructed by using a “Difference-in-difference” model (DID) to evaluate the impact of the policy on reducing the carbon emission intensity. The following regression model was constructed:

$$c{i}_{it}=\alpha +\beta Polic{y}_{it}+\gamma Contro{l}_{it}+C_i+T_t+{\epsilon }_{it}$$

where i denotes the city, t denotes the year, $c{i}_{it}$ denotes the carbon emission intensity, $Polic{y}_{it}$ denotes the national innovative city pilot policy, $Contro{l}_{it}$ denotes a set of control variables, $C_i$ is the city fixed effect, $T_t$ is year fixed effects, and ${\epsilon }_{it}$ is the random disturbance term. The coefficient $\beta$ reflects the net effect of innovative city construction on carbon emission performance. If $\beta$ is <0 and significant, it means that the innovative city construction can significantly reduce the carbon emission intensity. If $\beta$ is not significant, it indicates the impact of innovative city construction on carbon emission is not obvious.

3.2. Variable Selection

Carbon emission intensity ($c{i}_{it}$) is the explained variable in this paper; the ratio of CO₂ emissions to the GDP of each city in each year was selected to measure the carbon emission intensity. The national innovative city pilot policy ($Polic{y}_{it}$) is the core explanatory variable of this paper; it indicated whether city i was established as an innovative pilot city in year t. The interaction term of the city dummy variable and the policy implementation time dummy variable (City × Time) was used to measure the policy benefits of the national innovative city pilot policy ($Polic{y}_{it}$): The national innovative pilot city variable City (the experimental group) was set to 1, and the non-national innovative pilot city (the control group) was set to 0. The time variable before the implementation of national innovative pilot city policy was set to 0, and the time variable after the implementation of the policy was set to 1.

Considering that the development of different cities has great heterogeneity, many factors will also have an impact on urban carbon emission intensity, so the following control variables for $Contro{l}_{it}$ were set. (1) Industrial structure (Ind): the differences in industrial structure among cities can reflect a series of differences in different urban functions, resource endowment, science and technology levels, labor quality and quantity, and consumption habit preference, etc. These differences will have an impact on urban carbon emission intensity. In this paper, the ratio of industrial added value to GDP was used to measure the change of industrial structure in each city. (2) Green coverage rate of built-up area (Green): greening is undoubtedly an important way to reduce carbon emission. The photosynthesis of plants and their input to the soil carbon pool during the growth process have become an important carbon sink that can offset the carbon emission to some extent. (3) Population density (Po): an empirical analysis by Chai (2013) found that the per capita carbon emission first decreased and then increased during the process of population increase [41]. In this paper, the ratio of total urban population to area was used to measure the population density. (4) Environmental regulation (ER): this is an important policy tool for ecological governance nowadays. The government reduces the use of fossil fuels and controls carbon emission intensity by taxing high-emission enterprises and subsidizing enterprises using clean energy. This paper measured environmental regulation using the proportion of words related to environmental protection to the total number of words in the work reports of each local government. (5) Foreign direct investment intensity (Fdi): from a national perspective, China’s carbon emission intensity decreases significantly after the introduction of foreign investment, but the effect varies across regions and there are threshold conditions such as the per capita income level, environmental regulation intensity, energy consumption structure, etc., which will result in a significant decrease in carbon emissions when some conditions are met. This paper measured the intensity of foreign direct investment intensity using the ratio of the actual amount of foreign investment used in the year to the regional GDP. The actual amount of foreign investment in the current year was obtained by multiplying the actual amount of foreign investment used in the current year (in USD millions) by the average exchange rate between the RMB and the USD in the current year.

To exclude the effect of heteroskedasticity on the regression results, the above variables were logarithmically treated in the regression.

3.3. Data Sources

According to the implementation time of the national innovative city pilot policy and considering the completeness and availability of relevant data, this paper took the prefecture-level cities as samples, excluded prefecture-level cities newly established due to the adjustment of the administrative division (such as Bijie City and Tongren City in Guizhou Province), and some western cities with serious data deficiency. Finally, the balanced panel data of 266 prefecture-level cities in 16 years from 2003 to 2018 were selected, and the missing data in the panel were combined with local statistical yearbooks or statistical bulletins or were supplemented by linear interpolation and the moving average method. The data were mainly obtained from the China City Statistical Yearbook, provincial statistical yearbooks, urban statistical bulletin, EPS database, and CSMAR database (

Table 1. Descriptive statistics of main variables.

Variable	Variable Meaning	Obs	Mean	SD	Min	Max
$c{i}_{it}$	Logarithm of carbon emission intensity	4256	−13.12854	0.594217	−15.79712	−10.90308
$Polic{y}_{it}$	National innovative city pilot policy	4256	0.0850564	0.2789984	0	1
LnInd	Logarithm of industry structure	4155	−4.506574	1.317381	−13.25711	−0.978727
LnGreen	Logarithm of greening coverage of built-up areas	4256	3.564658	0.3885407	−0.52763	5.957494
LnPo	Logarithm of population density	4256	−2.760653	1.668129	−14.5289	3.558215
LnER	Logarithm of environmental regulation intensity	4256	−1.378301	1.155832	−5.771594	1.795783
LnFdi	Logarithm of foreign direct investment intensity	4256	−10.15975	0.41433	−14.78662	−7.762087

).

4. Empirical Results and Analysis

4.1. Impact of Innovation-Driven Policy on Carbon Emission Intensity

Table 2. Overall impact of national innovative city pilot policy on carbon emission intensity.

	(1)	(2)	(3)	(4)	(5)
	$\mathit{c}{\mathit{i}}_{\mathit{i}\mathit{t}}$	$\mathit{c}{\mathit{i}}_{\mathit{i}\mathit{t}}$	$\mathit{c}{\mathit{i}}_{\mathit{i}\mathit{t}}$	$\mathit{c}{\mathit{i}}_{\mathit{i}\mathit{t}}$	$\mathit{c}{\mathit{i}}_{\mathit{i}\mathit{t}}$
Policyit	−0.0591 ***	−0.0650 ***	−0.0613 ***	−0.0620 ***	−0.083 ***
	(−5.37)	(−6.31)	(−5.96)	(−6.05)	(−5.67)
LnFdi		−0.0108 ***	−0.0103 ***	−0.0123 ***	−0.0119 ***
		(−4.40)	(−4.24)	(−5.01)	(−4.86)
LnGreen		−0.0358 ***	−0.0351 ***	−0.0386 ***	−0.0379 ***
		(−4.47)	(−4.39)	(−4.82)	(−4.76)
LnER		0.0306 ***	0.0294 ***	0.0320 ***	0.0307 ***
		(13.79)	(13.20)	(14.43)	(13.83)
LnPo		−0.2301 ***	−0.2260 ***	−0.2309 ***	−0.2265 ***
		(−8.86)	(−8.74)	(−8.94)	(−8.81)
LnInd		−0.0354 ***	−0.0375 ***	−0.0314 ***	−0.0333 ***
		(−3.29)	(−3.49)	(−2.93)	(−3.11)
Smart City			−0.0410 ***		−0.0432 ***
			(−4.92)		(−5.22)
Low Carbon Cities				−0.0506 ***	−0.0518 ***
				(−6.56)	(−6.75)
_Cons	−12.62227 ***	−13.16513 ***	−13.18269 ***	−13.1186 ***	−13.13436 ***
	(−1527.14)	(−107.95)	(−108.40)	(−107.96)	(−108.43)
Fixed effects	Control	Control	Control	Control	Control
Observations	4256	4155	4154	4155	4154
R2	0.8674	0.8910	0.8919	0.8922	0.8931

Note: t-statistic is given in parentheses; *** was significant at the levels of 1%.

reports the regression results of the impact of the national innovative pilot policy on carbon emission intensity. Column (1) shows the estimation results without considering the control variables when fixed effects were controlled, and column (2) shows the estimation result while considering both fixed effects and control variables. In the regression results, it can be seen that the estimated coefficient of the core explanatory variable $Polic{y}_{it}$ was significantly negative, which initially indicated that the national innovative pilot cities can significantly reduce the local carbon emission intensity.

During the implementation period of the national innovative city pilot policy, the smart city construction policy implemented in China will promote low-carbon development [42]; in addition, the low-carbon city pilot policy will also lead to a reduction in carbon emissions. Therefore, the reduction effect of the national innovative city pilot policy on carbon emission intensity may be overestimated due to the impact of other policies beyond consideration. To further enhance the accuracy and robustness of the empirical results, we successively added the policy dummy variables of the implementation of the smart city policy and the low-carbon city pilot policy into the benchmark regression mode; the results are shown in Table 2 (columns (3)–(5)). Based on the estimation results, it can be seen that after controlling for the two policies, the coefficient of the national innovative city pilot policy remained significantly negative, which indicated to a certain extent that the carbon emission reduction effect of the innovative city pilot policy really exists.

4.2. Parallel Trend Test and Dynamic Effects Analysis

In this paper, the time-varying DID model was used to conduct regression and analysis on each variable, which required the experimental group and the control group to maintain a consistent trend before the policy shock; that is, to meet the parallel trend test. This paper drew on the practice of Ren et al. (2019) and used an event analysis to conduct the test by setting a time dummy variable of 1 for the year of policy implementation and 0 for the remaining years [43] then cross-multiplying it with the grouped dummy variable of the policy shock. These interaction terms were put into the regression equation for estimation by using the second year before policy shock as the base period. If the interaction term before the policy shock in the regression results was statistically significant, it indicated that there was a large difference in the trend of change between the experimental group and the control group before the policy shock. Otherwise, it indicated that there was no significant difference in the trend of change between the experimental group and the control group before the policy shock; that is, the national innovative city pilot policy conformed to the hypothesis of a parallel trend. Figure 1 shows a plot of the estimated coefficients of the interaction terms and their 95% confidence intervals.

According to the results shown in Figure 1, it was clear that the national innovative city pilot policy effectively reduced the carbon emission intensity in general but with some lag. The reduction degree of the innovative city pilot policy on carbon emission intensity was not obvious in the early stage of the implementation of the policy, and the influence coefficients of the pilot policy were significantly negative and all smaller than the current period after 3 years of the pilot policy implementation. The overall trend is accelerated reduction, which may be due to the fact that the facility construction and investment return of the pilot policy require a certain period.

4.3. Sensitivity Analysis

The main source of CO₂ is the combustion of fossil fuels, which is often accompanied by the production of SO₂. Therefore, in this paper, the ratio of CO₂ emissions to GDP in each city in each year in the benchmark regression model was replaced with the ratio of SO₂ emissions to GDP in each city in each year for the robustness test, and the ratio of SO₂ emissions to GDP was logarithmically treated in the regression in order to exclude the effect of heteroskedasticity on the regression results. The test results are shown in columns (1) and (2) in

Table 3. Robustness test results.

	Sensitivity Analysis		PSM-DID
	(1)	(2)	(3)	(4)
	LnSO2	LnSO2	$\mathit{c}{\mathit{i}}_{\mathit{i}\mathit{t}}$	$\mathit{c}{\mathit{i}}_{\mathit{i}\mathit{t}}$
Policyit	−0.1854 ***	−0.1907 ***	−0.0497 ***	−0.0629 ***
	(−3.98)	(−4.76)	(−4.80)	(−6.19)
Control variables	No	Yes	No	Yes
Fixed effect	Control	Control	Control	Control
Observations	4256	4155	3833	3833
R2	0.4925	0.6444	0.8848	0.8949

Note: t-statistic is given in parentheses; *** was significant at the levels of 1%. Due to limited space, we omitted the estimated coefficients of the constants and control variables.

. Column (1) reports the regression results for the case without added control variables, and column (2) reports the regression results for the case with added control variables. As can be seen in the regression results, the estimated coefficients of the core explanatory variables were significantly negative in both cases, demonstrating that the national innovative city pilot policy significantly reduced the emission intensity of SO₂. In a sense, it showed that the causal identification results of this paper were relatively robust.

4.4. Propensity Score Matching–Difference-in-Difference (PSM-DID)

Using the time-varying difference-in-difference model required the experimental group and control group to satisfy the randomness assumption for the selection and to avoid systematic differences in the trends of changes between national innovative pilot cities and ordinary cities. To ensure the robustness of the regression results, this paper drew on the practice of Fu et al. (2018) and further used the propensity score matching–difference-in-difference (PSM-DID) method for robustness testing to reduce the selectivity error of the sample [44]. The specific practice followed was: the control variables in the benchmark regression were used as covariates, and the kernel matching method was applied to match the samples year by year to obtain matching results that were as similar as possible before the experimental group and control group received policy shocks with the loss of a small portion of the sample. Then, using the matched city samples as the basis, the benchmark regression model was used to verify the impact of the national innovative city pilot policy on carbon emission intensity reduction. The regression results are shown in columns (3) and (4) in Table 3; column (3) reports the case without added control variables, and column (4) reports the case with added control variables. As can be seen in the results, the coefficient of Policy_it was still significantly negative, which was not significantly different from the above difference-in-difference regression results, thereby indicating that the empirical conclusion of this paper was robust: the national innovative city pilot policy has the effect of reducing carbon emissions.

4.5. Placebo Test

In the above section, a quasi-natural experiment was used to control the characteristics of a large number of sample cities, and possible adverse effects of other policies were excluded. The parallel trend test, which was the assumed premise of the time-varying difference-in-difference model was satisfied; however, the problem of variable omission could not be completely avoided. To further ensure the robustness of the regression results, this paper conducted a placebo test while referring to the practice of Cai et al. (2016), Chen et al. (2018), and Li et al. (2016) [45,46,47] using Stata software to conduct 500 random shocks on 266 sample cities while selecting some cities as the experimental group and others as the control group. Under this experimental condition, if the coefficient of Policy_it remained significant, it indicated that the empirical results above were caused by other factors; if the coefficient was no longer significant, it indicated that the empirical results above were reliable; that is, the national innovative city pilot policy can reduce carbon emission intensity. This paper added the control variables and controlled fixed effects when conducting the placebo test; the results are shown in Figure 2. The figure shows the t-value density distribution of Policy_it after 500 random sampling times. It can be seen that the t-statistic of Policy_it was mainly distributed around the point 0. The t-value of Policy_it was −6.31 in the regression of adding the control variables and controlling the fixed effects in the above paper, so it can be concluded that the empirical results of this paper truly reflected the innovative city pilot policy’s reduction effect on carbon emission intensity by comparing the t-value of Policy_it with the t-value distribution of the placebo test results.

5. Further Analysis: Mechanism Test and Heterogeneity Discussion

5.1. Mechanism Test

Based on the previous analysis, it can be seen that the national innovative city pilot policy had a significant effect on reducing carbon emission intensity, but what was the mechanism of this policy to reduce carbon emission intensity? In order to verify the action mechanism of the national innovative city pilot policy on reducing carbon emission, the mediation model constructed in this paper was as follows:

$$Inte{r}_{it}=\alpha +\beta Polic{y}_{it}+\gamma Contro{l}_{it}+C_i+T_t+{\epsilon }_{it}$$

$$c{i}_{it}=\delta +\epsilon Polic{y}_{it}+ϵContro{l}_{it}+\theta Inte{r}_{it}+C_i+T_t+{\epsilon }_{it}$$

where $Inte{r}_{it}$ is the mediator variable, which is replaced by green technology innovation (GTI) and energy efficiency (EE) in turn; and the other variables remained the same as given above. If the coefficient $\beta$ and the coefficient $\theta$ in the regression results were both significant, then the intermediary effect was established.

(1) Green Technology Innovation (GTI): this paper used the number of green patents of the sample cities in each year from the China City Statistical Yearbook to measure green technological innovation. In order to exclude the influence of heteroskedasticity on the regression results, the number of green patents was logarithmically treated in the regression. The regression results are shown in columns (1) and (2) in

Table 4. Results of the mechanism test.

	(1)	(2)	(3)	(4)
	GTI	$\mathit{c}{\mathit{i}}_{\mathit{i}\mathit{t}}$	$\mathit{E}\mathit{E}$	$\mathit{c}{\mathit{i}}_{\mathit{i}\mathit{t}}$
$Polic{y}_{it}$	0.4860 ***	−0.0481 ***	−0.0714 ***	−0.0581 ***
	(9.46)	(−4.60)	(−3.10)	(−5.77)
GTI		−0.0121 ***
		(−3.33)
EE				0.0959 ***
				(13.68)
Control variables	Control	Control	Control	Control
Fixed effect	Yes	Yes	Yes	Yes
Proportion of mediator effect	10.89%		10.54%
Observations	3353	3353	4155	4155
R2	0.7526	0.8884	0.3128	0.8960

Note: t-statistic is given in parentheses; *** was significant at the levels of 1%. Due to limited space, we omitted the estimated coefficients of the constants and control variables.

. The results in column (1) show that the estimated coefficient of the core explanatory variable $Polic{y}_{it}$ was significantly positive, indicating that the national innovative city pilot policy significantly improved the level of green technology innovation. Column (2) replaces the explained variable with carbon emission intensity ($c{i}_{it}$), and the core explanatory variable $Polic{y}_{it}$ and the mediator variable green technological innovation (GTI) were put into the equation for regression at the same time. The results showed that both the core explanatory variable of $Polic{y}_{it}$ and the mediator variable of green technological innovation (GTI) were significantly negative. This indicated that innovation-driven policy can reduce carbon emission intensity by promoting the level of green technology innovation.

(2) Energy efficiency (EE): drawing on the extensive practice of the existing literature, this paper used electricity consumption per unit of GDP to measure energy efficiency [48,49,50]. The logical basis was that electricity consumption is the most extensive direct energy consumption in current production and life. In China, the vast majority of fossil energy such as coal, oil, and natural gas is also used for electric power generation. This paper calculated the ratio of electricity consumption to GDP of the sample cities in each year from the China City Statistical Yearbook. A larger numerical value represented a higher unit energy consumption and a lower energy efficiency. In order to exclude the influence of heteroskedasticity on the regression results, the ratio of electricity consumption to GDP was logarithmically treated in the regression. In addition, in order to exclude extreme values, the data were winsorized by 5%. The regression results are shown in columns (3) and (4) in Table 4. The results in column (3) show that the estimated coefficient of the core explanatory variable $Polic{y}_{it}$ was significantly negative, which indicated that the national innovative city pilot policy significantly reduced the unit energy consumption and improves energy efficiency. In column (4), the explained variable was replaced by carbon emission intensity ($c{i}_{it}$), and the core explanatory variable of $Polic{y}_{it}$ and the mediator variable of energy efficiency (EE) were put into the equation at the same time. The results showed that the core explanatory variable was significantly negative and the intermediary variable energy efficiency (EE) was significantly positive. This indicated that innovation-driven policy can reduce carbon emission intensity by improving energy efficiency.

5.2. Heterogeneity

The main purpose of the heterogeneity test in this paper was to clarify whether the effect of the national innovative city pilot policy on reducing carbon emission intensity would differ significantly due to the different innovation capacities of different cities, whether the differences in city size would cause the policy to have a differential effect on reducing carbon emissions, and whether the differences in the levels of industrial structure of different cities would lead to different effects of the policy on reducing carbon emissions.

5.2.1. Heterogeneity Analysis While Considering the Characteristic of Urban Innovation Capability

The construction of national innovative pilot cities requires the realization of innovation-driven development as the guidance, and innovation capability drives the iterative upgrading of production technologies, which can lead to higher production efficiency and lower carbon emission intensity with the same output, thus technological innovation plays an important role in reducing carbon emissions. In this paper, urban innovation index was calculated and averaged over it [51], and then selected the median to divide the total sample into two groups of “strong innovation cities” and “weak innovation cities”; columns (1) and (2) in

Table 5. Heterogeneity analysis results.

	Innovation Ability		City Size		Industry Structure
	(1)	(2)	(3)	(4)	(5)	(6)
	Strong	Weak	Large	Small	High Level	Low Level
Policy	−0.0611 ***	−0.0482	−0.1196 ***	0.0420 *	−0.0293 **	−0.1963 ***
	(−5.93)	(−0.79)	(−12.04)	(1.92)	(−2.43)	(−8.48)
Control variables	Control	Control	Control	Control	Control	Control
Fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Observations	2124	2031	2126	2029	2084	2071
R2	0.9076	0.8780	0.9281	0.8726	0.8916	0.9004

Note: t-statistic is given in parentheses; ***, **, and * are significant at the levels of 1%, 5%, and 10%, respectively. For limited space, we omit the estimated coefficients of constants and control variables.

report the respective regression results. Based on the results, it can be seen that the reduction effect of carbon emissions by the national innovative city pilot policy was only significant in the group of “strong innovation cities”, which indicated that the policy had a better reduction effect on carbon emissions in the cities with a better innovation foundation. This enabled the cities with a better innovation foundation to make the “snowball” of carbon emission reduction increasingly larger due to the opportunity of the national innovative city pilot policy.

5.2.2. Heterogeneity Analysis While Considering the Characteristic of City Size

The implementation effect of the national innovative city pilot policy depends on the city size to a certain extent. The agglomeration effect generated by the expansion of the city size can bring about efficiency gains in many aspects, but it is also accompanied by an increase in carbon emissions due to the expansion of the population and the number of vehicles, so this paper analyzed the heterogeneity of the city size. The China State Council classifies city sizes according to population. In this paper, the total population of each sample city in each year was averaged, and the median was selected to divide the total sample into two groups: “large cities” and “small cities”; columns (3) and (4) in Table 5 report the respective estimation results. It can be seen in the regression results that the reduction effect of the national innovative city pilot policy on carbon emission intensity was greatly related to the city size: the policy only had a significant reduction effect on carbon emission intensity in “large cities” and even exacerbated the carbon emission intensity in “small cities” to a certain extent. The reason for this may be that larger cities with larger population bases had a larger radius of people to implement the innovative policy, which made the implementation of the policy more effective. This indicated that the innovative policy should focus on larger cities to bring about a greater reduction in carbon emission intensity.

5.2.3. Heterogeneity Analysis While Considering the Characteristic of Industrial Structure

The industrial structure varies from city to city, and different industries have great differences in the use of different energy, resulting in different carbon emission intensities. This paper took the average of the industrial structure upgrade index of each city in the sample in different years and then selected the median to divide the total sample into “high level of industrial structure” and “low level of industrial structure”. The upgrade index of industrial structure adopted the calculation method of Fu (2010) [52]; the larger the index, the higher the level of industrial structure of the city. The regression results are reported in columns (5) and (6) of Table 5, respectively. Based on the regression results, it can be seen that the national innovative city pilot policy had a better effect on reducing carbon emission intensity for cities with a low level of industrial structure and a more general effect on reducing carbon emission intensity for cities with a high level of industrial structure. The reason may be that cities with a low level of industrial structure had a high concentration of industry and agriculture; their development tended to be extensive, so there was more room for progress to reduce carbon emissions. Therefore, the national innovative city pilot policy had a “rain on the drought” effect on the reduction of carbon emission intensity in cities with a low level of industrial structure, and it was easier for the policy to be effective. This also indicated that for cities with a low level of industrial structure, the implementation of the national innovative city pilot policy is a good strategy to reduce carbon emission intensity.

6. Conclusions and Policy Implications

Global warming is a huge challenge for human society, and the increasing carbon emissions have introduced a series of serious problems to human production and life as well as economic and social development. How to promote energy conservation and carbon reduction in the context of carbon peaking and carbon-neutral targets in China in the new development period has become a major issue that needs to be solved urgently. This paper regarded the implementation of the national innovative city pilot policy in different cities at different times as a quasi-natural experiment and used a time-varying DID model, a PSM-DID model, and a normative casual identification strategy to systematically examine the causal relationship and internal mechanism of innovation-driven policy on reducing carbon emission intensity. The main research conclusions were as follows. The national innovative city pilot policy significantly reduced carbon emission intensity during the investigation period, which was still valid after a series of robustness tests such as a parallel trend test and a placebo test. The heterogeneity analysis found that the differences in the reduction effect of the national innovative city pilot policy on carbon emission intensity were related to the innovation capacity, city size, and industrial structure of cities. In the group of cities with a higher innovation capacity and a larger size, the reduction effect of the policy on carbon emissions was more obvious; in the grouping by industrial structure, the results showed that the policy had a more significant reduction effect on carbon emission intensity for cities with a low level of industrial structure. The results also enriched the previous research results on the heterogeneity of innovation-driven policy [53,54]. At the same time, in contrast to the previous literature on the mechanism of innovation-driven policy [55,56], this paper distinguished the contribution of promoting green technology innovation and improving energy efficiency to reducing carbon emission intensity in the mechanism analysis and calculated the contribution degree of the two mechanisms. The mechanism test showed that innovation-driven policy could reduce the carbon emission intensity—mainly because the policy could promote the level of green technology innovation and improve energy efficiency.

The conclusions of this paper have an important guiding significance for supporting the national innovative city pilot policy and reducing carbon emission intensity. The policy implications of the conclusions of this paper are as follows. Firstly, the new round of scientific and technological revolution and industrial transformation are reconstructing the global innovation landscape and reshaping the global economic structure. In order to achieve the goal of entering the group of innovative countries, China has placed science and technology innovation at the core of the overall national development and implemented the innovation-driven policy. This paper found that the implementation of innovation-driven policy had a significant effect on reducing carbon emission intensity, which means that the policy can not only increase the economic welfare by improving the efficiency of economic growth and providing new basis point for economic growth (as suggested by the existing literature) but also increase the social welfare by reducing carbon emission intensity. This provides new theoretical support and empirical evidence to promote high-quality development through the implementation of innovation-driven policy.

Secondly, China’s economy has entered a new stage of development after achieving “miracle growth”, the economic sectors are no longer “interest-only”, and the economic growth mode has been transformed from extensive growth to intensive growth. At the same time, with the growing material and cultural needs, people have put forward higher and higher requirements for their ecological living environment. How to maintain economic growth while providing effective protection for the ecological living environment and realize a “win-win” situation of having cake and eating it is a focus issue that has attracted much attention. Under the existing research, how to improve both the “economy” and the “environment” is a controversial issue. In the past experience of environmental governance, the government usually used the administrative right to intervene directly. While this traditional governance approach can reduce the destruction of the ecological environment in a short term, the administrative regulation for the environment also may have a negative impact on long-term economic growth. According to the research conclusions of this paper, in addition to using administrative regulation, the government can also implement innovation-driven policy to protect the environment that is beneficial to both economic welfare and social welfare to earn “gold and silver mountains” while retaining “clear waters and green mountains”.

Finally, the results of the heterogeneity analysis in this paper also provided some directional guides for the strategic selection of the national innovative city pilot policy. When investigating the pilot cities for the national innovative city pilot policy, it is important to entirely analyze the city characteristics, tailor them to local conditions, and take overall factors into account such as the innovation capacity, industrial structure, and city size in order to efficiently release the economic and social benefits brought by the innovation-driven policy.