3.3.3. Random Effects Model

For the coldiag2 test, the condition number using scaled variables was 13.02; thus, the variables passed the collinearity test [49]. To ensure the accuracy and credibility of the estimated results of the F test, the fixed-effect model is considered to be significantly better than mixed regression. Further, considering the addition of dummy variables and according to the least squares dummy variables (LSDV) method, a Hausman test was used to determine the use of random effects model.

#### 3.3.4. Spatial Regression Model

Due to the significant spatial correlation of environmental innovation with cities in China, a spatial panel regression model was considered for the detection of the determinants. By comparing the spatial lag model (SLM), the spatial error model (SEM), and the spatial Durbin model (SDM) based on Stata 12.0 analysis, it was found that the goodness of fit and credibility of the SDM (*R*<sup>2</sup> = 0.863) was the highest among the three models, and the Hausman test results showed that the SDM random effects passed the robustness test. Therefore, this paper selected the SDM with random effects to analyze the determinants of environmental innovation of cities in China:

Ln*EICi*,*<sup>t</sup>* = *αi*+ *β*1Ln*ERi*,*<sup>t</sup>* + *β*2Ln*U* − *Si*,*<sup>t</sup>* + *β*3Ln*UFDIi*,*<sup>t</sup>* + *β*4Ln*U* − *TICi*,*<sup>t</sup>* + *β*5Ln*U* − *IEICi*,*<sup>t</sup>* + *β*6Ln*U* − *CSi*,*<sup>t</sup>* <sup>+</sup>*β*7Ln*<sup>U</sup>* <sup>−</sup> *EDLi*,*<sup>t</sup>* <sup>+</sup> *<sup>β</sup>*8Ln*<sup>U</sup>* <sup>−</sup> *ISi*,*<sup>t</sup>* <sup>+</sup> *<sup>β</sup>*9Ln*<sup>U</sup>* <sup>−</sup> *ALi*,*t*+*<sup>λ</sup>* <sup>11</sup> ∑ *k*=1 *wij*Ln*EICj*,*<sup>t</sup>* + *θ*<sup>1</sup> 11 ∑ *j*=1 *wij*Ln*ERj*,*<sup>t</sup>* +*θ*<sup>2</sup> 11 ∑ *j*=1 *wij*Ln*U* − *Sj*,*<sup>t</sup>* + *θ*<sup>3</sup> 11 ∑ *j*=1 *wij*Ln*U* − *FDIj*,*t*+*θ*<sup>4</sup> 11 ∑ *j*=1 *wij*Ln*U* − *TIOCj*,*<sup>t</sup>* +*θ*<sup>5</sup> 11 ∑ *j*=1 *wij*Ln*U* − *IEICj*,*<sup>t</sup>* + *θ*<sup>6</sup> 11 ∑ *j*=1 *wij*Ln*U* − *CSj*,*<sup>t</sup>* + *θ*<sup>7</sup> 11 ∑ *j*=1 *wij*Ln*U* − *EDLj*,*<sup>t</sup>* +*θ*<sup>8</sup> 11 ∑ *j*=1 *wij*Ln*U* − *ISj*,*<sup>t</sup>* + *θ*<sup>9</sup> 11 ∑ *j*=1 *wij*Ln*U* − *ALj*,*<sup>t</sup>* (4)

where *i* indexes the city, and *t* indexes time.

### **4. Empirical Results**

#### *4.1. Environmental Innovation of Cities in China*

From 2007 to 2017, the number of cities participating in environmental innovation in China increased from 330 in 2007 to 338 in 2017, among which the number of cities engaging in innovation around building technology was always the largest, while the number of cities engaging in innovation around greenhouse gas technology was always the smallest. Spatially, China's environmental innovation activities showed significant spatial heterogeneity, highly concentrated in a few cities (Table 5).

Specifically, in 2007, the top 10 cities in environmental innovation accounted for 48.1% of the environment-related patent applications in China. There were four cities with more than 5000 environment-related patents, namely Shanghai (9003), Beijing (8073), Guangzhou (6689), and Shenzhen (6174). Shanghai also ranked first in patent applications in the field of greenhouse gases, building, water adaptation, and transportation technology, while Beijing ranked first in the field of environmental management and energy technology. In different technical fields, most of the cities with outstanding performance were in eastern China. Cities in central and western China were generally backward in environmental innovation. In 2012, the proportion of the top 10 cities in environmental patents fell to 43.3%. Beijing not only surpassed Shanghai in terms of total volume, but also ranked first in all six technical fields. Chengdu and Xi-An in the central and western regions ranked sixth and seventh with 3852 and 3822 patent applications, respectively. Similarly, in different technical fields, cities with outstanding performance were mostly located in east China. Environmental innovation in central and western China was still generally backward. By 2017, Beijing continued to rank first with 26,224 patent applications and Shenzhen ranked second with 18,997 patents. Shanghai, Guangzhou, and Suzhou ranked third, fourth, and fifth with 14,801, 11,800, and 10,659 patent applications, respectively. Foshan ranked tenth with 7059 environmental patent applications. Its environmental patent applications mainly came from the field of water-related adaptation, accounting for 57.7% of the total. This is mainly because the Midea Group, which is headquartered in Foshan, applied for more than 2000 patents in the field of water-related adaptation technology in 2017. In the six categories of environment-related technologies, Beijing still ranked first in environmental management, energy, greenhouse gases, and transportation technologies, while Shenzhen ranked first in the technologies of building and water-related adaptation.

**Table 5.** Top 10 cities with the most environment-related technologies (2007, 2012, and 2017).


Note: The underlined numbers indicate that the corresponding city ranked first in patent applications for this type of technology.

#### *4.2. Regression Results*

Our initial data covered all cities at the prefecture level and above in China; however, due to the limitation of variable data, we finally selected 274 cities to enter the regression model. Table 6 shows the variable descriptive statistics, interpreted variables including urban EIC (total number of environmental patents), and the innovation ability of the cities in six types of environmental technology (the number of patent applications in each technology). Table 7 shows the descriptive statistics of the variables in three major regions of China. There were 114 cities in the eastern region, 108 in the central region, and 52 in the western region.

**Table 6.** Descriptive statistics of China city panel data and six types of environment-related technologies.


**Table 7.** Descriptive statistics for panel data of cities in three major regions of China.



**Table 7.** *Cont.*

Table 8 shows the regression results of the random effects model. The quadratic coefficients of environmental regulation were greater than 0, and the coefficients of the primary term were less than 0, which indicated that the impact of environmental regulation on the urban environmental innovation capability was a positive U-shaped curve that first declined and then rose. In other words, our research showed that urban environmental regulation had a restraining effect on urban environmental innovation in the short term [8]; however, they presented a positive correlation in the long term [12]. To avoid the interference of regional differences and improve the robustness and accuracy of the findings, we performed the same analysis on the sample data from three regions in China. Table 9 shows the regression results of different regions of China. We obtained the same results from the analysis, which showed that at the city level, the U-shaped relationship between China's environmental regulation and environmental innovation has a certain degree of stability.

Furthermore, we continued to investigate whether the negative effect of environmental regulation on environmental innovation had spatial spillover effects or not. Table 10 reports the results of the SDM. Since the SDM contains the spatial lag terms of both the explanatory variables and the explained variables, the partial differential method was adopted to decompose the spillover effects of the SDM into direct effect, indirect effect, and total effect. Among them, the direct effect reflects the impact on the city's environmental innovation, that is, the local effect. The indirect effect reflects the impact on the environmental innovation of the surrounding cities, that is, the spillover effect. The total effect is equal to the sum of the direct effect and the indirect effect. The R<sup>2</sup> was 0.852 and the spatial rho was significant at a significance level of 1%, with a coefficient of 0.213, indicating that environmental innovation in China city systems has a significant positive spatial spillover effect. Regarding the relationship between environmental regulation and environmental innovation, the regression results of the SDM were consistent with the results of the random effects model. That is, the effect of environmental regulation on environmental innovation was first to inhibit and then to promote.

Through the above regression results, we verified some existing findings. Firstly, there was a significant positive correlation between urban size and urban environmental innovation capability, which is consistent with the existing studies on the relationship between enterprise environmental innovation and enterprise size [21,22], and the relationship between regional environmental innovation and regional size [3,23,24]. Secondly, the higher the level of urban economic development was, the stronger the capability of urban environmental innovation was. This is also consistent with the existing research results of environmental economics and innovation economics. Thirdly, urban FDI was not only positively correlated with the urban environmental innovation capability, but also showed positive local effects and local spillover effects. This is consistent with existing research results [38,45,47,48]. A large amount of foreign capital investment promotes the improvement of local environmental innovation ability, and also forms a strong trickle-down effect

to the surrounding areas, promoting the environmental innovation of surrounding cities. Fourthly, the urban initial environmental innovation capability was obviously conducive to the urban environmental innovation capability. The direct effect of the urban initial environmental innovation capability was 18.288, and the indirect effect was −22.658, both of which were significant at the 1% significance level, indicating that urban environmental innovation had strong path dependence characteristics, while it also reflected that the cities with good performance in environmental innovation in the early stage would form a siphon effect, thus restraining the environmental innovation of surrounding cities.


**Table 8.** Regression results of the random effects model.

Note: \*\*\* *p* < 0.01, \*\* *p* < 0.05, \* *p* < 0.1.

**Table 9.** Regression results from different regions of China.


Note: \*\*\* *p* < 0.01, \*\* *p* < 0.05, \* *p* < 0.1.

In addition, we also found some very interesting conclusions about the control variables. Firstly, the industrial structure dominated by the secondary industry not only did not promote urban environmental innovation, but also played an obvious inhibitory effect. This finding is clearly different from the existing related research [3,12]. To verify the correctness of this result, we replaced this variable with the proportion of tertiary industry and performed a regression analysis again and found that the industrial structure dominated by the tertiary industry had a significant role in promoting the environmental innovation of the city itself and its surrounding cities. The reason for this result may be that the traditional manufacturing industry is still the pillar industry in China's cities dominated by secondary industry, and patent applications are mostly completed by the tertiary industry, represented by the information and communication industry, real estate industry, and scientific research. Secondly, the scale of urban construction measured by the scale of urban fixed asset investment had a significant negative effect on urban environmental innovation capability. As an important driving force of economic growth, China's urban fixed asset investment is growing rapidly. However, while stimulating economic growth, it has also increased energy consumption and environmental pollution. Many studies have found

that there is a significant positive correlation between urban fixed asset investment and urban environmental pollution emissions in China [50]. Thirdly, there was a negative correlation between the urban administrative level and urban environmental innovation capability. The reason for this may be that some non-capital cities have performed very well in environmental innovation, such as Suzhou, Ningbo, Shenzhen, and Foshan. However, this negative correlation did not pass the significance test, which also showed that the relationship between the two needs more verification.


**Table 10.** Results of the spatial Durbin model (SDM).

Note: \*\*\*, *p* < 0.01; \*\*, *p* < 0.05; \*, *p* < 0.1.

### **5. Conclusions**

Exploring the relationship between environmental regulation and environmental innovation is the core topic of environmental economics, innovation economics, and other research fields, and it is also one of the emerging topics in the field of environmental economic geography in recent years. This study used the number of environmental patent applications to measure urban environmental innovation, and analyzed the role of urban environmental regulation on urban environmental innovation, from which we summarize the following key findings.

Firstly, the number of environmental patents in China has grown rapidly, from 87,691 in 2007 to 307,929 in 2017. From a technical perspective, technologies related to buildings have always dominated the development of environmental innovation of cities in China, while technologies in the field of greenhouse gases and water adaptation were quite unpopular throughout China. China's urban environmental innovation showed a significant spatial correlation. The Moran's I index of both the whole (urban EIC) and the six technical fields were significant at the 1% level and greater than 0. Additionally, in the time sequence, the values of Moran's index were increasing, which indicated that the spatial correlation was strengthened with the passage of time.

Secondly, the number of cities participating in environmental innovation in China increased from 330 in 2007 to 338 in 2017, among which the number of cities engaging in innovation around building technology was always the largest, while the number of cities engaging in innovation around the greenhouse gas technology was always the smallest. Spatially, China's environmental innovation activities showed significant spatial heterogeneity, highly concentrated in a few cities [43]. From 2007 to 2017, Beijing not only surpassed Shanghai in the total number of environmental patent applications, but also ranked first in the fields of environmental management, energy, greenhouse gas treatment, and transportation technology. Shenzhen also surpassed Shanghai in the total number of environmental patent applications, ranking second in China. At the same time, Shenzhen had the largest number of environmental patent applications in the field of building and water-related adaptation in China.

Thirdly, both the random effects model and the SDM model showed that there was a U-shaped relationship between China's urban environmental regulation and urban environmental innovation, which was not only consistent with what the Porter hypothesis advocates [9,12], but is also consistent with existing research on the relationship between green innovation and urban green development [51]. Moreover, the regression results of different technical fields and different regions verified this result. These results showed that environmental regulations will first restrict environmental innovation due to increased production costs. After a period of adaptation, environmental regulations will induce environmental innovation. We also found some interesting results in the control variables. Urban size, urban economic development level, and urban FDI all played positive roles in promoting urban environmental innovation, while the urban fixed asset investment scale and industrial structure dominated by the secondary industry significantly inhibited urban environmental innovation. In addition, we also found that urban environmental innovation had significant path dependence characteristics.

### **6. Discussion**

The discussion of the relationship between environmental regulation and environmental innovation in this article gives us many policy implications. Firstly, local governments should adhere to the innovation-driven development strategy, increasing R&D investment to promote the continuous growth of urban technological innovation capability. Secondly, local governments should continue to promote the transformation and upgrading of the urban industrial structure, especially increasing the supports for the producer service industry. Thirdly, local governments should increase the introduction of foreign capital to promote the upgrading of local production and management methods. Fourthly, local governments should conduct environmental assessments on the increasing urban fixed asset investment under the background of rapid urbanization. In addition, increasing investment in fixed assets will also reduce other financial expenditures. Fifthly, in view of the U-shaped relationship between environmental regulation and environmental innovation, local governments should adopt strategies that adapt to time and local conditions in environmental regulation.

We noticed that enterprises are becoming increasingly important as the largest actors in China's environmental innovation. Based on applicant information for environmental patents, we identified environmental innovation actors in four categories, namely universities and scientific research institutions, enterprises, individuals, and others. We found that the proportion of enterprises increased rapidly from 39.7% in 2007 to 70.2% in 2017. This showed that the difference in the spatial distribution of China's environmental innovation was seriously affected by the distribution of enterprises engaged in environmental innovation. Therefore, future research on China's environmental innovation must fully consider the impact of the spatial distribution characteristics of different types of enterprises.

There are also some limitations that may exist in our study. First, we limited our data to the patent applications from the Wanfang Patent Database, which cannot avoid criticism about neglecting other data sources. Second, using only one indicator (environmental patent applications) to measure the capacity of a city's environmental innovation is relatively weak, and building a richer evaluation system for city environmental innovation is a direction we will explore in the future. Third, although six types of environmental technologies were identified based on the work of the Organization for Economic Co-operation and Development (OECD), there are still some environmental technologies that we have not considered, such as adsorption cooling technology [52,53], advanced combustion technology, and emission reduction technology [54,55]. A broader environmental technology identification system should be constructed. Fourth, we considered the spatial relevance of urban environmental innovation but ignored the spatial diffusion of environmental innovation. In the era of innovation networking, cities participate in environmental innovation activities not only relying on their own development, but also relying on networks to

obtain innovation resources. Therefore, future environmental innovation research must fully consider the externality effect brought by the innovation network.

**Author Contributions:** Conceptualization, D.D.; methodology, D.D. and Q.X.; software, Q.X.; validation, D.D.; formal analysis, D.D. and Q.X.; investigation, D.D.; resources, D.D.; data curation, D.D. and Q.X.; writing—original draft preparation, D.D..; writing—review and editing, D.D.; supervision, Q.X.; project administration, D.D.; funding acquisition, D.D. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the National Natural Science Foundation of China grant number 41901139.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.
