### Samples and data collection

In order to promote the implementation of the “Made in China 2025” strategy, From 2015 to 2018, the Ministry of Industry and Information Technology of China selected 305 enterprises nationwide as pilot enterprises for smart manufacturing. Because the data of listed enterprises is relatively complete and easy to obtain, based on the availability of data, this paper selects the enterprises listed in the Shanghai and Shenzhen stock markets as the research sample among the intelligent manufacturing pilot enterprises. The smart manufacturing pilot has been implemented since 2015, and the latest data is up to 2018. Among the smart manufacturing pilot enterprises, there are 59 enterprises listed on the Shanghai and Shenzhen stock exchanges. Excluding stocks with severe data loss and special handling (ST) and risk of delisting (* ST), In this paper, 52 listed enterprises are selected, and their data for 4 consecutive years are selected, with a total of 204 effective observations.

Data Sources. The data of “green invention patents” and “green utility model patents” used to measure green innovation come from Chinese Patent Full-text Database (CPFD) and are obtained through query and manual sorting. The environmental protection input data comes from the annual reports and social responsibility reports of listed enterprises. The names disclosed mainly include environmental protection expenditures, environmental protection inputs, greening environmental protection fees, and pollution control support. Control variables such as the size of the enterprise, the age of the enterprise, and the asset-liability ratio are all from China Stock Market & Accounting Research Database (CSMAR). At the same time, in order to prevent the abnormal value of each variable from affecting the estimation efficiency, the two ends of the variables with large differences were subjected to 0.95 and 0.05 tailing treatment.

### Research methods and variable measures

#### Research methods

Based on the analysis of the relationships among Intelligent Upgrade (INU), Environmental regulations (ER), Green Innovation (GI), and Environmental dynamism (ED), this paper proposes research hypotheses among INU, ER, GI, and ED. The data in China Stock Market & Accounting Research Database (CSMAR), Chinese Patent Full-text Database (CPFD), annual reports and social responsibility reports of listed companies are used. Constructed the correlation model between variables, and analyzed the correlation between variables. The research method is shown in Fig. 2.

#### Dependent variable “intelligent upgrade (INU)”

Since the enterprises selected in this article are all pilot enterprises for smart manufacturing identified by the Ministry of Industry and Information Technology of China, all the enterprises in the sample are enterprises in the process of intelligent transformation. Effectively ensure that the selected enterprises are in the process of intelligent upgrade. As for the measurement indicators of enterprise transformation and upgrading, scholars mostly use indirect indicators for alternative measurements. With regard to alternative variables for transformation and upgrading, Szirmai et al.^{56} and Ren et al.^{57} proposed that the result of enterprise transformation and upgrading, one of which is product transformation, is generally measured by the company’s new product sales, and the proportion of new product sales in total product sales (NPS) can be used as a substitution variable. The other is the transformation of production methods or production efficiency of enterprises. The transformation of production methods or efficiency will lead to the transformation of products from low added value to high added value. The essence is that there are differences in the productivity of enterprises, and the total factor productivity (TFP) of enterprises can be used as a substitute variable. The intelligent upgrading of manufacturing enterprises studied in this article refers to the transformation of production methods or products brought to listed enterprises through the use of advanced manufacturing technology, artificial intelligence and other new-generation information technologies. It includes both the transformation of production methods and the transformation of enterprise products, Therefore, this paper uses two alternative indicators to measure the intelligent transformation and upgrading of manufacturing enterprises, namely the proportion of NPS and TFP.

Enterprise TFP calculation. There are various methods for calculating the enterprise’s total factor productivity. This article is based on the research results of Jieyu et al.^{58}, Add a new input to the traditional Cobb–Douglas production function, called intermediate input (FI), By taking the natural logarithm of each variable on both sides of the traditional Cobb Douglas production function function, the following model Eq. (1) is obtained:

$$ln TO_{{{text{i}},{text{t}}}} = c + alpha ln LI_{{{text{i}},{text{t}}}} + beta ln {text{CI}}_{{{text{i}},{text{t}}}} + gamma ln FI_{{{text{i}},{text{t}}}} + mu_{{{text{i}},{text{t}}}}$$

(1)

where TO_{i,t} is the total output of the “i” company in the “t” year, This article uses the inventory change of the enterprise and the total amount of main business income to measure. LI_{i,t} is the labor input of the “i” company in the “t” year. The labor input includes not only the input of labor factors, but also the efficiency of labor output and the quality of input factors. And labor compensation can better explain the changes in labor input. Therefore, this article uses paid employee salaries (listed in the company’s cash flow statement) to measure. CI_{i,t} is the capital investment of the “i” company in the “t” year, mainly refers to the total capital stock of listed enterprises, Not only includes the current and fixed assets related to the company’s production and operation, but also other assets that provide guarantee services for the company’s production. This article uses the year-end fixed assets (listed company’s balance sheet) to measure. FI_{i,t} is the input of various other factors of the “i” company in the “t” year, This article uses purchased products and services (listed in the company’s cash flow statement) to measure. “α” represents the output elasticity of labor input (LI), “β” is the output elasticity of capital input (CI), “γ”is the output elasticity of the input of other factors (FI), “μ”is the random error term. The value range of “i” is 1–52, The value range of “t”is 2015, 2016, 2017, 2018. Based on the existing data of listed enterprises, Eviews 6.0 software was used to analyze the above Cobb–Douglas production function, Regression of panel data using fixed effect methods, The residuals obtained are TFP of each listed company. The specific values of TFP of each listed company are shown in Table 1.

#### The independent variable “environmental regulations (ER)”

Environmental regulations is a necessary way for government departments to maintain the environment. Current research is mainly measured from the following perspectives. The first is the policy perspective, which can measure the change of the “three wastes” before and after the environmental policy is formulated, use this to determine the intensity of environmental regulations^{59}; Secondly, from the perspective of environmental protection costs, enterprises can use the expenses incurred in the treatment of solid waste, water and air as indicators for measuring environmental regulations^{60}; Third, from the perspective of pollution intensity, the proportion of pollution emissions in industrial output value can be used as an indicator for measuring environmental regulations^{61}; Fourth, from the perspective of national income, the per capita income of a region or GDP per capita (GDP) can be used as an indicator to measure the environmental regulations of the region^{62}. Since this article analyzes the impact of environmental regulations on green innovation and intelligent upgrading from an enterprise perspective, Therefore, this article selects the measurement method from the perspective of environmental protection costs, and considers that the enterprise’s environmental protection investment is an important form of environmental regulations, At the same time, in order to exclude the influence of the scale of the enterprise, the company’ s main business income index is introduced, The company’s environmental protection investment accounts for the proportion of the company’s main business income as an indicator for specific measurement of environmental regulations.

#### Intermediary variable “green innovation (GI)”

At present, the measurement of green innovation is mostly based on data analysis from the perspective of the industry as a whole, often measured by the level of energy consumption reduced during the production of new products^{63}, But this method is only suitable for measuring from the perspective of the industry as a whole. Some scholars use the number of green patents of each enterprise to measure the green innovation level of the enterprise^{64}. Since this paper needs to measure the level of green innovation of enterprises, this paper draws on the methods of Xu et al.^{64} to measure green innovation by the number of green innovation patents. Taking the natural logarithm of the number of green innovation patents of each listed company represents the green innovation of the enterprise. Among them, the number of green patents includes green invention patents and green utility model patents.

#### Adjustment variable “environmental dynamism (ED)”

Cheng et al.^{65} proposed that the performance fluctuation of an enterprise can approximately measure the environmental dynamism, Ghosh et al.^{66} proposed to use the adjusted industry sales revenue standard deviation to measure the dynamics of the corporate environment. On the basis of previous research, Shen Huihui and others deleted some of the content of the business income that was rising steadily, and achieved a more accurate measurement of environmental dynamism. This article refers to Huihui et al.’s^{67} measurement method of environmental dynamism, using ordinary least squares. Based on the sales revenue of each listed company in the past 5 years, the standard deviation of each listed company’s abnormal sales revenue in the past 5 years is estimated. The ratio of the standard deviation to the average of the company’ s sales revenue over the last five years is taken as the unadjusted environmental dynamism (UED) of the company. Then, calculate the median of the unadjusted environmental dynamism of all enterprises in each industry in each year to obtain the annual environmental dynamism (AED) of each industry. Using the method adopted by Ghosh et al.^{66}, dividing the enterprise’s UED by AED of each industry, the final value of the environmental dynamism (ED) is obtained.

#### Control variables “enterprise size (ES), business age (BA), asset-liability ratio (AL)”

ES is expressed by the natural logarithm of the ending value of the total assets of listed enterprises in the manufacturing industry, BA is expressed by the natural logarithm of the time from the establishment of the listed manufacturing company to the statistical year, AL is expressed as the ratio of total liabilities and total assets of a listed manufacturing company.

### Empirical model

In order to examine the relationship between green innovation, environmental dynamism and intelligent upgrading of manufacturing enterprises, this paper refers to the method proposed by Huang and Song^{68}, At the same time, considering the elimination of possible heteroscedasticity and the hysteresis of environmental regulations and green innovation^{69}, the measurement model is set as follows:

Without adding the independent variable and the adjustment variable, Only considering the influence of control variables (ES, BA, AL) on GI, Establish metrology model 1 as shown below:

$$ln GI_{{text{i,t}}} = alpha_{{text{i,t1}}} + beta_{11} ln ES_{{text{i,t}}} + beta_{12} ln BA_{{text{i,t}}} + beta_{13} ln AL_{{text{i,t}}} + varepsilon_{{text{i,t1}}}$$

(2)

where α_{i,t1} is a constant term, β_{11}–β_{13} regression coefficients for control variables ES, BA, and AL, respectively. “t” is the year,” i” is the ”i” company, ε _{i,t1} is the random error term.

Based on the model 1, the independent variable ER is included in the model, Establish metrology model 2 as shown below:

$$ln GI_{{text{i,t}}} = alpha_{{text{i,t2}}} + beta_{21} ln ER_{{text{i,t} – 1}} + beta_{22} ln ES_{{text{i,t}}} + beta_{23} ln BA_{{text{i,t}}} { + }beta_{24} ln AL_{{text{i,t}}} + varepsilon_{{text{i,t2}}}$$

(3)

Without adding independent variables and adjustment variables, only consider the influence of control variables ES, BA, AL on INU, Establish metrology model 3 as shown below:

$$ln INU_{{text{i,t}}} = alpha_{{text{i,t3}}} + beta_{31} ln ES_{{text{i,t}}} + beta_{32} ln BA_{{{text{i}},{text{t}}}} { + }beta_{33} ln AL_{{text{i,t}}} + varepsilon_{{text{i,t3}}}$$

(4)

Based on the model 3, the independent variable ER is included in the model, and the econometric model 4 is established, as follows:

$$ln INU_{{text{i,t}}} = alpha_{{{text{i}},{text{t}}4}} + beta_{41} ln ER_{{text{i,t} – 1}} + beta_{42} ln ES_{{text{i,t}}} + beta_{43} ln BA_{{text{i,t}}} { + }beta_{44} ln AL_{{text{i,t}}} + varepsilon_{{text{i,t4}}}$$

(5)

Based on the model 3, the intermediary variable GI is included in the model, and the econometric model 5 is established, as follows:

$$ln INU_{{text{i,t}}} = alpha_{{text{i,t5}}} + beta_{51} ln GI_{{text{i,t} – 1}} + beta_{52} ln ES_{{text{i,t}}} + beta_{53} ln BA_{{text{i,t}}} { + }beta_{54} ln AL_{{{text{i}},t}} + varepsilon_{{text{i,t5}}}$$

(6)

Based on the model 4, the intermediary variable GI is included in the model, and the econometric model 6 is established, as follows:

$$begin{aligned} ln INU_{{text{i,t}}} & = alpha_{{text{i,t6}}} + beta_{61} ln ER_{{text{i,t} – 1}} + beta_{62} ln GI_{{text{i,t}-1}} hfill \ & quad + beta_{63} ln ES_{{text{i,t}}} + beta_{64} ln BA_{{text{i,t}}} { + }beta_{65} ln AL_{{text{i,t}}} + varepsilon_{{text{i,t6}}} hfill \ end{aligned}$$

(7)

Based on the model 3, the intermediary variable GI and manipulated variable ED are included in the model, and the econometric model 7 is established, as follows:

$$begin{aligned} ln INU_{{text{i,t}}} & = alpha_{{text{i,t7}}} + beta_{71} ln GI_{{text{i,t}-1}} + beta_{72} ln ED_{{text{i,t}}} + beta_{73} ln ES_{{text{i,t}}} hfill \ & quad + beta_{74} ln BA_{{text{i,t}}} { + }beta_{75} ln AL_{{text{i,t}}} + varepsilon_{{text{i,t7}}} hfill \ end{aligned}$$

(8)

Based on the model 7, in order to check the adjustment result of the environment dynamics, Product of mediator variable green innovation and moderator variable environmental dynamism is introduced as the interaction term, and the measurement model 8 is established as follows:

$$begin{aligned} ln INU_{{text{i,t}}} & = alpha_{{text{i,t8}}} + beta_{81} ln GI_{{text{i,t}- 1}} + beta_{82} ln ED_{{text{i,t}}} + beta_{83} ln GI_{{text{i,t} – 1}} hfill \ & quad times ED_{{text{i,t}}} + beta_{84} ln ES_{{text{i,t}}} + beta_{85} ln BA_{{text{i,t}}} { + }beta_{86} ln AL_{{text{i,t}}} + varepsilon_{{text{i,t8}}} hfill \ end{aligned}$$

(9)

### Descriptive statistics and correlation coefficient analysis

Before conducting hypothesis testing, this article has performed descriptive statistics and correlation analysis on each relevant variable. The results are shown in Tables 2 and 3. Among them, Table 2 provides statistics and descriptions of the mean, standard deviation, minimum and maximum values of dependent variable intelligent upgrade, independent variable environmental regulations, intermediary variable green innovation, moderating variable environmental dynamism, and control variable enterprise size , business age , asset-liability ratio. Table 3 examines the correlation between the dependent variable intelligent upgrade, independent variable environmental regulations, intermediary variable green innovation, moderating variable environmental dynamism and control variable enterprise size , business age , asset-liability ratio. The correlation coefficient between the intelligent upgrade of manufacturing enterprises and environmental regulations is 0.301, which is significant at the level of 0.01; The correlation coefficient between the intelligent upgrade of manufacturing enterprises and the environmental dynamism is 0.193, which is significant at the level of 0.05. There are different degrees of correlation between the intelligent upgrading of manufacturing enterprises and various control variables.