Path Analysis of Green Finance-Driven Regional Technological Innovation in China: The Mediating Effect of R&D Investment

2. Methodology

2.1. Data Sources

This study uses data from 2008 to 2021, spanning a 14-year time span, in order to analyse the impact of China’s green finance practices on regional innovation capacity. Data sources include official publications and authoritative databases such as China Statistical Yearbook, China Regional Innovation Capacity Evaluation Report, China Financial Statistical Yearbook, China Science and Technology Statistical Yearbook, China Energy Statistical Yearbook, CSMAR database, CCEER database, and others. The 5easonn for choosing this time period is that the Opinions on Implementing Environmental Protection Policies and Regulations to Prevent Credit Risks, issued in 2007, is regarded as the starting point of China’s comprehensive green finance practice. The sample covers 30 provinces in mainland China (excluding Hong Kong, Macao, Taiwan and Tibet), totalling 420 observations. Data were selected based on availability and modelling needs, aiming to ensure the accuracy, representativeness and validity of the findings.

2.2. Introduction to Variables

2.2.1. Explained Variables

Regional Innovation Capacity (RIC). As a proxy variable for the explanatory variable regional innovation level, the composite score of the provincial-level ‘Comprehensive utility value of China's regional innovation capacity’ in the Report on the Evaluation of China's Regional Innovation Capacity (2008-2021), jointly published by the China Science and Technology Development Strategy Group and the Chinese Academy of Sciences, was used. Its measurements include the dimensions of provincial knowledge creation, knowledge acquisition, enterprise innovation, innovation environment and innovation performance, which can objectively evaluate the regional innovation capacity (see Table 1 for details).

Specifically, knowledge creation (15 per cent) is the cornerstone of innovation activities; knowledge acquisition (15 per cent) reflects the importance of open innovation; enterprise innovation (25 per cent), as the core driving force, directly reflects the degree of activity and effectiveness of regional innovation activities; innovation environment (25 per cent), as a supportive indicator, is the necessary guarantee for the continuation of innovation activities; and finally, innovation performance (20 per cent), as an outcome indicator, is the key basis for assessing the effectiveness of regional innovation capacity enhancement.

In summary, by adopting this comprehensive utility value as a proxy variable for regional innovation level and combining it with its multi-dimensional measurement indicators, we are able to understand and evaluate the regional innovation capacity in a more comprehensive and in-depth way, which provides a precise quantitative tool for empirical analyses, and helps to explore in-depth the internal mechanisms and external factors affecting the regional innovation capacity.

Table 1. Regional innovation capacity indicator system.

Level 1 indicators	Secondary indicators	Description of indicators	causality
Regional Innovation Capacity (ric)	Knowledge creation 15 per cent	Measuring a region's ability to generate new knowledge.	positive
	Knowledge acquisition 15 per cent	Measurement of a region's ability to utilise external knowledge and cooperation between industry, academia and research.	positive
	Enterprise innovation 25 per cent	Measures the ability of firms within a region to apply new knowledge, develop new technologies, utilise new processes, and manufacture new products.	positive
	Innovation environment 25 per cent	Measure the ability of a region to provide the appropriate environment for the generation, flow and application of technology.	positive
	Innovation performance 20 per cent	The ability to measure the benefits of innovation for the economic and social development of a region.	positive

Table 1. Regional innovation capacity indicator system.

Level 1 indicators	Secondary indicators	Description of indicators	causality
Regional Innovation Capacity (ric)	Knowledge creation 15 per cent	Measuring a region's ability to generate new knowledge.	positive
	Knowledge acquisition 15 per cent	Measurement of a region's ability to utilise external knowledge and cooperation between industry, academia and research.	positive
	Enterprise innovation 25 per cent	Measures the ability of firms within a region to apply new knowledge, develop new technologies, utilise new processes, and manufacture new products.	positive
	Innovation environment 25 per cent	Measure the ability of a region to provide the appropriate environment for the generation, flow and application of technology.	positive
	Innovation performance 20 per cent	The ability to measure the benefits of innovation for the economic and social development of a region.	positive

2.2.2. Explanatory Variables

Green Finance Development Index (GF), the index is constructed based on the entropy method, aiming to objectively reflect the level of green finance development. Green finance covers green loans, green securities, green insurance and green investment. In order to construct the green financial development index, this paper designs a set of index system, and objectively assigns weights to each index by entropy value method, in order to avoid the subjectivity of traditional expert scoring method. The steps of constructing green financial index by entropy value method are as follows:

①Indicator standardisation: The raw data are standardised to eliminate the effect of the scale. For positive indicators, the following formula is used for standardisation:

$$Y_{ij}={\displaystyle \frac{max\left(X_{ij}\right)-X_{ij}}{max\left(X_{ij}\right)-min\left(X_{ij}\right)}}$$

For negative indicators, the normalisation formula is adjusted to:

$$Y_{ij}={\displaystyle \frac{X_{ij}-min\left(X_{ij}\right)}{max\left(X_{ij}\right)-min\left(X_{ij}\right)}}$$

where$X_{ij}$ represents the observed value of the ith indicator in the jth region;$Y_{ij}$ is the standardised value.

②Information Entropy Calculation: The information entropy value calculates the information entropy value of each indicator using the standardized indicator value, which reflects the degree of dispersion of the indicator:

$$z_{ij}={\displaystyle \frac{Y_{ij}}{\sum _{j=1}^n Y_{ij}}}$$

$$E_i=-\mathrm{l}\mathrm{n}(n{)}^{-1}\sum _{j=1}^n z_{ij}\mathrm{l}\mathrm{n}{z}_{ij}$$

③Determination of indicator weights: The weights of the indicators are determined by the information entropy value $W_i$ ;

$$W_i={\displaystyle \frac{1-E_i}{\sum _{i=1}^n {(1-E}_i)}}$$

④Synthesis of green financial development index: Finally, the green financial development composite index is calculated by combining the weights and standardized values of each indicator G:

$$G=a_{ni}{W}_i$$

where$a_{ni}$ is the standardised value of the ith indicator.

Through the above steps, we obtained an objective quantitative green finance development index, which is a detailed and specific index system (as shown in Table 2), comprehensively covers multiple dimensions of green finance, accurately reflects its level of development, and lays a solid foundation of explanatory variables for the subsequent empirical analyses.

2.2.3. Control Variables

In order to ensure the accuracy of the estimation of the impact of green finance on regional innovation, the selection of control variables in this paper takes full account of the region's basic innovation environment, actual resource inputs and external environment. Specifically, a series of control variables including industrial structure (ind), human capital (lnhes), urbanisation level (ur), science and technology emphasis (techi), carbon emissions (lnco2) and capital investment (capi) are introduced, and the following theoretical and empirical bases support the choice of these variables.

Firstly, industrial structure plays a key role in technological innovation. According to the theory of industrial economics, regions with a higher share of service and high-technology industries usually have stronger innovation capacity (Ge et al. 2023) [41]. Therefore, incorporating industrial structure as a control variable can more accurately measure the innovation potential of the region. Human capital directly affects the quality of regional human capital, which is the basic element to promote technological innovation. The higher the human capital, the stronger the region's innovation and technology absorption capacity, so the article needs to use it as a control variable. The level of urbanisation is also an important factor affecting regional innovation, as the urbanisation process is able to gather innovative resources and promote the diffusion and application of knowledge and technology, and controlling for this variable is crucial to accurately assessing innovation capacity.

In addition, the level of emphasis on science and technology is an important indicator of the level of support for innovation activities in a region. Higher S&T investment usually implies a stronger incentive to innovate, so this variable is crucial for an accurate assessment of regional innovation capacity. Carbon emissions, as a control variable for environmental factors, influence the demand and direction of green technology innovation, and controlling for this variable helps to exclude the external effects of environmental pressures on innovation activities. Finally, capital investment is a key resource for firms to carry out R&D and innovation activities, and regions with sufficient capital are more likely to support large-scale innovation activities, so controlling for capital investment helps to avoid overestimation or underestimation of innovation capacity due to regional capital differences.

By introducing these control variables, the analyses in this paper can reflect the actual impact of green finance on regional innovation more comprehensively and accurately, ensuring the robustness and reliability of the research results. In addition, this paper adopts a fixed-effects double clustering (Cluster) regression model to control for the invariant effects of province and time, thus further enhancing the robustness of the research results. Details of each variable are detailed in Table 3.

2.2.4. Mediating and Moderating Variables

In this study, the mediating variable is the intensity of R&D investment (RD), which serves as a key indicator of regional innovation capacity and reflects the level of regional financial investment in S&T innovation. This variable is quantified by the ratio of R&D expenditure to regional gross domestic product (GDP), which reveals the level of regional emphasis on S&T R&D activities and its support for the innovation-driven development strategy. Theoretical analyses anticipate that the development of green finance may indirectly promote regional technological innovation capacity by enhancing the intensity of R&D investment, thus establish an intrinsic link between green finance and regional technological innovation.

Research and development (R&D) investment intensity (RD) is selected as a mediator variable, aiming to explore how green finance can enhance regional innovation capacity by promoting R&D activities. The importance of R&D investment in promoting technological innovation has been widely recognised, and both the Diffusion of Innovation Theory (ROGERS 1962) [42]and the Resource Base Theory (Wernerfelt 1984)[43]emphasise that R&D investment is the core driver of technological innovation and maintaining competitive advantage for enterprises. By increasing R&D investment, firms can develop new technologies and improve products more effectively, thus enhancing their overall innovation capability. Therefore, RD as a mediating variable is of great significance in this study. In addition, existing empirical studies have shown that RD investment has a significant mediating role between green finance and technological innovation, especially in resource-intensive and technology-intensive industries, and this mechanism is particularly obvious (Tang et al. 2023) [44]. The introduction of RD as a mediating variable not only helps to reveal the path of green finance affecting regional innovation, but also fills the research gap in the existing literature in exploring the specific mechanism between green finance and innovation.

In this paper, the regional innovation and entrepreneurship capacity (IU) was selected as a moderating variable, aiming to deeply explore the impact of green finance on regional innovation. As an intrinsic factor, regional innovation and entrepreneurship capacity (IU) reflects the ability of enterprises and individuals in the region in innovative ideas, technological development and entrepreneurial activities. According to the theory of diffusion of innovation, regions with stronger innovation and entrepreneurship capabilities are more active in technological innovation activities, and green finance is more likely to play a significant facilitating role in such regions. Therefore, IU can moderate the relationship between green finance and regional innovation, and further promote the research, development and application of green technologies by enhancing the innovation atmosphere.

Combining these mediating and moderating variables, this study aims to reveal the complex interaction mechanism between green finance and regional technological innovation. Through rigorous empirical analyses, this study can not only provide a supplement to existing theories, but also provide policymakers with strong decision support in promoting green transformation and technological innovation in regional economies.

2.3. Data Description

To ensure the stability and accuracy of the data, this paper performs logarithmic transformation on the sample variables and applies a 1% Winsorizing treatment to the continuous variables. Descriptive statistical analysis based on 420 observations shows that the mean value of regional innovation capacity (lnric) is 3.359, with a standard deviation of 0.309, indicating that there are significant differences in the innovation capacity of regions; the median is 3.315, implying that the innovation capacity of most of the regions is relatively low, and needs to be further improved. The mean value of the green financial development index (gf) is 0.152, with a standard deviation of 0.063, and the median is slightly lower than the mean, suggesting that the development of green finance in most regions is still at a low level and is more widely distributed (the minimum value is 0.072, and the maximum value is 0.450). Control variables, such as industrial development (ind), human capital (hes), urbanisation rate (ur) and technological level (techi), show some regional variability, which may affect regional innovation capacity. The large fluctuations in carbon dioxide emissions (co2) suggest the need to consider the role of environmental factors on technological innovation, while the relative stability of capital investment (capi) may indicate its supportive role in technological innovation. Detailed statistical results are shown in Table 4.

To detect the problem of multicollinearity in the model, the variance inflation factor (VIF) and tolerance (Tolerance) were calculated in this study. Table 5 shows the VIF values and Tolerance of the respective variables. From the results in Table 5, the VIF values for all the variables are below 3, with a maximum value of 2.93 (variable techi), which is much lower than the commonly used threshold value of 10. The average VIF value of 2.01 is also at a low level, which indicates that the problem of multicollinearity in this model is not significant. In addition, the tolerance of each variable is greater than 0.1 (minimum value of 0.341, variable techi), which also indicates that there is no significant covariance problem among the independent variables of the model.

2.4. Empirical Model

2.4.1. Fixed Effects Model

In order to comprehensively explore the impact of green finance on regional innovation capacity, this study adopts a fixed effects model with the basic model set as follows:

$${RIC}_i={\alpha }_0+{\alpha }_{1}G{F}_{it}+{\alpha }_{2}contro{l}_{it}+{\lambda }_i+{\mu }_t+{ϵ}_{it}$$

(1)

which${RIC}_i$ Represents regional innovation capacity;$G{F}_{it}$ represents green finance, and$contro{l}_{it}$ are the control variables, the${\lambda }_i$ and${\mu }_t$ represent regional and year control variables, respectively;${ϵ}_{it}$ represents the random error term.

2.4.2. Mediated Effects Model

In order to deeply investigate the internal transmission mechanism of "green finance - regional innovation capacity", this study constructs the following recursive equation system to analyse how green finance affects regional innovation capacity through R&D investment intensity (RD):

Effect of green finance on mediating variables (path a).

$${RD}_{it}={\mathsf{\delta }}_0+{\mathsf{\delta }}_{1}G{F}_i+{\mathsf{\delta }}_{2}Control{s}_{it}+{\lambda }_i+{\mu }_t+{ϵ}_{it}$$

(2)

Impact of mediating variables on regional innovation capacity (path b).

$${RIC}_i={\epsilon }_0+{\epsilon }_1{RD}_{it}+{\epsilon }_{2}G{F}_i+{\epsilon }_{3}Control{s}_{it}+{\lambda }_i+{\mu }_t+{ϵ}_{it}$$

(3)

where ${RD}_{it}$ is the mediating variable. Following the research line of Wen and Ye (2014) [45], the first step is to verify the coefficient ${\mathsf{\delta }}_1$ in equation (2) and the coefficient ${\epsilon }_1$ in equation (3).If both coefficients exist, the mediating effect is significant; the second step verifies the coefficient ${\epsilon }_2$ in equation (3) , if it is significant, it indicates the existence of partial mediation effect, otherwise it is a full mediation effect; the third step compares${\mathsf{\delta }}_1$ and${\epsilon }_1{\epsilon }_2$ sign, if the sign is same, it is a partial mediation effect, the weight of the mediation effect is ${\mathsf{\delta }}_1{\epsilon }_1$ /${\alpha }_1$ .The third step is to compare the signs of ${\mathsf{\delta }}_1\mathrm{a}\mathrm{n}\mathrm{d}{\epsilon }_1{\epsilon }_2$. If the sign is different, it is a masking effect, and the proportion of the mediating effect is |${\mathsf{\delta }}_1{\epsilon }_1$ /${\alpha }_1$ |.

2.4.3. Modelling the Mediating Effects of Regulation

In order to investigate the effect of moderating variables on the mediating effect, this study constructed a mediating effect model of moderation:

$${RD}_{it}=a_0+a_{1}G{F}_i+a_2{W}_{it}+a_3(G{F}_{it}\times W_{it})+a_{4}Control{s}_{it}+{\lambda }_i+{\mu }_t+{ϵ}_{it}$$

(4)

$${RIC}_i=b_0+b_{1}G{F}_i+b_2{M}_{it}+b_3{W}_{it}+b_4(G{F}_{it}\times W_{it})+b_5({RD}_{it}\times W_{it}){+b}_{6}Control{s}_{it}+{\lambda }_i+{\mu }_t+{ϵ}_{it}$$

(5)

According to the test of Wen and Ye (2014) [46], the coefficient $a_3$ of the interaction term in equation (4) is first tested . If it is significant, it indicates that the moderator variable (W) has a moderating effect on the relationship between the independent variable (GF) and the mediator variable (${RD}_{it}$ ) has a moderating effect on the relationship between the independent variable (GF) and the mediator variable (${RD}_{it}$); then verify the coefficient $b_5$ of the interaction term in Equation (5), which, if significant ,indicates that the moderator variable (W) has a moderating effect on the relationship between the mediator variable (${RD}_{it}$ ) and the dependent variable (RIC), thus proving that there is a mediating effect of moderation. The path diagram of the relevant influence mechanism is shown in Figure 1:

3. Empirical Results

3.1. Analysis of Baseline Regression

This study first tests the direct effect of green finance (gf) on regional innovation capacity (lnric), hypothesising that green finance can promote regional innovation capacity. To this end, this paper constructs a variety of fixed effects models to test the hypothesis. The following are the results of the baseline regression analyses of the impact of green finance on regional innovation capacity, which are detailed in Table 6.

In Model 1, only the explanatory variable green finance (gf) is included in this paper. The results show that the regression coefficient of gf is 1.315 and significant at the 1% statistical level, indicating that the development of green finance has a significant positive contribution to regional innovation capacity. However, the goodness of fit of model 1 is low, with an adjusted R² of 0.067, indicating that the explanatory power of the model is more limited when only green finance is considered.

Model 2 adds province fixed effects and year fixed effects to model 1. The results show that the regression coefficient of green finance decreases to 0.214, but is still significant at the 1% statistical level. The adjusted R² of Model 2 significantly increases to 0.944, showing that the introduction of province and year fixed effects significantly enhances the explanatory power of the model.

Model 3 further introduces control variables based on model 1. After adding these control variables, the regression coefficient of green finance is 0.148, although no longer significant, but other control variables such as industrial structure, science and technology inputs and capital inputs have a significant positive impact on regional innovation capacity at the 1% statistical level, while carbon emissions have a significant negative impact on regional innovation capacity at the 5% level. The adjusted R² of model 3 is 0.771, indicating that the control variables have a certain enhancement on the explanatory power of the model.

Model 4 further adds control variables to model 2, and the results show that the regression coefficient of gf is 0.207 and is significant at the 1 per cent statistical level. This indicates that green finance still has a significant positive effect on regional innovation capacity even when controlling for both province and year fixed effects and multiple control variables. The adjusted R² of model 4 further increases to 0.954, indicating that the model has stronger explanatory power and robustness.

Overall, the results of the baseline regression analyses indicate that green finance has a significant positive contribution to regional innovation capacity, and the finding remains robust after the inclusion of control variables and fixed effects. This suggests that green finance is an important factor in promoting regional innovation.

Table 6. Baseline regression results.

	Model 1	Model 2	Model 3	Model 4
	lnric	lnric	lnric	lnric
gf	1.315*** (0.235)	0.214*** (0.052)	0.148 (0.133)	0.207*** (0.052)
ind			0.742*** (0.118)	0.757*** (0.133)
lnhes			0.174*** (0.017)	0.142** (0.057)
ur			0.337*** (0.089)	0.252 (0.300)
techi			13.187*** (0.850)	2.097*** (0.477)
lnco2			-0.070*** (0.013)	-0.029* (0.015)
capi			0.427*** (0.138)	0.500*** (0.082)
Constant	3.159*** (0.039)	3.326*** (0.008)	2.142*** (0.095)	2.309*** (0.211)
N	420	420	420	420
R^2	0.069	0.950	0.775	0.960
Prov FE	NO	YES	NO	YES
Year FE	NO	YES	NO	YES
r2_a	0.067	0.944	0.771	0.954

Note: *, **, *** indicate significant at the 10 per cent, 5 per cent, and 1 per cent levels, respectively; standard errors are in () in the table.

Table 6. Baseline regression results.

	Model 1	Model 2	Model 3	Model 4
	lnric	lnric	lnric	lnric
gf	1.315*** (0.235)	0.214*** (0.052)	0.148 (0.133)	0.207*** (0.052)
ind			0.742*** (0.118)	0.757*** (0.133)
lnhes			0.174*** (0.017)	0.142** (0.057)
ur			0.337*** (0.089)	0.252 (0.300)
techi			13.187*** (0.850)	2.097*** (0.477)
lnco2			-0.070*** (0.013)	-0.029* (0.015)
capi			0.427*** (0.138)	0.500*** (0.082)
Constant	3.159*** (0.039)	3.326*** (0.008)	2.142*** (0.095)	2.309*** (0.211)
N	420	420	420	420
R^2	0.069	0.950	0.775	0.960
Prov FE	NO	YES	NO	YES
Year FE	NO	YES	NO	YES
r2_a	0.067	0.944	0.771	0.954

Note: *, **, *** indicate significant at the 10 per cent, 5 per cent, and 1 per cent levels, respectively; standard errors are in () in the table.

3.2. Analysis of Mediating Effects

This study further explores the mediating role of R&D investment intensity (rd) in the impact of green finance (gf) on regional innovation capacity (lnric). We hypothesise that green finance enhances regional innovation capacity by promoting an increase in R&D investment. To test this hypothesis, we constructed a regression model (Model 5) with mediating variables.

In Model 5, we find that green finance has a significant positive effect on R&D investment intensity (coefficient of 0.004, p < 0.1), indicating that green finance can effectively incentivise regions to increase R&D investment. In addition, the positive effect of R&D investment intensity on regional innovation capability (lnric) is also confirmed (coefficient of 9.464, p < 0.01), suggesting that R&D investment is an important driver of regional innovation capability. Nevertheless, the coefficient of green finance is still significant (coefficient of 0.168, p < 0.05), suggesting that there is a partial mediating effect. The Sobel Z-value of mediating effect is 2.347 (p < 0.05), the bootstrap Z-value is 2.13 (p < 0.05), and the mediating effect accounts for 49% of the total, and all these results indicate that R&D investment intensity plays a partially mediating role in the impact of green finance on regional innovation capacity. In short, green finance not only directly promotes the enhancement of regional innovation capacity, but also indirectly promotes it by increasing the intensity of R&D investment, where the intensity of R&D investment plays a significant mediating role.

In addition, this study explores the moderating mediating effect of regional innovation and entrepreneurship (lniu) in the mechanism of green finance affecting regional innovation capacity through R&D investment intensity. We hypothesise that regional innovation and entrepreneurship capacity may influence the impact of green finance on R&D inputs, which in turn changes the mediating effect of R&D inputs on regional innovation capacity. To test this hypothesis, we constructed a moderated mediation effect model (Model 6) containing interaction terms based on the mediation effect model.

In Model 6, the coefficient of the interaction term (gf × lniu) for regional innovation and entrepreneurship capacity (lniu) is 0.015, which is significant at the 10% significance level (p < 0.10), suggesting that regional innovation and entrepreneurship capacity significantly enhances the positive impact of green finance on R&D inputs, which may further enhance the mediating effect of green finance on regional innovation capacity through R&D inputs. In addition, the positive effect of R&D investment on regional innovation capacity is also significant (coefficient of 11.469, p < 0.01).The results of both the Sobel test and the bootstrap test indicate that the mediation effect of regulation is significant (Sobel Z value of 2.103, p < 0.05; bootstrap Z value of -1.98, p < 0.05 ), and the mediation effect under moderation is 39.8%. All these results indicate that regional innovation and entrepreneurship capacity plays a moderating role in the mechanism of green finance's influence on regional innovation capacity, and this moderating role is realised through the mediating variable R&D input.

Table 7. Mediated effects test.

	Model 5		Model 6
	rd	lnric	rd	lnric
gf	0.004* (0.002)	0.168** (0.068)	-0.063* (0.031)	-1.609* (0.891)
rd		9.464*** (2.781)		11.469*** (3.368)
lniu			-0.006*** (0.001)	0.051 (0.046)
gf×lniu			0.015* (0.007)	0.412* (0.197)
control variable	YES	YES	YES	YES
Constant	0.015*** (0.004)	2.171*** (0.203)	0.026*** (0.005)	2.299*** (0.293)
N	420	420	420	420
R^2	0.945	0.961	0.951	0.963
Prov FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
r2_a	0.938	0.956	0.944	0.957
Sobel Z	2.347		2.103
Sobel Z-p value	0.019		0.035
bootstrap Z	2.13		1.98
bootstrap Z-p value	0.033		0.048
Percentage of intermediary effects	49 per cent		39.8 per cent

Note: *, **, *** indicate significant at the 10 per cent, 5 per cent, and 1 per cent levels, respectively; standard errors are in () in the table.

Table 7. Mediated effects test.

	Model 5		Model 6
	rd	lnric	rd	lnric
gf	0.004* (0.002)	0.168** (0.068)	-0.063* (0.031)	-1.609* (0.891)
rd		9.464*** (2.781)		11.469*** (3.368)
lniu			-0.006*** (0.001)	0.051 (0.046)
gf×lniu			0.015* (0.007)	0.412* (0.197)
control variable	YES	YES	YES	YES
Constant	0.015*** (0.004)	2.171*** (0.203)	0.026*** (0.005)	2.299*** (0.293)
N	420	420	420	420
R^2	0.945	0.961	0.951	0.963
Prov FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
r2_a	0.938	0.956	0.944	0.957
Sobel Z	2.347		2.103
Sobel Z-p value	0.019		0.035
bootstrap Z	2.13		1.98
bootstrap Z-p value	0.033		0.048
Percentage of intermediary effects	49 per cent		39.8 per cent

Note: *, **, *** indicate significant at the 10 per cent, 5 per cent, and 1 per cent levels, respectively; standard errors are in () in the table.

The quantile moderated mediation effect analysis further reveals the moderating role of regional innovation and entrepreneurship capability (lniu) on the mediation effect of research and development (R&D) investment (RD) at different quantile levels. Figure 2 shows that the coefficient of RD investment is higher at lower quartile levels of lniu, indicating that RD investment plays a more significant role in promoting regional innovation in regions with lower regional innovation and entrepreneurship capabilities. However, this impact coefficient shows a gradual decline as the lniu quartile increases and is no longer significant at the 90 per cent quartile. This finding suggests that the marginal promotional effect of R&D investment is weakened in regions with higher regional innovation and entrepreneurship capabilities, probably because other innovation drivers are more prominent in a high level of lniu, which relatively reduces the strength of the impact of R&D investment.

In summary, the results of the mediation effect analysis and quartile moderated mediation effect analysis in this study indicate that regional innovation and entrepreneurship (lniu) plays a significant moderating role in the process of green finance affecting regional innovation capacity through R&D investment intensity. The mediating effect of R&D inputs is more significant at lower quartile levels of lniu, while it is weakened at higher quartile levels. These findings provide insights into understanding the mechanism of green finance's effect on regional innovation and provide important references for the formulation of more targeted policies.

Figure 2. Quantile Moderation Effect Plot.

Figure 2. Quantile Moderation Effect Plot.

4. Robustness Tests

In Table 8, the impact of green finance on regional innovation capacity (lnric) is analysed through three types of robustness tests, and the results show that the model is highly robust and consistent.

Firstly, the results of Bootstrap method (column 1 of Table 8) show that the regression coefficient of green finance (gf) is 0.207 and significant at 1% significance level (p<0.01). This indicates that the positive impact of green finance on regional innovation capacity is effectively verified by sampling test through Bootstrap method, which enhances the reliability of the model results.

Secondly, robustness is verified by the replacement variable method (columns 2-5 of Table 8). In column 2, replacing regional innovation capacity with corporate innovation (lnci) as the explanatory variable, the regression coefficient of green finance is significantly increased to 0.684 (p<0.01), indicating that green finance plays a more prominent role at the micro level (e.g., corporate innovation), which supports the theory that green finance promotes corporate-level innovation. In column 3, using the first-order lagged term of green finance (gfl1) as the dependent variable, the regression coefficient is 0.243 (p<0.01), which maintains significance, confirming that the lagged effect of green finance has a positive impact on regional innovation capacity, further confirming the robustness of the lagged effect. In column 4, after adjusting and replacing the control variables, the regression coefficient of green finance is 0.614 (p<0.01), indicating that the impact of green finance on regional innovation capacity remains significant and stable even if the control variables are changed. In column 5, we further enrich the model by including the following additional control variables:

The degree of openness to the outside world (od), measured as the ratio of total imports and exports to GDP of the location of the business unit, reflects the degree of openness of the regional economy and the degree of integration into the international market, which may have an impact on regional innovation capacity.

The degree of attention to environmental protection (ep), measured by the ratio of local financial expenditure on environmental protection to the general budget expenditure of local finance, reflects the local government's financial investment in environmental protection, which may be related to the effect of the role of green finance.

Traditional financial development (fd), measured as the ratio of value added of the financial industry to GDP, represents the level of development of the traditional financial sector, and its interaction with green finance may affect regional innovation capacity.

Area size (lnsize), which uses the natural logarithm of the number of household population at the end of the year as a proxy variable, reflects the size of the area, which may affect the clustering of innovation resources and the development of innovation activities.

In the model in column 5, the regression coefficient for green finance is 0.573 (p<0.01), and the effect of green finance on regional innovation capacity remains significant even in the complex model setup that takes into account these additional control variables, which provides further evidence of the original findings and demonstrates the robustness of the model under more comprehensive variable considerations.

Finally, the robustness test conducted by shortening the time window (column 6) to the last ten years (2012-2021) shows that the regression coefficient of green finance increases to 0.773 (p<0.01), and the significance level remains at 1%. This result indicates that green finance contributes more significantly to regional innovation capacity in a shorter time frame, further demonstrating the applicability and robustness of the model across different time periods.

In summary, the results of the above three types of robustness testing methods all show the significance and consistency of the impact of green finance on regional innovation capacity, proving the robustness of the research model and the reliability of the results. These findings further support the important role of green finance in promoting regional innovation and provide strong theoretical support for the formulation of related policies.

Table 8. Robustness tests.

	(1)	(2)	(3)	(4)	(5)	(6)
VARIABLES	lnric	lnci	lnric	lnric	lnric	lnric
gf	0.207***	0.684***		0.614***	0.573***	0.773***
	(0.053)	(0.180)		(0.156)	(0.163)	(0.200)
gfl1			0.243***
			(0.059)
ind	0.757***	1.562***	0.834***		1.286**	3.019***
	(0.135)	(0.500)	(0.131)		(0.519)	(0.590)
is				-1.705***
				(0.462)
lnhes	0.142**	-0.213	0.144*		-0.413	-0.552
	(0.059)	(0.255)	(0.071)		(0.263)	(0.330)
hep				-35.255
				(26.260)
ur	0.252	0.636	0.100	0.294	2.822**	2.742
	(0.301)	(0.760)	(0.308)	(0.842)	(1.094)	(1.694)
techi	2.097***	4.302**	2.225***	5.425***	2.040	4.526*
	(0.475)	(1.762)	(0.493)	(1.483)	(1.358)	(2.284)
lnco2	-0.029*	-0.028	-0.036*		-0.069	-0.095
	(0.015)	(0.049)	(0.018)		(0.048)	(0.058)
lnso2				0.086*
				(0.042)
capi	0.500***	0.997**	0.497***	1.135***	0.844**	0.730*
	(0.088)	(0.354)	(0.111)	(0.348)	(0.339)	(0.374)
od					-0.521**
					(0.176)
ep					-0.516
					(1.455)
fd					-0.323
					(1.347)
lnsize					0.887*
					(0.449)
Constant	2.309***	3.076**	2.380***	3.524***	-3.983	3.066***
	(0.213)	(1.060)	(0.219)	(0.454)	(3.251)	(0.887)
Observations	420	420	390	420	420	300
R-squared	0.960	0.868	0.960	0.871	0.880	0.893
Prov FE	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
r2_a	0.960	0.851	0.955	0.853	0.863	0.874

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Table 8. Robustness tests.

	(1)	(2)	(3)	(4)	(5)	(6)
VARIABLES	lnric	lnci	lnric	lnric	lnric	lnric
gf	0.207***	0.684***		0.614***	0.573***	0.773***
	(0.053)	(0.180)		(0.156)	(0.163)	(0.200)
gfl1			0.243***
			(0.059)
ind	0.757***	1.562***	0.834***		1.286**	3.019***
	(0.135)	(0.500)	(0.131)		(0.519)	(0.590)
is				-1.705***
				(0.462)
lnhes	0.142**	-0.213	0.144*		-0.413	-0.552
	(0.059)	(0.255)	(0.071)		(0.263)	(0.330)
hep				-35.255
				(26.260)
ur	0.252	0.636	0.100	0.294	2.822**	2.742
	(0.301)	(0.760)	(0.308)	(0.842)	(1.094)	(1.694)
techi	2.097***	4.302**	2.225***	5.425***	2.040	4.526*
	(0.475)	(1.762)	(0.493)	(1.483)	(1.358)	(2.284)
lnco2	-0.029*	-0.028	-0.036*		-0.069	-0.095
	(0.015)	(0.049)	(0.018)		(0.048)	(0.058)
lnso2				0.086*
				(0.042)
capi	0.500***	0.997**	0.497***	1.135***	0.844**	0.730*
	(0.088)	(0.354)	(0.111)	(0.348)	(0.339)	(0.374)
od					-0.521**
					(0.176)
ep					-0.516
					(1.455)
fd					-0.323
					(1.347)
lnsize					0.887*
					(0.449)
Constant	2.309***	3.076**	2.380***	3.524***	-3.983	3.066***
	(0.213)	(1.060)	(0.219)	(0.454)	(3.251)	(0.887)
Observations	420	420	390	420	420	300
R-squared	0.960	0.868	0.960	0.871	0.880	0.893
Prov FE	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
r2_a	0.960	0.851	0.955	0.853	0.863	0.874

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

5. Conclusions

This study constructs a green finance development index and employs fixed effects models, mediation models, and moderated mediation models to conduct an empirical analysis of panel data from 30 provinces in China from 2008 to 2021. The analysis investigates the mechanisms through which green finance influences regional technological innovation, with a particular focus on the mediating role of R&D investment and the moderating role of regional innovation and entrepreneurship capabilities. The results indicate that green finance has a significant positive impact on regional technological innovation, with part of this effect being realized through increased R&D intensity. Moreover, regional innovation and entrepreneurship capabilities significantly moderate this relationship. Specifically, higher innovation and entrepreneurship capabilities enhance the positive effect of green finance on R&D investment, though the marginal effect of R&D on innovation may diminish in regions with stronger capabilities.

This research uncovers the multi-level mechanisms by which green finance affects regional technological innovation, highlighting the critical mediating role of R&D investment and the complex interactions of moderating variables in different contexts. These findings offer new insights and empirical evidence for understanding how green finance drives regional innovation. Based on these results, the study proposes several policy recommendations. First, the government should establish dedicated green finance innovation funds to support R&D and innovation projects in environmental technology companies, particularly those that significantly enhance regional innovation capacity. This will alleviate financial pressures on companies engaged in green technology development and foster technological breakthroughs. Second, the government should further improve green finance-related policies by offering more favorable tax and loan incentives, encouraging financial institutions and enterprises to actively participate in green finance initiatives, thus promoting regional technological innovation. Third, local governments and businesses should increase their support for innovation and entrepreneurship, especially in regions with weaker capabilities, by providing policy guidance and resource allocation to strengthen overall regional innovation capacity and amplify the effect of green finance on R&D investment. Lastly, collaboration between green finance institutions and scientific research institutes, universities, and other innovation stakeholders should be promoted to establish synergistic innovation mechanisms. Such cooperation will optimize the allocation of green finance resources and accelerate the industrialization and application of research outcomes.

These policy recommendations will not only contribute to the deepening of green finance but will also effectively promote regional technological innovation, facilitating the green transformation of the economy and sustainable development. Their implementation will ensure that the potential of green finance to drive innovation is fully realized, providing robust support for green transition and high-quality development.

While this study provides new perspectives on the relationship between green finance and regional technological innovation, there remains room for further exploration. The current research focuses on provincial-level data, and future studies could extend to the firm level, offering more detailed insights into how green finance shapes the trajectory of firm-level technological innovation. Additionally, future research could examine other potential mediators, such as human capital and intellectual property protection. Given the ongoing development of green finance policies, it is also essential to consider the long-term effects. Therefore, future research should adopt a more micro-level perspective, explore a broader range of mediating factors, and extend the time frame to build a more comprehensive and in-depth understanding.