Mineral resource dependence and green transformation: the impact and challenges of green finance in Chinese cities

Multicollinearity test

To address potential multicollinearity and its impact on regression accuracy, we present our test results in Table 3. The average VIF (Variance Inflation Factor) is 2.46, well below the threshold of 5, indicating no significant multicollinearity among our chosen variables.

Impact of MRD on GEE

Firstly, we empirically tested Eq. (1) using the IV method. To ensure the robustness of the estimation results, we gradually added control variables. As shown in the first column of Table 4, in the absence of any control variables, the coefficient of MRD is significantly negative, suggesting that the level of MRK is closely related to the reduction in urban GEE. This indicates that, overall, the GEE of cities at the prefectural level in China has not received the blessing of resources. Despite the introduction of control variables, this inverse relationship remains significant, as evident in columns (2) through (6). Particularly in column (6), the regression coefficient of MRD is −0.014, implying that for every 1% increase in MRD, the city’s GEE decreases by 0.014%, thereby preliminarily validating our H1 hypothesis. Additionally, the coefficient of MRD fluctuates between −0.014 and −0.021, demonstrating the robustness of the estimation results. This finding is consistent with the mineral resource curse proposed in this paper, where high dependence on resources hinders green economic development (Sachs and Warner 1995; Xu 2022).

Table 4 shows that the Kleibergen–Paap rk LM statistics significantly reject the under-identification null hypothesis, indicating a strong correlation between the instrumental and endogenous variables. Additionally, first stage regression F statistics exceeding 10 further reject the weak IV hypothesis. Therefore, the IV in this study is valid. The discussion of the validity of the IV in the following sections follows the same logic as presented here.

Validation of the green finance pathway

To explore the role of green finance in the influence of MRD on GEE, we selected green investment (GI) and green credit (GC) as mediating variables. In validating H2, we first used green investment (GI) as a mediating variable and performed regression analysis on Eqs. (2) and (3). As shown in column (2) of Table 5, MRD has a significant negative crowding-out effect on GI. In column (3), the negative effect of MRD continues, while GI has a significant positive impact on GEE. This suggests that an increase in MRD inhibits the improvement of GEE by crowding out green investments, thereby validating H2. According to H3, we incorporated green credit (GC) as a mediating variable. Results in columns (4) and (5) of Table 5 indicate that MRD has a significant negative effect on GC, and GC positively affects GEE. This implies that an increase in MRD negatively impacts GEE by inhibiting green credit, thus confirming H3. The beneficial influence of green credit and investment on GEE aligns with Muganyi et al. (2021).

In summary, the positive effects of green investment and green credit on urban green development are hindered by the crowding-out effect under mineral resource dependence. These results are consistent with the theoretical frameworks proposed Hu et al. (2021b) and Zhang et al. (2021), which suggest that green finance can stimulate green technological innovation and support sustainable industrial transformation. This finding emphasizes the need for city governments to strengthen investments in clean energy, renewable energy, and environmental industries, and to promote green finance development to ensure coordinated economic growth and environmental protection.

Robustness tests

Replacement of dependent variable

In the baseline regression model, we gradually added control variables, initially confirming the robustness of the estimation results. To further verify this, we remeasured green economic efficiency (GEE_sc) using the Super-CCR method. The estimation results are shown in column (1) of Table 6. Results indicate that even after replacing the dependent variable, the negative impact of MRD on GEE remains significant, further supporting the validity of the baseline regression results.

Alternative measurement of explanatory variables

We employed the ratio of mining industry employment to overall employment (MRD-e) as a substitute measure for mineral resource dependence. Table 6, column (2), reveals that MRD-e significantly and negatively affects GEE.

Substitution of the IV

We substituted the original IV with the lagged one period MRD (L.MRD) as a new IV. Since L.MRD is highly correlated with MRD and is predetermined, it can avoid reverse causality. Table 6, column (3), illustrates that the findings remain consistent after this IV substitution, verifying the robustness of the empirical results.

IvTobit model estimation

Acknowledging the dependent variable’s constraints, we re-estimated the parameters using the IvTobit model, as outlined by Wang et al. (2019a), to correct for endogeneity. The outcomes, displayed in Table 6, column (4), show a significant negative effect of MRD, aligning with previous findings and confirming the robustness of our estimates.

Exclusion of special cities

Given the distinct nature of municipalities, we recalculated after omitting data from Beijing, Tianjin, Shanghai, and Chongqing. The results, seen in Table 7, column (1), indicate that MRD’s negative impact is significant at the 5% level, further solidifying the baseline regression’s robustness.

Shortening the sample interval

To address any influence from the choice of sample period, we reanalyzed using data exclusively from 2007 to 2016. The findings, presented in Table 7, column (2), demonstrate that the main explanatory variables’ coefficients align with the baseline regression, even over this reduced timeframe.

Trimming

Considering the potential interference of outliers, we re-estimated the data after 1% trimming. The outcomes, seen in Table 7, column (3), confirm that MRD’s negative influence on GEE persists significantly.

Generalized method of moments estimation

The results from the Generalized Method of Moments (GMM) estimation, which utilized lagged values of the endogenous variables, are reported in the final column of Table 7. The p-value of the Hansen test (0.128) that we conducted to verify the validity of the instrumental variables indicates that our set of instruments is valid. The p-value of the AR(2) test (0.703) indicates that there is no correlation between the lagged explained variable and the error term.

City heterogeneity analysis

To delve deeper into how MRD in urban areas affects GEE, we categorized the 262 prefecture-level cities into two groups: 100 resource-based and 162 non-resource-based cities, according to the “Sustainable Development Planning for Resource-Based Cities (2013–2020).” Analysis of Table 8 reveals that in column (1), MRD’s coefficient is significantly negative for resource-based cities, while in column (2), it is significantly positive for non-resource-based cities. This suggests that resource-based cities, due to their heavy reliance on resource industries, often develop a path-dependent approach, leading to the displacement of innovative factors and skilled talents, negatively impacting non-resource industries. Conversely, non-resource-based cities, with varied industrial sectors, are less dependent on a single natural resource. Their focus on extensive green investments and technological advancements boosts green economic efficiency.

Next, considering the potential heterogeneity due to the different stages of resource development in various resource-based cities, we further divided resource-based cities into growing, mature, declining, and regenerative types for analysis. As shown in columns (3)–(6) of Table 8, in growing, mature, and declining cities, natural resource dependence inhibits the improvement of GEE. Only in regenerative cities does MRD promote the enhancement of GEE. Currently, growing, mature, and declining cities are in the initial, stable, and exhaustion stages of resource development, respectively. The economic development of these cities has not yet escaped dependence on mineral resources. This over-reliance on resources is unfavorable for the development of non-resource industries and cannot promote the enhancement of GEE. However, the economic development of regenerative cities has largely decoupled from mineral resources.

Overall, the differences in policy formulation and investment decisions among different types of cities significantly impact their green development efficiency. Non-resource cities and regenerative cities, such as those in the eastern coastal regions (like the Yangtze River Delta and Pearl River Delta), usually have more developed economies and higher technological levels. They also place greater emphasis on environmental protection and sustainable development. These areas generally do not rely on mineral resources but on manufacturing, high-tech industries, and services, thus the negative impact of MRD on green development is relatively smaller. These regions are also more capable of adopting and promoting green technologies, achieving green growth. In contrast, growing, mature, and declining cities face different challenges at various stages of economic development. Mature and declining cities, like the old industrial bases in Northeast China, traditionally reliant on heavy industry and resource extraction, are currently facing the challenge of economic restructuring. Meanwhile, emerging cities and city clusters (such as the Chengdu-Chongqing economic zone) are in a rapid process of urbanization and industrialization, facing challenges of balancing environmental protection and sustainable resource use while pursuing economic growth. From the above, it can be seen that the impact of MRD on GEE exhibits considerable heterogeneity among different types of cities, necessitating that policymakers and investors employ more nuanced and context-specific strategies to foster green growth in accordance with the unique characteristics of each urban area.

Further analysis

Recognizing the significance of economic status, resource allocation, and policy context in assessing various city types, and mindful of the diversity across Chinese cities, it becomes evident that custom green development strategies are crucial. This understanding leads us to a vital transition towards a more detailed analysis. To this end, we utilize data from 252 cities between 2006 and 2017, applying a dynamic threshold model for regression. This methodology is essential to unravel the intricate effects of green finance amidst the interplay of MRD and GEE.

Table 9 presents the regression results based on the threshold values of green investment and green credit. In the regression with green investment and green credit thresholds, the relationship between MRD and GEE demonstrates significant dynamic threshold effects, with threshold values of 0.073 and 0.034, significant within the 95% confidence intervals [0.067, 0.079] and [0.030, 0.039], respectively. The P-values of the linearity test also reject the null hypothesis, further confirming the dynamic non-linear relationship between MRD and GEE. Additionally, the P-value for Kink is statistically significant at the 1% level, thus avoiding the issue of knots in linear relationships.

The results with green investment as the threshold value show that the MRD coefficient is significantly negative in both the lower and upper regime. This result indicates that even if green investment develops to a certain extent, it still cannot offset the inhibitory effect of MRD on GEE. In the initial stages, insufficient investment in green technology and sustainable development results in low-level green investments in resource-based cities, which are insufficient to drive the economy towards a more sustainable model. Even with increased green investment in later stages, resource-based cities continue to face green transformation challenges due to the inertia of economic structure and the entrenched nature of traditional industries. Furthermore, reliance on resources typically results in environmental degradation and ecological damage, necessitating sustained effort and investment for resolution. As a result, the beneficial influence of GI on Green transformation might not be instantly observable. In the lower portion of Table 9, results indicate that when GC is less than or equal to 0.034, MRD positively influences GEE. When GC exceeds 0.034, MRD hinders the improvement of GEE. These findings suggest that at low levels of green credit, the development and utilization of mineral resources initially drive economic growth and provide funds for infrastructure and development. However, with increasing levels of green credit, resource-based cities may encounter challenges such as resource depletion and environmental degradation, adversely affecting sustainable development. During this phase, MRD tends to lead to overinvestment in traditional resource extraction at the expense of green economic development.

Further discussion on the results

The further analysis aligns with the actual situation in China. For instance, some cities in the northeastern region, such as Anshan in Liaoning, Jilin City in Jilin Province, and Datong in Shanxi, traditionally depend on coal and heavy industries. These areas have attempted transformation through increased green investments but often face challenges such as insufficient capacity for technology absorption, unclear market demand, or lack of policy support, making it difficult for high-level green investments to achieve the expected impact. Resource-based city clusters in western regions, such as Xi’an in Shaanxi Province and Lanzhou in Gansu Province, despite being rich in resources, encounter significant obstacles in achieving green development. This challenge is compounded in cities heavily dependent on conventional energy and resource-intensive industries. In such cities, even increased green investments may not effectively stimulate green growth unless supported by suitable policy frameworks and market mechanisms.

In contrast, emerging city clusters like the Chengdu-Chongqing region, although undergoing rapid urbanization and industrialization, are also striving for greener and more sustainable development. However, these cities face difficulties in translating green investments into actual green growth. For example, cities like Suzhou in Jiangsu Province and Ningbo in Zhejiang Province, despite rapid economic development, are also confronted with industrial pollution and overuse of resources. In these cities, even with increased green credit investment, the negative impact of MRD on GEE persists due to limitations in industrial structure and technological capacity. Additionally, the marginal effects of green finance decrease beyond the threshold value, meaning that each additional amount of green investment may contribute less to promoting green growth, especially in areas heavily dependent on mineral resources.

It is noteworthy that the Chinese government has implemented a series of important policies in the green finance sector, such as the carbon trading market launched in 2021, the establishment of green finance pilot zones, and the update of green bond standards. Since most of these policies were implemented after 2017, their impacts are not yet reflected in our dataset. The current analysis may not fully capture the potential effects of recent policy dynamics on green economic growth. Therefore, the rapid changes in the green finance policy environment need to be considered when interpreting our study results.