Role of green finance in regional heterogeneous green innovation: Evidence from China

China’s development strategy has shifted from rapid growth to high-quality, low-carbon development. Despite progress, reliance on fossil fuels remains high, with CO2 emissions around 12.1 billion tons in 2022. Green innovation (GI) is key to reducing emissions and improving energy efficiency but faces high costs, long cycles, and financing constraints. China has built a comprehensive green finance (GF) system (green credit, bonds, insurance, trust) to address these constraints. Existing studies suggest GF promotes GI and shows spatial spillovers, yet the mechanisms across heterogeneous GI types and regional differences are insufficiently understood. This study investigates how GF affects different categories of GI, whether environmental supervision intensity (ESI) generates threshold effects in the GF–GI relationship, and whether spatial spillovers exist. It aims to inform region-specific strategies for China’s green transformation.

Prior work shows GI supports low-carbon development but is complex, costly, and sensitive to institutional conditions. Barriers include high development costs, commercial uncertainty, and limited willingness to pay. GF can mitigate financing constraints (especially for SMEs), offset environmental compliance costs, and incentivize higher-quality GI via market mechanisms. However, heterogeneity in GI outcomes by type is observed; some policies improve GI diversity but not quality, and technology-intensive versus structural GI types may respond differently to GF. Regional disparities in resources, supervision, and industry structures mean policy effectiveness varies across China, with eastern coastal regions leading renewable energy innovation. Literature on environmental supervision presents both Porter-hypothesis-based positive effects and compliance-cost-based negative effects; outcomes depend on whether innovation benefits offset higher regulatory costs, potentially yielding nonlinear (threshold) dynamics. Spatial literature indicates GI exhibits spillovers through knowledge, networks, and industrial transfers, but studies often focus on innovation types rather than specific green technology paradigms. Hypotheses: (H1) GF’s influence differs across heterogeneous GI; (H2) ESI has a threshold effect on the GF–GI relationship; (H3) The impact of GF on heterogeneous GI exhibits spatial effects.

- Empirical strategy: The study uses provincial panel data for 30 Chinese provinces (2010–2019) and estimates (1) a baseline negative binomial count model to relate GF to heterogeneous GI, (2) a panel threshold model (Hansen, 1999) to test for ESI thresholds in the GF–GI relationship, and (3) a spatial error model (SEM) to assess spatial spillovers after confirming spatial correlation (Moran’s I) and selecting SEM via LM tests. - Dependent variable (GI): Number of authorized green invention patents (lagged one year) by region, classified per China’s 2022 Green Low-Carbon Technology Patent Classification into five categories: GI_1 fossil energy carbon reduction, GI_2 energy saving and recovery, GI_3 clean energy, GI_4 energy storage, GI_5 GHG capture, utilization, and storage. Invention patents are used to reflect higher-level, substantive innovation. - Independent variable (GF): A composite green finance index constructed from 10 primary and 26 secondary indicators covering policy promotion, market incentives/constraints, and government investment across provincial–municipal–county levels. Indicators are standardized and aggregated using a spatiotemporal entropy-weighting method to form the GF index. - Threshold variable (ESI): Intensity of environmental supervision proxied by the number of environment-related administrative penalties in each region, capturing regulatory pressure. - Controls: Per capita GDP (PGDP), government support (GOV: fiscal expenditure/GDP), foreign direct investment (FDI), financial development (FIN: deposits and loans/GDP), population density (PPL), and industrial structure (STR: share of secondary and tertiary industries in GDP). - Spatial analysis: Global Moran’s I on GI and its categories using a geographical proximity matrix to test spatial correlation; LM and robust LM tests to select SEM over SLM. The SEM models include provincial and year fixed effects. - Data sources: China Statistical Yearbook, China Energy Statistical Yearbook, China Environmental Statistical Yearbook, China Urban Statistical Yearbook, incoPat patent database, and Peking University legal information database. - Robustness: Alternative lags (0, 2, 3 years), alternative count/OLS models (Poisson; OLS with log-transformed GI), and alternative ESI measures (inverse of industrial wastewater, SO2, and dust emissions).

- Baseline effects (negative binomial): GF positively impacts total GI; coefficient for GF on GI is 0.119 (p<0.01). By category, GF significantly promotes fossil energy carbon reduction (GI_1), energy saving and recovery (GI_2), and clean energy (GI_3); effects on energy storage (GI_4) and GHG capture/utilization/storage (GI_5) are not significant in the baseline. - Threshold effects (ESI): Single-threshold tests confirm ESI thresholds for most GI categories (Table 3). Estimated thresholds (ESI): GI_1=6.2519; GI_2=6.4677; GI_3=8.1764; GI_4=6.2519; GI_5=6.6946 (significant at least at 10% for all except GI_5 in the narrative, with table showing significance). Threshold regressions show when ESI is below the threshold, GF significantly affects only GI_2 (energy saving and recovery). When ESI exceeds the threshold, GF significantly and more strongly promotes GI across categories. - Spatial correlation: Global Moran’s I for GI and categories is generally positive and significant across years, indicating spatial agglomeration. - Spatial spillovers (SEM): LM tests favor SEM over SLM. SEM estimates show GF has strong positive effects on GI and all categories: GF coefficients range from 0.293 to 0.506 (all p<0.01). The spatial error term λ is negative and significant for total GI (λ=−0.165, p<0.01) and most categories, indicating a notable negative spatial spillover (siphoning) effect—regions near high-GI areas tend to have lower GI due to resource and talent flows and industry relocation. - Robustness: Results hold under alternative lags, models (Poisson, OLS with log GI), and alternative ESI measures.

The study confirms that green finance increases regional green innovation but with heterogeneous effects across technology categories. GF is most effective for technologies closer to production and with quicker returns—fossil energy carbon reduction, energy saving and recovery, and clean energy—aligning with policy definitions and investment incentives of GF. Environmental supervision intensity conditions the efficacy of GF: below threshold levels, GF has limited impact (notably only on energy-saving and recovery), while above threshold, GF significantly spurs broader GI. Spatial analysis reveals GI exhibits positive spatial correlation yet generates negative spatial spillovers in errors, consistent with siphoning of innovation resources, talent, and emission-intensive industry relocation due to stricter local regulation. These findings address the research questions by demonstrating heterogeneity, identifying an ESI threshold mechanism, and documenting spatial externalities, underscoring the need for regionally tailored GF and coordinated inter-regional policies.

The paper contributes by: (1) distinguishing heterogeneous GI via a national green technology classification and showing GF’s strongest impacts on fossil energy carbon reduction, energy saving and recovery, and clean energy; (2) identifying an environmental supervision intensity threshold beyond which GF’s promotion of GI becomes pronounced; and (3) evidencing negative spatial spillovers in GI, indicating a siphoning effect. Policy implications include strengthening local green financial systems; prioritizing GF support for technologies with clearer and faster returns (fossil carbon reduction, energy saving and recovery, clean energy) while not neglecting longer-cycle technologies; tailoring GF deployment to regional ESI, as significant gains require adequate regulatory intensity; and coordinating across regions to mitigate negative spillovers and ensure broader sustainability of green transformation.