Deconstruction of green product innovation drivers for regional export-oriented industrial enterprises in China: the product space perspective

The paper addresses how export-oriented industrial enterprises in China can achieve green growth through green product innovation (GPI) amid high uncertainty and compliance costs from tightening environmental regulations. It integrates Sustainable Development Goals with industry and trade by positing GPI as a route to competitive advantage in global green markets. The research questions are: (1) whether endogenous improvements in regional green production capacity directly drive enterprises’ GPI, and (2) how government environmental and financial policies should coordinate with firms’ development and innovation of green products. The study motivates a product space (PS) perspective—linking relatedness and green complexity—to explain why regions with higher accumulated green productive knowledge can more feasibly diversify into new green products and gain international comparative advantage, thus operationalizing GPI at the regional enterprise level.

The paper situates firm GPI antecedents within two strands: (a) internal organizational drivers from the natural resource-based view (NRBV), emphasizing resources, capabilities, and especially absorptive capacity; and (b) external institutional drivers via environmental regulation and stakeholder pressures, including green finance. Prior work documents mixed effects of environmental regulation on innovation and limitations of traditional finance in supporting green innovation, while highlighting needs for fiscal and policy instruments (e.g., tax rebates, equity investment). Gaps identified include insufficient articulation of regional characteristics in firms’ GPI and a disconnect between policy design and implementation for green growth. The authors propose that PS theory’s principles of relatedness and complexity can reconcile internal and external drivers: relatedness captures knowledge diffusion and absorptive capacity effects; green economic complexity captures accumulated regional green productive knowledge. Hypotheses: H1, increases in green production capacity directly drive GPI; H2a/H2b, this effect operates indirectly through green technology innovation and human capital; H3a/H3b, environmental regulation and green credit positively moderate the capacity–GPI link.

Data: Panel data for 30 Chinese provinces (excluding Tibet, Hong Kong, Macao, Taiwan) from 2004–2018 on regional export-oriented industrial enterprises (REIEs). Micro export data matched from the EPS microenterprise export database by aligning China Customs Database and China Industrial Enterprises Database; aggregated to provincial level. Green products identified using OECD/Sauvage (2014) Comprehensive List of Environmental Goods (CLEG). Additional data from CEPII-BACI (global trade), CSMAR, national/provincial yearbooks, China Industrial Statistical Yearbook, and Statistical Yearbook on Science and Technology Activities of Industrial Enterprises. Variables: Explained variable (GPI): annual total green product exports for which REIEs in a province newly acquire regional comparative advantage (RCA≥1), with robustness using export quantity counts. Core explanatory variables (green production capacity): (1) Green complexity index (GCI), measuring regional green economic complexity as the specialization-weighted sum of normalized product complexity indices for green products; (2) green product-related density (Gdensity) measuring relatedness between a province’s existing exports and green products, used as an instrument. Driving channels: Green technology proxied by the number of granted green patents (IPC Green List); Human capital measured by the present value of lifetime income (CHLR, Jorgenson–Fraumeni method). Moderators: Environmental regulation index (composite of standardized industrial wastewater, SO2, soot with weights); Green credit proxied by the ratio of interest expenditures of all industries excluding six high-energy-consuming industries to total industrial interest expenditures. Controls: regional enterprise size (output per enterprise above scale), innovation climate (share of firms conducting R&D), logistics level (cargo turnover), FDI (inward/out-of-province), digitization (entropy-weighted composite of infrastructure, usage, and industry development), local government expenditure (fiscal expenditure/GDP), trade openness (total trade/GDP). Models: (1) Direct drive: two-way fixed effects panel regression of log GPI on GCI and controls; endogeneity addressed via IV-2SLS with green product-related density as instrument for GCI. (2) Indirect push (channels): three-stage least squares (3SLS) system where GCI predicts channel variables (green patents or human capital), which then predict GPI, extending the IV setup. (3) Regulatory pull: interaction models including GCI×environmental regulation and GCI×green credit. (4) Heterogeneity: panel threshold regressions (Hansen bootstrap, 300 replications) allowing slopes of GCI to vary across thresholds in GCI, log FDI, log digitization, local government expenditure, and trade openness. Robustness: alternative explained variable; additional controls (technology market transactions, industrial agglomeration, regional per capita GDP); winsorization (1% and 5%) to handle outliers; alternative IV addressing potential reverse causality by using the mean GCI of the same economic-geographic region as instrument; weak-IV and overidentification diagnostics reported.

- Direct effect (H1): In fixed effects, GCI positively predicts GPI (coef ≈ 1.510, p<0.01). IV-2SLS strengthens the effect (coef ≈ 2.650, p<0.01). Instrument validity: first-stage IV (log Gdensity) strongly predicts GCI (coef ≈ 0.359, F≈92.57; underidentification LM≈11.70, p<0.01). Endogeneity of GCI rejected at 1% level, supporting causal interpretation. - Indirect channels (H2): Green technology channel: GCI → green patents (coef ≈ 0.322, p<0.01); green patents → GPI (coef ≈ 7.976, p<0.01). Human capital channel: GCI → human capital (coef ≈ 0.091, p<0.01); human capital → GPI (coef ≈ 28.118, p<0.01). Both channels significantly mediate the capacity–GPI relationship. - Moderation (H3): Environmental regulation positively moderates the GCI→GPI effect (interaction coef ≈ 0.086, p<0.10). Green credit has a strong positive moderating effect (interaction coef ≈ 1.515, p<0.01). - Threshold heterogeneity: Inverted U-shaped promoting effects found for GCI, FDI, digitization, and trade openness—GCI’s marginal effect rises across the first two thresholds (to ≈2.335) then declines (≈1.031). For local government expenditure share, the effect shows diminishing marginal utility: slope decreases from ≈1.963 to ≈1.301 beyond the threshold. Logistics level shows no threshold effect. - Robustness: Results hold when replacing the explained variable (export quantity), adding extra controls (enterprise/industry/region), winsorizing outliers, and using alternative IVs to address reciprocal causation (first-stage F≈44.26; Hansen J p≈0.356).

The findings demonstrate that regional accumulation of green productive knowledge—captured by green economic complexity—directly and indirectly drives enterprises’ green product innovation, aligning with product space theory that relatedness and complexity shape diversification into new products. The positive moderating roles of environmental regulation and green credit show how institutional settings translate potential capacity into realized GPI, consistent with the Porter hypothesis’s innovation compensation effect and the role of green finance in lowering innovation costs and risks. The inverted U-shaped heterogeneity suggests that while greater capacity, openness, FDI, and digitalization initially enhance GPI, beyond certain thresholds coordination costs, saturation, or crowding may dampen marginal gains; similarly, fiscal expansion displays diminishing returns, indicating the need for targeted rather than uniformly expansive policies. These insights bridge internal capability-building with external policy design, offering a PS-grounded framework to operationalize regional strategies for green industrial upgrading and export competitiveness.

This study shows that improvements in regional green production capacity are a direct and significant driver of green product innovation among China’s regional export-oriented industrial enterprises. It identifies green technology advancement and human capital accumulation as key indirect channels and evidences that environmental regulation and green credit strengthen the capacity–GPI link. Threshold analyses reveal inverted U-shaped effects for capacity, FDI, digitization, and openness, and diminishing marginal returns to local government expenditure. Contributions include: (a) extending product space theory to subnational green complexity and linking it to enterprise-level GPI outcomes; (b) reconciling internal capability accumulation with external institutional drivers; and (c) providing a quantitative, policy-relevant framework for regional green industrial strategy. Future research could: incorporate production data (not just exports) to construct PS; refine enterprise-level identification and variables as microdata access improves; and extend analyses beyond 2018 to assess evolving digital-green synergies and policy reforms.

- Data limitations: Reliance on export-based PS construction due to lack of comprehensive production data; microenterprise identification constrained by EPS linkage using enterprise or customs codes rather than unique firm identifiers. - Temporal coverage: China Customs ceased providing firm-level export product details after 2016, requiring aggregation that may affect enterprise-level inference for later years. - Measurement choices: Proxy measures for green credit, environmental regulation, and human capital may not capture all dimensions; potential unobserved regional shocks despite extensive controls. - Generalizability: Results pertain to Chinese provincial REIEs and green product lists per CLEG (HS1996 harmonization), which may limit direct transferability to other contexts or classification schemes. - Methodological path dependency: PS measures depend on trade data; future work should explore PS methods less reliant on exports and incorporate updated datasets.