The spatial spillover effect of financial growth on high-quality development: Evidence from Yellow River Basin in China

The study addresses how financial growth influences high-quality development (HD) within China’s Yellow River Basin (YRB), emphasizing spatial dependence, nonlinearity, and transmission mechanisms. Against China’s policy push from high-speed to high-quality growth—particularly the 2021 Outline of Ecological Protection and High-Quality Development Plan (OEHP) for the YRB—and substantial recent financial inflows, the paper highlights gaps in empirically evaluating HD and identifying how finance affects it spatially. The authors construct a comprehensive HD evaluation framework aligned with China’s “demand–development” contradiction theory, and propose three hypotheses: H1, financial growth has an N-shaped impact on HD; H2, green innovation (GI) mediates the effect of financial growth on HD; and H3, higher economic development strengthens the impact of financial growth on HD. The purpose is to quantify HD, test spatial autocorrelation, estimate spatial spillovers using spatial econometric models, and assess mediation and heterogeneity across economic strata.

Three strands are reviewed. (1) Definitions of HD: Prior work views HD as a systemic, multidimensional, coordinated mode balancing economic, social, ecological, and cultural dimensions; however, a unified definition is lacking. This paper defines HD as comprehensive development meeting long-term human needs across stable economic growth, social welfare, cultural support, and ecological protection. (2) Evaluation of HD: Calculation methods include efficiency models (e.g., super-efficient SBM) and composite index methods (entropy, entropy-TOPSIS, PCA, grey methods). Indicator selection ranges from single indicators to multi-dimensional systems; in the YRB, studies often include resources, ecology, economy, and society. (3) Financial growth’s impact on HD in the YRB: Finance can bolster innovation, reduce poverty and inequality, and improve environmental outcomes, but some argue it can increase energy-intensive consumption and pollution. The literature is divided on net effects and lacks spatially explicit, systematic empirical testing in the YRB. The paper aims to fill these gaps by building a demand–development-based HD index, and by empirically testing nonlinear, spatial, and mediating effects of financial growth.

Design: Four steps: (1) Construct an HD evaluation system for YRB cities; (2) Test spatial autocorrelation of HD; (3) Specify spatial econometric models; (4) Conduct empirical analyses including nonlinearity, robustness, mediation (GI), and heterogeneity. Data and sample: Panel of 99 prefecture-level cities in the YRB across nine provinces (excluding 14 autonomous prefectures and Haidong City for data scarcity), 2006–2019. Data from China Statistical Yearbook and China Urban Statistical Yearbook. HD evaluation: Entropy-weighted TOPSIS using 12 indicators across four dimensions—Economy (A1: GDP per capita; A2: service VA share; A3: labor productivity; A4: R&D/GDP), Society (A5: employee salary; A6: medical beds per capita; A7: unemployment rate), Ecology (A8: public green space per capita; A9: harmless waste treatment rate), Culture (A10: centralized sewage treatment rate; A11: public library collections per capita; A12: education expenditure share). Weights derived via entropy; HD index computed by TOPSIS distance to positive/negative ideal solutions. Key variables: HQ (HD index, dependent). Core explanatory: financial growth (FG), defined as (bank loans + deposits)/GDP. Controls: human capital (HC = higher education students/population), informatization (INT = internet user scale), foreign direct investment (FDI, RMB converted), and government control (GC = fiscal expenditure/GDP). GI is used in mediation analysis as an intermediary variable. Spatial methods: Spatial weight matrix (W) is binary contiguity (adjacency). Global Moran’s I assesses spatial autocorrelation of HD. Spatial Durbin Model (SDM) is selected based on LM, robust LM, Wald, LR, and Hausman tests. Modeling: Nonlinear SDM specifications include FG, FG^2 (Model 1), and FG, FG^2, FG^3 (Model 2), with spatial lags of dependent and independent variables and controls. Effects are decomposed into direct, indirect (spillover), and total effects. Heterogeneity analysis splits cities by average real GDP per capita into high-economy (36) and low-economy (63) subsamples (Models 3 and 4). Robustness checks include instrumental variables (IV-GMM using lagged loans component of FG as instrument), OLS substitution, and sample cropping excluding five geographically isolated cities to address spatial discontinuity. Estimation software: ArcGIS for mapping; STATA for Moran’s I and regressions.

- Spatial dependence: HD exhibits significant positive spatial autocorrelation across 2006–2019; Moran’s I ranges from 0.257 to 0.469 with Z-values 4.204–6.944, all significant at 1%. - Nonlinear local effects: Financial growth shows an N-shaped relationship with HD. In SDM with cubic term (Model 2), FG > 0, FG^2 < 0, FG^3 > 0, all statistically significant (Table 6), confirming H1. Controls: HC and FDI positively and significantly affect HD; INT and GC negatively affect HD. - Spatial spillovers: Decomposition indicates an inverted U-shaped spillover of FG to neighboring cities’ HD—FG and FG^2 significant in indirect effects in Models 1–2, while FG^3 is not (Table 8). Thus, local N-shape but spatial inverted U-shape. - Magnitudes and spatial lag: The spatial lag of HD (W*HQ) is positive and highly significant (e.g., ~0.528 and ~0.476 in Models 1–2), underscoring strong spatial feedback. - Mediation via green innovation (GI): GI is significantly affected by FG with an inverted U-shape (FG positive, FG^2 negative). Given GI’s documented multifaceted impact on HD, this supports GI as a partial mediator (H2) (Table 10). - Heterogeneity by economic level: The N-shaped local effect of FG on HD persists in both high- and low-economy subsamples (Models 3–4). Effect sizes are stronger in high-economy areas (e.g., larger absolute coefficients). In high-economy areas, spillovers exhibit N-shaped characteristics (Table 13). In low-economy areas, the nonlinear spillover pattern is largely absent (Table 14). These results validate H3: higher economic development strengthens both the local impact and the transmission of nonlinear effects. - Robustness: IV-GMM (lagged loans as instrument) addresses endogeneity; Cragg–Donald Wald F ≈ 641 indicates strong instruments. GMM, OLS substitution, and sample cropping all confirm the N-shaped nexus. - Descriptive highlights: Financial growth increased over time (e.g., maximum value from 5.339 in 2016 to 11.173 in 2019); HD values range ~0.04–0.80 with spatial clustering of high and low values.

The findings confirm that HD in the YRB is spatially clustered, justifying spatial econometric modeling. The stable N-shaped local relationship indicates that early-stage financial expansion promotes economic and social aspects of HD, mid-stage expansion can hinder HD via environmental degradation and structural frictions, and further maturation in finance supports greener industry, cultural and social improvements, and efficiency, re-boosting HD. Spatially, financial growth initially disseminates benefits to nearby cities through factor flows, technology diffusion, and commuting/remittance patterns, but later stages can concentrate talent and resources, reversing benefits and inhibiting neighbors—hence an inverted U-shaped spillover on HD. GI serves as a transmission channel: financial growth stimulates GI up to a threshold, which then affects HD’s economic, social, and environmental pillars, partially explaining the overall N-shaped path. Economic development level conditions these dynamics: high-economy regions amplify finance’s local effects and enable nonlinear spillovers; in low-economy settings, weak demand, smaller industrial scale, and lower agglomeration reduce finance’s spatial transmission. These results address the research questions and support H1–H3, offering nuance on the spatial allocation and structure of financial development for promoting HD.

This paper develops a four-dimensional (economy, society, ecology, culture) HD index via entropy-weighted TOPSIS for 99 YRB cities (2006–2019) and applies spatial econometrics to assess how financial growth affects HD locally and across space. Main contributions: (1) documenting strong positive spatial autocorrelation in HD; (2) identifying a robust N-shaped local nexus between financial growth and HD; (3) revealing an inverted U-shaped spatial spillover of financial growth on neighboring HD; (4) establishing GI as a partial mediator; and (5) showing that higher economic development strengthens both the magnitude and spatial transmission of finance’s effects. Policy implications include coordinating inter-city financial planning, leveraging high-value hubs to catalyze low-value clusters, supporting demonstration cities in low-value agglomerations, and aligning financial development with structural efficiency and environmental goals, particularly considering local economic conditions. Future research should unpack the exact mechanisms linking GI to aggregate HD and to specific HD sub-dimensions, and examine the distinct roles of green finance and digital finance in shaping HD dynamics.

- Data coverage excludes 14 autonomous prefectures and Haidong City due to missing data, potentially affecting representativeness for the full YRB. - Financial growth is measured as (loans + deposits)/GDP, capturing scale but not financial structure/efficiency or green/digital finance nuances. - Spatial weights rely on simple contiguity; alternative distance- or economy-based matrices may yield additional insights. - Although IV-GMM addresses endogeneity, residual concerns may remain given complex two-way linkages between finance and HD. - The mediation mechanism via green innovation is only partially identified; the study acknowledges that the precise processes and links to HD sub-indicators warrant further exploration. - Robustness checks include excluding five geographically isolated cities; results may be sensitive to other sample or period choices.