西部论坛

科技创新是引领发展的第一动力，数据是数字经济时代的核心生产要素，“数据要素×科技创新”将驱动新质生产力持续跃升。采用我国30个省份2011—2023年的数据，以数据要素发展水平和科技创新发展水平的耦合协调度衡量“数据要素×科技创新”水平，利用图神经网络(GNN)提取融合特征，进而构建GNN-SDM模型检验，分析发现：“数据要素×科技创新”不仅显著提升了本地新质生产力发展水平，还带动了相邻地区的新质生产力发展，且间接效应大于直接效应，表明“数据要素×科技创新”的区域联动对新质生产力发展至关重要；GNN特征揭示出两类空间结构——“抑制本地、带动周边”的反向激励型空间扩散结构(GNN1)和“增强本地、抑制周边”的“中心-外围”型空间极化结构(GNN2),GNN1值较高地区应优化内部结构与资源配置，GNN2值较高地区则应强化区域联动；“数据要素×科技创新”对东部地区、“数据要素×科技创新”水平较高地区、新质生产力发展水平较高地区具有更强的新质生产力驱动作用，发展新质生产力应因地制宜。

In the digital economy era, “data elements ×” is emerging as the core engine driving the development of new quality productive forces by unleashing the multiplier effect of data. It not only effectively empowers industrial transformation and upgrading but also serves as a key pillar for achieving high-quality economic development. In this process, data elements and technological innovation are deeply integrated, jointly constituting the dual-wheel drive for the growth of new quality productive forces. However, empirical research on the intrinsic mechanisms linking data elements, technological innovation, and new quality productive forces remains scarce, particularly in identifying the pathways for their coordinated development and spatial diffusion.This study employs provincial-level panel data from 2011 to 2023 to measure the coupling synergistic effects between data elements and technological innovation. Drawing on the methodology of Zhu et al.(2022), it utilizes graph neural networks(GNN) to extract fusion features of both elements. These features, along with their coupling synergistic effects, are embedded into a spatial Durbin model(SDM) to establish the GNN-SDM model for empirical analysis. The findings reveal that the coupling synergistic effect not only significantly promotes the development of new quality productive forces within the region but also drives improvements in surrounding areas through spatial spillover mechanisms. Heterogeneity analysis confirms significant regional variations in the synergistic driving effect of data elements and technological innovation, with this effect being particularly pronounced in eastern regions, areas with high synergy levels, and regions leading in new quality productive forces. This highlights the critical role of regional foundational conditions in constraining the realization of synergy dividends. Furthermore, intergroup differences in GNN feature variables validate the diversity of regional spatial interaction mechanisms. Moreover, GNN reveals two spatial diffusion mechanisms: GNN1 exhibits an inverse incentive structure where structural imbalances within a region inhibit its own development but can stimulate surrounding areas through spatial spillovers; GNN2 reflects a central-peripheral innovation polarization structure where intra-regional innovation linkage significantly boosts local new quality productive forces while generating a “siphon effect” on neighboring areas.Compared to previous research, this paper makes two main contributions. First, from a synergy perspective, it proposes that data elements and technological innovation have a coupled-driving, spatially diffusing effect on new quality productive forces, thereby expanding the research perspective on new quality productive forces and providing empirical support and policy references for inter-provincial collaborative governance. Second, methodologically, by integrating GNN with SDM, it achieves precise characterization of complex spatial correlation structures and heterogeneous innovation diffusion pathways, enhancing the ability to identify the spatial spillover effects of synergistic development on new quality productive forces.This study introduces graph neural networks as the core feature extraction tool. Leveraging their powerful structural learning capability, the model adaptively captures complex nonlinear interactions and spatial dependencies among regions, effectively extracting highly representative spatial embedding features. Building upon traditional spatial econometric models, this approach further enhances the fitting accuracy and generalization ability of the model for actual economic and geographical phenomena and enhances the interpretability of the model for spatial interaction mechanisms. This method provides a new analytical perspective and methodological support for deeply revealing the internal mechanism of regional coordinated development and identifying the transmission paths of cross-regional linkage effects.