Digital–real integration and higher-quality productive capacity reinforce one another across Chinese provinces, but gains are locally concentrated: progress in one province tends to crowd out development in neighbors, producing negative spatial spillovers.

Main Finding

The study finds a significant bidirectional positive relationship between digital–real integration and the development of New Quality Productive Forces across 30 Chinese provinces (2011–2022). The promoting effect of digital–real integration on New Quality Productive Forces is slightly larger than the reverse effect. However, interactions between provinces exhibit significantly negative spatial spillovers: growth in one province tends to suppress the same variables in neighboring provinces. The results come from a spatial simultaneous-equations estimation using Generalized Spatial Three-Stage Least Squares (GS3SLS).

Key Points

• Bidirectional complementarity: Digital–real integration and New Quality Productive Forces reinforce each other—each helps raise the level of the other.
• Asymmetry in strength: The boosting effect runs slightly stronger from digital–real integration → New Quality Productive Forces than vice versa.
• Negative spatial spillovers: Instead of mutually beneficial regional diffusion, interregional effects are negative—development in one province can crowd out or reduce development in neighboring provinces.
• Policy implications emphasized by the authors: enhance interactive development, foster interregional cooperation, and optimize allocation of resources to reduce adverse spillovers.
• Conceptual mechanisms (implied): positive local complementarities (technology adoption raising productivity), but regional competition for scarce factors (talent, investment, firms) and uneven resource allocation likely cause negative spillovers.

Data & Methods

• Data: Panel of 30 Chinese provinces, annual observations 2011–2022.
• Dependent constructs: composite indices for (a) digital–real integration and (b) New Quality Productive Forces (the paper constructs/uses indexes capturing the depth of integration and the quality/tech-driven productive capacity).
• Econometric approach: spatial simultaneous-equations model estimated by GS3SLS.
  • GS3SLS addresses endogeneity and simultaneity between the two endogenous variables while allowing for spatially lagged dependent variables (spatial spillovers).
  • A spatial weight matrix is used to capture interprovincial spatial relationships (contiguity/distance-based weighting is typical).
• Identification features: simultaneous-equation framework identifies two-way causal links; GS3SLS corrects for bias from endogenous spatial lags and correlated errors.
• (Typical controls) The framework would account for standard province-level covariates (e.g., human capital, infrastructure, R&D intensity, industrial structure), and tests for spatial dependence and robustness are central to the approach.

Implications for AI Economics

• AI as a driver within digital–real integration: AI adoption and embedding into production processes are central components of digital–real integration and hence important levers for raising New Quality Productive Forces (higher productivity, quality and tech content).
• Expect bidirectional causality in AI studies: AI adoption both raises—and is facilitated by—the emergence of higher-quality productive capacities. Empirical work should allow for simultaneous effects rather than uni-directional assumptions.
• Account for spatial structure: AI benefits and harms can have spatially heterogeneous impacts. Negative spatial spillovers suggest regional competition (for AI talent, investment, leading firms) can limit positive diffusion; policies should seek mechanisms that transform potential negative spillovers into positive spillovers (knowledge sharing, joint infrastructure).
• Policy design for inclusive AI-driven growth:
  • Promote interregional coordination: cluster-level cooperation, shared research platforms, open data/infrastructure to spread AI benefits.
  • Invest in lagging regions’ absorptive capacity: digital infrastructure, human capital, and institutions to prevent siphoning of benefits to core regions.
  • Use incentives to internalize spillovers: fiscal transfers, joint R&D grants, mobility programs for AI talent.
• Methodological guidance for researchers:
  • Use spatial simultaneity models (e.g., GS3SLS or other spatial IV approaches) when analyzing AI adoption and regional economic outcomes to handle two-way causality and spillovers.
  • Construct multidimensional indices of digital–real integration and “quality” productive forces to capture complementarities beyond single indicators.
  • Complement provincial-level studies with firm- or city-level microdata to unpack mechanisms (competition for talent, investment diversion, technology diffusion).
• Broader takeaway: Scaling AI-driven productive transformation at the national level requires policies that both boost local complementarities and actively manage interregional interactions so that AI-led gains are aggregated rather than zero-sum across places.