Clusters of digital industries working with AI-like technologies boost green productivity by correcting innovation-resource misallocation, but returns fade beyond a threshold and can spill negatively onto neighbouring advanced regions; broader regional knowledge strengthens the gains.

Main Finding

Digital industry agglomeration (DIA) and digital–intelligent technology application (DIT) interact synergistically to improve green productivity (GP) by combining external regional innovation resources with firms’ internal innovation allocation. This synergy reduces innovation-resource misallocation and raises GP, but (i) it exhibits diminishing returns beyond a measurable synergy threshold, and (ii) its spatial spillovers differ by development stage—positive in less-developed regions and negative for neighbours of developed regions. Knowledge breadth and regional innovation intensity act as amplifiers once they exceed their own thresholds.

Key Points

• Synergy mechanism

  • DIA supplies external innovation resources and dense knowledge networks; DIT enables firms to orchestrate internal resources effectively.
  • The DIA–DIT interaction reduces misallocation of innovation resources and thereby elevates GP across three evaluated dimensions: labourers, means of labour, and objects of labour.
  • Broader regional knowledge (knowledge breadth) strengthens the positive DIA–DIT → GP pathway.

• Nonlinearity and thresholds

  • The positive marginal impact of DIA–DIT synergy on GP weakens after a specific synergy threshold (diminishing marginal returns).
  • Conversely, when regional knowledge breadth or innovation intensity cross their respective thresholds, the positive effect of DIA–DIT synergy on GP is amplified.

• Spatial heterogeneity

  • Spatial spillovers of DIA–DIT synergy vary by regional development stage:
    • Less-developed regions: synergy generates positive spillover effects on neighbouring regions’ GP.
    • Developed regions: synergy can exert negative spillovers on neighbours (possible crowding out or competitive displacement).

• Policy-relevant dynamics

  • Overcoming threshold constraints and managing spatial spillovers are crucial to achieving coordinated regional development and maximizing GP gains from digitalization and AI-driven technologies.

Data & Methods

• Outcome measurement

  • Green productivity (GP) evaluated along three dimensions: labourers (human capital/skills), means of labour (capital/technology), and objects of labour (products/processes/resources).

• Explanatory constructs

  • Digital industry agglomeration (DIA): regional clustering of digital firms/activities and related external innovation resources.
  • Digital–intelligent technology application (DIT): adoption and use of AI and digital-intelligent technologies within firms.
  • Knowledge breadth and regional innovation intensity: moderators/threshold variables reflecting the scope and depth of regional knowledge and R&D activity.

• Econometric approaches

  • Simultaneous equation models: capture two-way interactions and potential endogeneity between DIA and DIT.
  • Moderated mediation models: test mechanisms through which DIA and DIT jointly affect GP and how moderators (knowledge breadth) alter mediation.
  • Threshold models: identify nonlinearities and critical values beyond which marginal effects change (for synergy, knowledge breadth, innovation intensity).
  • Spatial Durbin models: estimate direct and indirect (spillover) effects across geographic neighbours and assess spatial heterogeneity by development stage.

• Data scope

  • The summary statement implies region-level analysis with spatial variation and measures of firm-level internal allocation, but specific datasets, sample periods, geographic coverage, and variable operationalizations are not provided in the brief.

Implications for AI Economics

• Micro–macro linkage: AI/digital technology adoption at the firm level (internal allocation) interacts with regional industrial structure (agglomeration) to determine productivity and environmental outcomes, highlighting the need for models that link firm decisions with regional ecosystems.
• Policy design: Simple promotion of AI adoption is insufficient—policies should promote balanced DIA and effective internal DIT deployment, expand regional knowledge breadth, and consider threshold levels to avoid wasted investment with limited marginal returns.
• Spatial policy coordination: Regional development strategies should account for spatial spillovers; pro-growth digital policies in developed regions may harm nearby areas unless coordinated (e.g., through complementary investments in neighbouring regions).
• Targeted interventions: Less-developed regions can benefit from incentives that foster DIA–DIT synergy to obtain positive spillovers; developed regions need policies that mitigate negative spillovers (e.g., support for peripheral firms, sharing of R&D platforms).
• Research directions: AI economics should further investigate:
  • Causal microfoundations of how DIT reallocates innovation inputs within firms.
  • The nature and measurement of threshold values for policy calibration.
  • Mechanisms behind negative spillovers from developed regions (competition for talent, market crowding).
  • Welfare- and distributional implications of DIA–DIT driven GP improvements across regions and sectors.

If you want, I can: (a) propose empirical specifications for testing these mechanisms, (b) outline likely operational definitions and data sources to replicate the study, or (c) draft policy recommendations tailored to a specific country/region.