China's twin innovation pilots drive listed high‑tech firms to boost overseas technology investment, with the policy pair producing a stronger effect than either alone; moderate levels of digital supervisory procurement amplify this effect but excessive supervision erodes it.

Main Finding

The joint implementation of two place-based innovation policies in China — the National Innovation Demonstration Zones and the Comprehensive Innovation Reform Pilot Zones — significantly increases A‑share listed high‑tech firms’ overseas technology investment. The combined (dual) policy produces a stronger effect than either policy alone, indicating a synergistic (complementary) relationship. Supervisory technology procurement moderates this effect with an inverted‑U relationship, and the policy impact is concentrated among non‑state firms, firms with weaker measured innovation capability, and firms located in eastern China.

Key Points

• Dual-policy effect: The dual pilot framework raises firms’ overseas technology investment more than either demonstration zone policy implemented in isolation, implying institutional complementarity across place‑based innovation policies.
• Moderation by supervision tech: Supervisory technology procurement (digital supervisory inputs) shows an inverted‑U shaped moderating effect — moderate levels amplify the dual policy’s impact, whereas too much procurement diminishes it.
• Heterogeneous impacts:
  • Stronger effects for non‑state‑owned enterprises (non‑SOEs).
  • Larger effects for firms with weaker internal innovation capabilities.
  • Concentrated effects among firms based in eastern China.
• Contribution: Adds evidence on policy bundle complementarities and on how governance/monitoring technologies interact with place‑based innovation policies.

Data & Methods

• Sample: Panel of A‑share listed high‑tech firms in China, 2009–2023.
• Treatment(s): Location in National Innovation Demonstration Zones and/or Comprehensive Innovation Reform Pilot Zones; analysis compares dual-policy exposure vs single/no policy exposure.
• Estimation approach: Double Machine Learning (DML) to estimate causal policy effects while flexibly controlling for many covariates and reducing model‑selection bias.
• Additional analyses:
  • Comparison of dual vs single policy effects to assess synergy.
  • Interaction tests with supervisory technology procurement to identify the inverted‑U moderation.
  • Heterogeneity checks by ownership type, firm innovation capability, and geographic region.

Implications for AI Economics

• Policy bundles matter: Analysis underscores that firms’ cross‑border technology decisions respond to combinations of place‑based policies; single‑policy evaluations may understate true policy impacts when complementarities exist.
• Role of digital supervision: The inverted‑U finding suggests digital supervisory tools can enhance firms’ outward technology investment up to a point, but excessive surveillance/regulation may crowd out investment — relevant for designing AI governance tools that aim to balance oversight and innovation.
• Targeting and equity: Stronger effects for non‑SOEs and weaker innovators imply place‑based policy bundles can help private and less‑capable firms access overseas technology, shaping international AI capability flows and competitive dynamics.
• Methodological takeaways: Use of DML demonstrates the value of machine‑learning–based causal methods for policy evaluation in AI economics, especially when handling high‑dimensional controls and complex treatment structures.
• Caution on generalizability: Results pertain to publicly listed high‑tech firms in China and two specific policy instruments; extrapolation to unlisted firms, other countries, or different policy mixes should be done cautiously.