Digital infrastructure strengthens agricultural resilience across Chinese provinces, with gains magnified once digital inclusive finance passes two critical thresholds and in more marketized provinces; effects are strongest in China’s central region.

Main Finding

Digital infrastructure significantly improves agricultural sustainable development resilience across Chinese provinces (2013–2024). This effect operates partly through digital inclusive finance (which shows a double‑threshold, nonlinear mediating role), is strengthened by higher levels of marketization, and is geographically heterogeneous (stronger in the central region).

Key Points

• Primary result: Long-term positive effect of digital infrastructure on agricultural sustainable development resilience.
• Mediation: Digital inclusive finance is a significant partial mediator; its mediating role exhibits a double‑threshold characteristic (i.e., the empowering effect changes nonlinearly across levels of digital inclusive finance).
• Moderation: Marketization positively moderates the impact—more market‑oriented provinces gain larger resilience benefits from digital infrastructure.
• Heterogeneity: The empowering effect is more pronounced in the central region of China.
• Policy prescriptions offered: optimize regional digital infrastructure layout, apply threshold‑based differentiated policies for digital inclusive finance, and strengthen market‑oriented support mechanisms.

Data & Methods

• Data: Panel dataset covering 30 Chinese provinces, 2013–2024.
• Index construction: Entropy method used to construct composite measures of (a) agricultural sustainable development resilience and (b) digital infrastructure development.
• Econometric strategy:
  • Two‑way fixed effects (FE) panel models to estimate baseline relationships controlling for province and year fixed effects.
  • Mediation analysis to test the role of digital inclusive finance as a channel.
  • Threshold models (double‑threshold) to identify nonlinearities in the mediating channel.
  • Moderation analysis to assess how marketization levels alter the effect.
  • Additional robustness checks and complementary econometric approaches (as reported) to support inference.
• Outcome: Robust evidence of causal pathways consistent with digital infrastructure → (via digital inclusive finance, modulated by marketization) → greater agricultural resilience, with regional variation.

Implications for AI Economics

• Infrastructure as a complementary asset: Findings illustrate that digital infrastructure (broadband, digital platforms, payments rails) is a necessary enabler of technology-driven productivity and resilience gains in agriculture. For AI economics, this underscores that returns to AI investment in agriculture depend critically on underlying digital infrastructure.
• Role of fintech / data platforms: Digital inclusive finance acts as a channel—AI‑driven credit scoring, risk assessment, and transaction platforms can amplify technology adoption and resilience. However, nonlinear threshold effects imply diminishing or accelerating returns depending on digital‑finance maturity; policymakers and firms should target critical mass thresholds for platform and data adoption.
• Institutions matter: Marketization amplifies tech benefits. AI deployment and markets for AI services will yield larger welfare/resilience gains where market institutions support competition, property rights, and factor reallocation—this should be modeled explicitly in AI diffusion and welfare analyses.
• Heterogeneity & targeting: Regional heterogeneity means one‑size‑fits‑all models of AI or digital policy are likely to misestimate impacts. Spatially explicit, heterogeneous-agent models and region‑specific cost–benefit analyses are appropriate.
• Methodological takeaway: Use of composite indices (entropy method), panel FE with mediation and threshold models is a useful empirical template for AI economics research studying infrastructure–technology–outcome channels. Future work should combine such macro panel approaches with microdata, causal identification strategies (natural experiments, instrumental variables), and equilibrium models to quantify welfare, distributional effects, and optimal policy sequencing for AI and digital infrastructure investments.