Digitalization measurably bolsters China’s grain supply-chain resilience: city-level evidence (2014–2024) shows digital technology raises resistance, speeds recovery and supports structural transformation mainly by improving resource allocation and industry integration; gains concentrate in eastern and plain areas and only fully emerge once digital infrastructure and farmer digital literacy pass critical thresholds.

Main Finding

Digital technology significantly increases grain industrial chain resilience in China (2014–2024). The effect is robust to endogeneity controls and several robustness checks, operates through two complementary mechanisms — (1) improving resource-allocation efficiency and (2) fostering industrial synergy and integration — and exhibits heterogeneous and non-linear (threshold) patterns: the resilience gains are stronger in eastern regions and plain areas, and are constrained until digital infrastructure and farmers’ digital literacy exceed critical thresholds.

Key Points

• Definition and decomposition: Grain industrial chain resilience is measured across three interlinked dimensions — resistance (ability to absorb shocks), recovery (speed/degree of post-shock restoration), and transformation (adaptive/evolutionary upgrading).
• Data: Panel of 285 prefecture-level Chinese cities, 2014–2024.
• Core independent variable: a composite digital technology index (constructed by the authors to capture local digitalization).
• Main empirical strategy: two-way fixed effects panel regressions, supplemented by instrumental-variable (IV) estimation to address endogeneity, panel threshold models to detect non-linear boundary effects, and mediation analysis to test mechanisms.
• Mechanisms:
  • Resource-allocation optimization — digital tools reduce information frictions, improve land/labor/capital/machinery matching, enable real-time dynamic scheduling, and thus raise resistance and recovery capacity.
  • Industrial synergy & integration — platforms, traceability, and coordination tools reduce transaction/coordination costs across production, processing, storage, logistics and markets, enabling better systemic coordination and longer-term transformation.
• Heterogeneity: Stronger positive effects in eastern regions and in plains; dimensional differences (across resistance/recovery/transformation) show point-estimate variation but lack full statistical confirmation.
• Thresholds: Two single-threshold effects identified — digital infrastructure and farmers’ digital literacy operate as boundary conditions; below certain levels, digitalization yields limited resilience gains; threshold patterns vary by resilience sub-dimension.
• Policy framing: Emphasizes whole-chain digital transformation targeted by region, infrastructure build-out, and human-capital (digital literacy) investments.

Data & Methods

• Scope: 285 prefecture-level cities in China, annual panel 2014–2024.
• Outcome: Composite grain industrial chain resilience index decomposed into resistance, recovery, and transformation capacities (multi-dimensional index rather than single-link metrics).
• Treatment variable: City-level composite digital technology index (constructed by the authors to capture local digital adoption/diffusion).
• Estimation techniques:
  • Two-way fixed effects (city and year) to control for time-invariant city heterogeneity and common shocks.
  • Instrumental-variable estimation to mitigate endogeneity of digitalization (IV details are reported in the paper).
  • Mediation (mediated-effects) framework to test resource allocation and industrial integration as channels.
  • Panel threshold models to formally test non-linear boundary conditions (digital infrastructure and farmers’ digital literacy).
  • Multiple robustness checks (alternative specifications, sample splits, etc.) — results are stable.
• Spatial/heterogeneity analysis: tests across regions (east/central/west) and terrain types (plain vs non-plain).

Implications for AI Economics

• Complementarities matter: Digitalization’s resilience benefits are conditional on physical infrastructure and human capital. For AI-driven agricultural interventions, returns will be limited without broadband/logistics platforms and sufficient user literacy. Economic models of AI adoption should include these complementarities and non-linear threshold effects.
• Mechanisms to model: AI/advanced digital tools enhance resilience primarily by (a) improving resource-allocation efficiency (better matching, dynamic reallocation, credit scoring) and (b) reducing coordination/transaction costs across firms and nodes (platform-mediated integration, traceability). Empirical work should explicitly model these mediated channels rather than treating AI as a black-box productivity shifter.
• Heterogeneous impacts & inequality risks: Stronger effects in better-endowed (eastern, plain) regions imply that AI/digital investments may widen regional disparities unless matched by targeted infrastructure and training in lagging areas. Policy evaluation and distributional analyses should account for spatial heterogeneity.
• Measurement guidance: Use granular (city- or subregional) panel data and multidimensional outcome metrics (resistance/recovery/transformation) to capture the systemic impacts of AI on food systems, rather than single-link productivity indicators.
• Policy & program design: For policymakers and donors aiming to use AI for food-system resilience, prioritize (1) digital infrastructure roll-out, (2) farmer digital literacy and training, and (3) platform/market integration policies that foster cross-segment coordination; incorporate threshold assessments to sequence investments efficiently.
• Research directions: Future AI-economics work should (a) identify valid instruments and causal strategies for AI/digital adoption, (b) explore dynamic/adaptive effects of AI on transformation capacity, and (c) study spillovers and network externalities within fragmented smallholder systems.

If you want, I can: (a) extract the paper’s exact mediating-variable measures and IV details from the methods/supplementary material, (b) produce a short slide-ready summary, or (c) map policy actions to budgetary/practical steps for closing the infrastructure and literacy thresholds.