China’s supply‑chain digitalization pilot improved resilience at listed manufacturers—non‑SOEs, mature firms and those in less marketized regions saw the biggest gains—but evaluative (nonrandom) selection into the pilot means the effect may not be purely causal.

Main Finding

Exposure to China’s Supply Chain Innovation and Application Pilot Program — a staggered, quasi-natural experiment in supply chain digitalization — is associated with a statistically significant increase in firm-level supply chain resilience for Chinese A‑share listed manufacturing firms (2012–2023). The result is robust to a battery of checks, but remains subject to possible selection on unobserved characteristics since pilot assignment was evaluative rather than randomized.

Key Points

• Identification: Uses the staggered rollout of a national pilot program as a quasi-experiment and difference‑in‑differences (DID) estimation to measure policy effects on supply‑chain resilience.
• Robustness: The positive coefficient on policy exposure survives parallel‑trends testing, placebo exercises, additional covariates, high‑dimensional fixed effects, and policy‑selection diagnostics.
• Caveat: Pilot assignment followed an evaluation process (not random), so estimates may reflect selection on unobservables or anticipatory behavior.
• Heterogeneity:
  • Larger positive effects for firms operating in less marketized regions.
  • Stronger effects for non‑state‑owned enterprises (non‑SOEs).
  • Stronger effects for mature firms (vs. young firms).
• Mechanisms:
  • Sobel mediation tests support a financing‑constraint pathway (digitalization alleviates financing frictions → greater resilience).
  • The information‑transparency pathway is only marginally significant.
• Operational correlates: Asset turnover, operating cycle, and inventory alignment correlate with resilience in outcome equations, but corresponding first‑stage links to policy exposure are not robust in firm‑ and year‑fixed effects specifications.
• Conclusion nuance: Evidence documents a positive quasi‑experimental association between supply‑chain digitalization and resilience, while highlighting institutional and firm‑level conditions that shape the effect.

Data & Methods

• Data: Panel of Chinese A‑share listed manufacturing firms covering 2012–2023.
• Treatment: Exposure to the Supply Chain Innovation and Application Pilot Program (staggered rollout across regions/firms).
• Empirical strategy:
  • Difference‑in‑differences with staggered adoption.
  • Event‑study / parallel‑trends tests to check pre‑trends.
  • Placebo exercises and additional controls to probe robustness.
  • High‑dimensional fixed effects (firm and year FE) and policy‑selection diagnostics to address selection concerns.
  • Subgroup regressions to assess heterogeneity.
  • Sobel mediation tests to probe financing and information transparency channels.
• Limitations: Nonrandom pilot assignment implies potential remaining bias from unobserved confounders; some first‑stage links for operational channels are not robust under preferred specifications.

Implications for AI Economics

• Digitalization as resilience policy: The paper provides quasi-experimental evidence that digital supply‑chain interventions raise firm resilience — relevant when modeling the economic value of AI/supply‑chain digitization (e.g., predictive analytics, demand forecasting, automated inventory control).
• Mechanism prioritization: Financing constraints appear a more important mediator than information transparency in this context. For AI economics, that suggests digital tools may create value partly by reducing financing frictions (e.g., better verifiable operational data that eases credit), not only by improving informational efficiency.
• Heterogeneous returns: Policy and investment in digital/AI tools are likely to yield larger resilience gains in less marketized regions, non‑SOEs, and mature firms. Targeting and adoption incentives should account for these heterogeneous marginal returns.
• Measurement & evaluation: Staggered DID and rich panel methods are appropriate for policy evaluation of digital/AI programs, but nonrandom program selection requires careful diagnostics, instruments, or experimental designs for causal claims.
• Research directions:
  • Test AI‑specific technologies (ML forecasting, optimization, automated contracting) separately from broader digitalization.
  • Use randomized rollouts or better instruments to address selection on unobservables.
  • Extend analysis to non‑listed firms and upstream/downstream suppliers to capture network spillovers.
  • Study dynamic resilience during acute shocks (e.g., pandemic, trade disruptions) to quantify short‑run versus long‑run benefits.
  • Quantify general‑equilibrium effects on credit markets if digitalization systematically reduces financing frictions.
• Policy takeaway: Investments and policy support for supply‑chain digitalization (including AI applications) can strengthen firm resilience, but effectiveness depends on institutional context and firm characteristics — and rigorous causal evaluation is still needed to guide scaling and targeting.