China’s manufacturing digital transformation is linked to larger foreign capital inflows by raising production efficiency, but gains are concentrated in central provinces; eastern and western regions show no significant effect, and innovation and upgraded human capital appear to mediate the relationship.

Main Finding

China's manufacturing digital transformation (2002–2023) significantly increases the manufacturing sector's ability to attract foreign direct investment (FDI), primarily by raising production efficiency. This positive effect operates partly through two mediators—technological innovation (measured by patent applications) and upgrading of human-capital structure (share of high‑tech manufacturing employees)—and shows pronounced regional heterogeneity: effects are relatively strong and significant in central China but insignificant in eastern and western regions.

Key Points

• Core result: Higher manufacturing digitization → larger logged FDI inflows into provincial manufacturing.
• Mechanisms confirmed:
  • Technological innovation (patent volume) mediates a portion of the effect.
  • Human-capital structure upgrading (share of high‑tech manufacturing employees) also mediates the effect.
• Regional heterogeneity:
  • Central region: significant positive effect.
  • Eastern region: effect statistically insignificant (possible FDI saturation / high baseline).
  • Western region: effect insignificant (weaker infrastructure and industrial support).
• Policy recommendations from the authors:
  • Continue advancing digital transformation (including “AI+” initiatives).
  • Strengthen R&D investment and talent development.
  • Implement regionally differentiated policies to optimize FDI structure and climb the global value chain.
• Caveats implied or partly discussed:
  • Measurement choices (digitization proxy based on ADB‑MRIO J62–63 sectors) may omit other digital inputs (e.g., robotics, 5G deployment).
  • Potential endogeneity and omitted-variable concerns; paper uses province and year fixed effects but causal identification beyond that is not detailed in the excerpt.

Data & Methods

• Sample: Provincial-level manufacturing panel, 31 Chinese provinces, yearly data 2002–2023 (N = 682).
• Dependent variable: Log of actual utilized FDI in manufacturing by province-year.
• Core explanatory variable (Digit / Dig): Manufacturing digitization measured using ADB-MRIO input‑output framework—ratio of intermediate goods and industrial value added invested in computer programming, consultancy and related activities, and information service activities (ADB sectors J62–63), aggregated to provinces by weighting.
• Mediators:
  • Technological innovation (TEIO): log of patent application counts.
  • Human capital structure (HC): ratio of high‑tech manufacturing employees to total urban manufacturing employees.
• Control variables: log GDP (market scale), fiscal revenue/GDP, openness (import+export/GDP), infrastructure (road mileage per capita), log average wage in urban manufacturing non-private entities.
• Econometric approach:
  • Baseline panel OLS with province fixed effects and year fixed effects: FDI_it = α0 + α1 Digit_it + α2 Controls_it + μ_i + λ_t + ε_it
  • Mediation tested using the Wen Zhonglin approach (sequential equations estimating Digit → Mediator, then Digit + Mediator → FDI).
• Data sources: ADB-MRIO tables (digitization), China Industrial Statistical Yearbook, China Labor Statistical Yearbook, China Science & Technology Statistical Yearbook, provincial yearbooks, China Statistical Yearbook.
• Descriptive statistics reported (mean, min, max, sd) for key variables across the sample.

Implications for AI Economics

• AI as central component of industrial digitization: The paper frames “AI+” within manufacturing digital transformation; its positive effect on attracting FDI implies that AI adoption is an investment magnet—particularly for efficiency-seeking, technology‑intensive foreign investors.
• Productivity and comparative advantage:
  • AI-driven productivity improvements raise returns to capital and can shift provincial comparative advantage toward higher‑value activities, altering FDI composition (more high‑tech, less labor‑intensive projects).
• Skill bias and labor-market impacts:
  • The mediating role of human-capital upgrading highlights that AI adoption increases demand for skilled workers; regions lacking skills may fail to convert AI adoption into FDI gains, reinforcing regional divergence.
  • Policy must address upskilling and reskilling to capture AI-related FDI benefits while managing displacement risks.
• Global value chains and location decisions:
  • Digitalization (including AI) lowers coordination and information frictions, making foreign firms more willing to locate complex activities remotely—this can move manufacturing up the value chain if matched with local innovation capacity.
• Policy design for AI-related FDI:
  • Investment attraction policies should combine digital infrastructure (5G, cloud, industrial internet) with human-capital programs and R&D incentives to maximize FDI quality rather than just quantity.
  • Regional differentiation matters: central China may be prime for scaling AI-enabled manufacturing investment, while eastern regions may need policies targeting high-end, knowledge-intensive FDI; western regions require infrastructure and capacity-building before AI policies pay off.
• Research implications for AI economics:
  • Measurement: studies should refine digitization metrics to explicitly capture AI adoption (e.g., firm-level AI use surveys, expenditures on AI systems, robot/automation density).
  • Identification: causal inference on AI → FDI would benefit from instruments or quasi‑experimental designs (e.g., staggered rollout of industrial internet/5G, policy shocks).
  • Distributional effects: further work should quantify how AI-driven FDI changes wage structure, employment composition, and regional inequality within and across provinces.

If you want, I can: - Produce a one‑page brief tailored for policy makers focusing on AI-related recommendations; or - Draft a short critique suggesting robustness checks and alternative identification strategies (instrumental variables, diff‑in‑diff, firm-level analysis) to strengthen causal claims.