Digital tools — from IoT to AI — materially strengthen supply-chain resilience by improving visibility, which accounts for roughly two-thirds of the resilience gains; the benefits are largest in complex, technology-intensive industries.

Main Finding

Digital technology adoption—specifically IoT, blockchain, AI/machine learning, big data analytics, and cloud computing—significantly improves supply chain visibility and resilience. Visibility is the dominant mechanism: it mediates 67.4% of the total effect of digital technologies on resilience. Results are robust to SEM, hierarchical OLS, IV, lagged models, and propensity score matching on a cross‑national sample of 742 manufacturing and logistics firms.

Key Points

• Sample and scope
  • N = 742 firms (manufacturing & logistics) across 23 countries (Asia‑Pacific, Europe, North America, emerging markets). Data collected Jan 2023–Jun 2024 via survey; some financials from Orbis/Compustat.
• Main quantitative results
  • DTA → SCV: β = 0.412, p < .001 (additional ΔR² = 0.204).
  • SCV → SCR: β = 0.486, p < .001 (ΔR² = 0.258).
  • DTA → SCR (direct): β = 0.298, p < .001; when SCV included direct effect attenuates to β = 0.097, p < .05.
  • Indirect effect (DTA → SCV → SCR): 0.201 (95% bootstrap CI [0.156, 0.253]); Sobel Z = 8.745, p < .001.
  • Proportion mediated by visibility: 67.4% (partial mediation).
• Technologies operationalized
  • DTA measured as 2nd‑order construct with 20 items across IoT, blockchain, AI/ML, big data analytics, cloud computing (4 items each). High reliability: α = 0.91; CR = 0.93; AVE = 0.62.
• Robustness & validity
  • CFA and measurement model fit: CFI = 0.963, TLI = 0.958, RMSEA = 0.048, SRMR = 0.052.
  • Endogeneity addressed via industry‑average adoption IV (first‑stage F = 196.34), lagged predictors, and propensity score matching; IV second‑stage β ≈ 0.314, p < .001.
  • Common method checks: Harman single factor = 38.2%; procedural remedies used.
• Heterogeneity
  • Effects stronger in high‑complexity supply chains and technology‑intensive industries (reported qualitative result in paper).
• Controls included
  • Firm size (ln employees), age, ROA, supply chain complexity, environmental dynamism, industry and region fixed effects.

Data & Methods

• Research design
  • Cross‑sectional survey of senior supply chain executives/operations managers; purposive sampling from Dun & Bradstreet and industry associations; 30.1% initial response rate (856), final N=742 after QC.
• Measurement
  • Constructs: Digital Technology Adoption (multi‑dimensional), Supply Chain Visibility (upstream/internal/downstream), Supply Chain Resilience (anticipation, resistance, response/recovery, adaptation). All Likert scales (7‑point). Objective resilience proxy: coefficient of variation of 3‑year revenue used as a robustness check.
• Analytical approach
  • Confirmatory factor analysis (lavaan) to validate measures.
  • Structural equation modeling and hierarchical OLS regressions to test hypotheses.
  • Mediation tested via Preacher & Hayes bootstrap (5,000 reps) and Sobel test.
  • Endogeneity strategies: instrumental variables (industry average DTA), lagged predictors, propensity score matching.
  • Standard errors clustered by industry.
• Limitations noted by authors
  • Cross‑sectional nature limits causal claims despite IV/lag/PSM efforts; reliance on self‑reported measures (though triangulated with secondary financials); generalizability beyond manufacturing/logistics needs further testing.

Implications for AI Economics

• Returns to AI and digital capital
  • AI contributes meaningfully to firm resilience primarily through improving information visibility—implying returns to AI investment depend on complementary systems (IoT sensors, data platforms, interoperability). Economists estimating returns to AI should model complementarity and mediation through information flows rather than treating AI in isolation.
• Complementarities and bundling
  • The study operationalizes DTA as a bundle (IoT, blockchain, AI, analytics, cloud). Cost–benefit analyses and investment models should account for complementary investments and joint production of visibility and resilience.
• Heterogeneous effects and targeting
  • Stronger effects in high‑complexity supply chains and technology‑intensive industries suggest heterogeneity in marginal returns to AI/digital capital. Policy subsidies or firm investment decisions could be targeted where gains (in resilience) are highest.
• Market structure and competition
  • Improved visibility and resilience can alter competitive dynamics (lower disruption costs, faster response), potentially affecting market concentration if early adopters secure durable advantages. Models of industrial dynamics should consider digitalization as a strategic barrier.
• Labor and skills
  • The mechanisms (real‑time monitoring, predictive analytics) imply demand shifts toward analytics, data engineering, and supply chain orchestration skills. Labor market impacts include upskilling needs and potential reallocation of tasks; economic estimates should include human capital complementarities.
• Measurement suggestions for economists
  • Use multi‑dimensional indices of digital adoption (as in this paper) and incorporate visibility as an intermediate outcome when linking digital adoption to productivity or performance.
  • Consider objective resilience proxies (revenue volatility, downtime costs) alongside survey measures; longitudinal or administrative data would strengthen causal inference.
• Policy implications
  • Infrastructure (connectivity, cloud/data centers) and interoperability/standards (for IoT/blockchain data exchange) materially enable the visibility channel—policy can improve social returns to private AI/digital investments.
  • Data governance and trust frameworks (e.g., standards for provenance, privacy) will affect adoption rates and the magnitude of resilience benefits.
• Research agenda for AI economics
  • Causal identification: randomized rollout or phased adoption designs to estimate local average treatment effects of AI components.
  • Macro and sectoral impacts: quantify how improved resilience via AI reduces aggregate supply shocks and their propagation.
  • Cost‑effectiveness: estimate marginal costs of achieving incremental visibility and the resulting reduction in disruption losses.
  • Distributional effects: assess which firms/workers capture the benefits and how digitalization affects welfare across regions and industries.

If you want, I can: - Extract the exact survey items or scales used for each construct (useful for replication). - Propose an empirical strategy (data sources and identification) for estimating causal returns to AI investments in supply chains using administrative data.