Digital infrastructure, not AI alone, drives sustainable development in the MENA region; AI helps but mainly where institutions and digital foundations are strong.

Main Finding

Digital transformation (DT) — interpreted as broad digital infrastructure and system-level digitalization — is the dominant driver of sustainable development in MENA (2010–2023). Artificial intelligence (AI) has a positive effect but substantially weaker and more conditional: its impact depends on maturity, complementary assets (data, human capital), and institutional quality. GCC countries show significantly stronger DT and AI effects than non‑GCC peers, owing to better infrastructure, governance, and absorptive capacity.

Key Points

• Primary result: DT → larger, more consistent positive effects on integrated SDG-related outcomes (economic, social, environmental) than AI.
• AI effect: positive but smaller and heterogeneous; functions mainly as an efficiency/complementarity enhancer within an established digital ecosystem.
• Regional heterogeneity: GCC countries experience stronger technology→sustainability effects than non‑GCC countries.
• Institutional moderating role: governance quality, regulatory frameworks, and absorptive capacity strengthen the translation of DT and AI into sustainable development.
• Policy analysis: paper supplements econometrics with desk-based review of national AI/digital strategies (e.g., Saudi Vision 2030, UAE strategies), SDG reports, and e‑government/SDI governance assessments.
• Hypotheses tested: H1–H5 (DT and AI positive effects; DT stronger than AI; stronger effects in GCC; institutional quality moderates impacts).

Data & Methods

• Coverage: Selected MENA economies, annual panel 2010–2023. Dependent variable: integrated sustainable development outcome proxied by SDG-related indicators (economic, social, environmental dimensions). (Paper couples econometrics with qualitative policy review; specific index construction not fully detailed in the extracted text.)
• Main estimator: System Generalized Method of Moments (System GMM) for dynamic panel data to:
  • Control for endogeneity (reverse causality, omitted variables),
  • Account for persistence via lagged dependent variable,
  • Remove unobserved country fixed effects.
• Instruments: internal instruments (lagged levels and differences) following standard System GMM practice; robustness checks via Fixed Effects and 2SLS.
• Qualitative component: literature synthesis, national strategy and SDG report review, institutional/governance assessment, and GCC vs non‑GCC comparative analysis.
• Reported robustness: results robust to alternative estimators; emphasis on instrument validity and addressing dynamic panel concerns (standard System GMM diagnostics implied, though not shown in the excerpt).

Implications for AI Economics

• Role of AI as a complement, not substitute: Economically, AI delivers gains primarily when digital infrastructure and complementary assets exist. Models of technology-driven growth should treat AI as a conditional general‑purpose technology whose productivity returns hinge on absorptive capacity (human capital, data ecosystems, regulation).
• Prioritization for policy and investment: For rapid SDG gains in emerging regions, prioritizing DT/infrastructure yields larger aggregate sustainability returns; targeted AI deployment becomes high‑value once DT foundations are in place.
• Institutional economics: Institutional quality is a key multiplier — economic models and policy cost–benefit analyses should incorporate governance and regulatory capacity as interaction terms when forecasting AI impacts.
• Distributional considerations: The weaker/more conditional effect of AI suggests potential for widening divergence (productivity and welfare) across countries and within countries; micro‑level and distributional studies (firm‑ or household‑level) are needed to quantify these dynamics.
• Measurement and empirical work:
  • Need for better, comparable country‑level measures of AI adoption and maturity (beyond broad ICT proxies) to identify non‑linear and sectoral effects.
  • Future causal work should link national/regional outcomes to firm‑level AI deployments and sectoral adoption to unpack mechanisms (productivity, employment, emissions).
  • Account for environmental externalities of AI (energy use, e‑waste) when evaluating net sustainability impacts; include “sustainable AI” metrics in welfare calculations.
• Policy design: Economic policy should bundle investments — digital infrastructure, data governance, human capital, and regulatory reform — rather than funding AI in isolation. Subsidies, standards, and capacity building targeted at non‑GCC countries can narrow regional gaps.
• Research gaps highlighted by the paper: integrated SDG outcome measurement, explicit quantification of AI’s conditional effects, heterogeneity across institutional contexts, and long‑run dynamics of AI diffusion in emerging economies.

Limitations to note for follow‑up work: country‑aggregate analysis may mask intra‑country and sectoral heterogeneity; the paper relies on System GMM assumptions (instrument validity, no second‑order autocorrelation); and specifics of the sustainable development index and AI/Digital measures should be inspected in the full paper for replication and extension.