Targeted AI R&D and governance, not overall AI vibrancy, raise tourism’s share of GDP; human-capital gains materialize with a one-year lag while COVID cut tourism by about 37%.

Main Finding

Overall national AI vibrancy (Stanford HAI AI Vibrancy Score) does not significantly affect tourism’s direct share of GDP within countries (β = 0.061, p = 0.622). However, disaggregating the AI ecosystem reveals that two pillars—AI-related R&D (β = 1.811, p = 0.005) and Policy & Governance (β = 0.353, p = 0.037)—significantly increase tourism’s GDP share. The Talent pillar shows a significant positive effect with a one-year lag (β = 0.183, p = 0.025). The COVID‑19 pandemic sharply reduced tourism’s GDP share (~37%; β = –0.455, p < 0.001), and AI development did not moderate that decline.

Key Points

• Sample and outcome: Panel of 33 countries, 2017–2023; dependent variable = tourism’s direct share of GDP.
• Main aggregate result: The composite AI Vibrancy Score has no measurable within-country effect on tourism’s GDP share once macro controls are included.
• Pillar decomposition: Significant positive within-country effects for:
  • AI R&D (large effect, β = 1.811, p = 0.005)
  • Policy & Governance (β = 0.353, p = 0.037)
• Human capital: Talent pillar’s effect manifests with a one-year lag (β = 0.183, p = 0.025), implying delayed returns from AI-related human-capital improvements.
• COVID-19: Pandemic reduced tourism’s share by ~37%; no evidence that stronger AI ecosystems mitigated this shock.
• Robustness: Results hold under alternative standard-error clustering and two-way fixed-effects specifications; dynamic and interaction models were estimated to probe timing and moderation.

Data & Methods

• Data: 33-country panel (2017–2023); Stanford HAI AI Vibrancy Score (aggregate and seven pillar indices) as main AI measures; tourism direct GDP share as outcome; macroeconomic controls included (unspecified here).
• Identification strategy: Within-country (country fixed-effects) panel estimation to capture changes over time within countries; clustered standard errors.
• Additional specifications: Lagged regressions to capture delayed effects (notably for Talent), dynamic models, and interaction terms (including tests for AI moderating COVID impacts).
• Robustness checks: Alternative standard-error procedures and two-way fixed-effects models (country and year)—pillar results remain robust.

Implications for AI Economics

• Aggregate AI indicators can conceal sector-specific pathways: decomposed metrics are necessary to identify which elements of an AI ecosystem matter for economic sectors like tourism.
• Policy focus: Investments in AI R&D and clear governance frameworks appear most effective at increasing tourism’s contribution to GDP; talent development matters but shows lagged effects—expect delayed benefits from human-capital policies.
• Shock resilience: AI ecosystem development, as measured here, did not buffer the pandemic’s negative impact on tourism; resilience may require different policies or faster-adopting firm- and destination-level interventions.
• Research implications: Future work should (a) unpack mechanisms (e.g., product/service innovation, productivity in hospitality, marketing and personalization), (b) examine heterogeneity by countries’ baseline tourism dependence, and (c) use firm- or destination-level data to validate channels and timing.
• Policy caution: Relying on composite AI vibrancy scores to forecast sectoral economic impacts risks missing actionable levers—targeted metrics and policy design are preferable.