2025年欧洲人工智能与生产力报告_37页_1mb

Methodology: Uses Acemoglu (2024) framework to estimate medium-term productivity gains from AI adoption, calibrated for 31 European countries based on task automatability, adoption rates, and labor cost savings. Adoption rates are driven by wage levels, with Europe-wide rates 5 percentage points lower than the US due to wage differences.

Core Findings:

• Medium-term productivity gains (5 years cumulative) average 1.1% across Europe, slightly higher than the US's 0.7% under Acemoglu's baseline.
• Variation is high: Luxembourg gains up to 2.95% while Romania sees only 0.3%. Wealthier countries benefit more due to higher exposure in AI-prone sectors (e.g., finance) and wage-driven adoption incentives.
• Regulation (occupation-level, EU AI Act, data privacy) reduces gains by up to 30% if AI exposure is halved in affected tasks/sectors.

Regulatory Impact:

• National regulations, EU AI Act, and data privacy laws collectively reduce Europe's productivity gains by over 30% compared to a preferred optimistic scenario. Data privacy laws have the smallest impact (~10% reduction).

Policy Implications:

• AI may not be a "silver bullet" for Europe's productivity growth or productivity gap with the US. Regulation could hinder gains, but cost-benefit trade-offs must be considered. Upside risks are higher in wealthier nations.