reference
The Projected Impact of Generative AI on Future Productivity Growth
Penn Wharton Budget Model analysis (September 2025) projecting AI's impact on Total Factor Productivity using Acemoglu's framework. Estimates ~42% of jobs exposed to AI automation, but only ~10% of GDP likely impacted due to adoption constraints. Projects modest TFP contribution of 0.01pp in 2025.
Source Details
- Publisher: Penn Wharton Budget Model
- Date: September 10, 2025
- Framework: Based on Acemoglu (2024) derivation
- Methodology: Task-level exposure analysis using Eloundou et al. (2024) classification
Key Findings
Exposure vs. Impact Gap
- 42% of jobs potentially exposed to AI (β₯50% of tasks automatable)
- Only ~10% of GDP likely to be impacted in practice
- Gap explained by adoption feasibility and profitability constraints
Adoption Reality Check
Based on Svanberg et al. (2024):
- Only 23% of exposed tasks will eventually be automated
- Feasibility and profitability filter out most theoretical exposure
Labor Cost Savings
Real-world studies show gains ranging 10-55%, with average of ~25% on adopted tasks.
TFP Contribution Projections
| Year | TFP Impact |
|---|---|
| 2025 | 0.01 pp |
| Future | Growing to ~15% GDP share over two decades |
Employment Shifts Already Visible
- Jobs with 100% AI exposure: Employment fell sharply in 2024 (0.75% below 2021)
- Jobs with 90-99% exposure: Growth slowed significantly since 2022
Caveats Noted by Authors
The analysis does not account for:
- AI-driven changes in product quality
- Emergence of new products and labor tasks
- Potential impact on innovation/TFP feedback loops
Relevance to MPDrexel Framework
Validates Jobless Boom Thesis
- Employment effects are real but concentrated
- GDP growth can decouple from labor expansion
- "Silent attrition" pattern emerging in exposed roles
Informs Engagement Expectations
- Client expectations should be calibrated to ~25% task-level savings
- Not all efficiency potential will be captured (adoption constraints)
- Long-term macro effects still unfolding