It is important to note that we capture predicted changes based on which tasks could be automated. Since many of these tasks have not been automated yet, the effects are likely to grow significantly larger as AI becomes increasingly powerful.
📢 Call for papers!
We are organizing the 6th Early Career Workshop in Quantitative Political Economy on 14-15 May 2026 at King’s College London!
Keynote: Shanker Satyanath (NYU)
No fee, travel grants might become available!
Submit at: tinyurl.com/qpe2026
This is exactly what we find. Occupations where AI targets low-expertise tasks experience wage increases compared to occupations where high-expertise tasks are exposed to AI automation.
Second contribution: Building on Autor and Thompson (2025), we argue it matters not just how many tasks are exposed, but which tasks. We distinguish:
• AI that automates low-expertise tasks → raises skill requirements → wages ⬆️
• AI that automates high-expertise tasks → lowers barriers → wages ⬇️