Date context: March 2026 revision; original working paper issued May 2025
The March 2026 working-paper revision links Danish adoption surveys to administrative labour records. Workers report benefits and new AI-related tasks. Difference-in-differences estimates show precise early null effects on earnings and recorded hours, ruling out effects larger than 2% over the reported horizon.
NBER; Humlum and Vestergaard · 2026-03
Primary publisher text checked on 2026-11-02. Scope and limits stated below; independent results not reproduced.
What the source shows
The March 2026 working-paper revision links Danish adoption surveys to administrative labour records. Workers report benefits and new AI-related tasks. Difference-in-differences estimates show precise early null effects on earnings and recorded hours, ruling out effects larger than 2% over the reported horizon. Source: NBER; Humlum and Vestergaard, 2026-03
Limits and reading context
This is a working paper and a particular early Danish setting. Earnings, recorded hours, task speed and service quality are different outcomes. Survey responses about expected use of saved time do not demonstrate observed reallocation. The publisher abstract and version record were read; a fresh full-method replication was not performed. Source: NBER; Humlum and Vestergaard, 2026-03
For HR, AI + JJ
AI + JJ, application: Track both task-level changes and longer-term work outcomes. What benefit matters locally if faster tasks do not immediately change paid hours or wages? Source: NBER; Humlum and Vestergaard, 2026-03