Skip to content
Slow Variables

Ask the data

Answers come from the same store as the site; every number is checked against the record it cites.

Adaptation · Labour & consumers

Chatbot effect on earnings and hours in exposed occupations (Denmark)

consistent with normal80/95 confidence, good evidence, some ambiguitygrade Blagging

1What this measures

Difference-in-differences estimate of the effect of chatbot adoption on earnings and recorded hours for workers in exposed occupations, two years after ChatGPT, from Humlum and Vestergaard's Danish register study; the point estimates centre on zero with confidence intervals excluding effects above 2%.

Why it matters. The cleanest labour-market null on record, from administrative data covering the whole population. A normal technology shows up in earnings slowly; this is the slow showing.

Proxy types
welfare
Unit
share
Cadence
annual
Valve
adoption to adaptation

2How we track this

Normal band
≥ -2.0% and ≤ 2.0%
Fast band
≤ -5.0%
Falsifying
—

Normal = earnings and hours effects within two percent two years in; fast = a decline of five percent or more for exposed workers. Between is `emerging`.

Applied to humlum_pdf.denmark_exposed_workers.chatbot_effect_on_earnings.pt.

3Tracker interpretation

Zero, precisely. Adoption is high, time savings are real and modest, and none of it reaches pay yet.

4Evidence

Latest point
0.0%as of 2024-12-31
denmark_exposed_workers
Value the bands apply to
0.0%as of 2024-12-31

3 observations. Hollow points are disputed (see counterevidence). Every point links to its observation.

5Status and reasoning

consistent with normalsince 2026-09-10 · evaluate

Evaluator: 0 (Humlum working-paper PDFs (author site), 2024-12-31) is inside the consistent band (lo=-0.02 hi=0.02). Auto-reason; band rationale: Normal = earnings and hours effects within two percent two years in; fast = a decline of five percent or more for exposed workers. Between is `emerging`.

6Timeline notes

  • 2024-12-31 denmark_exposed_workers · 0.0%as of 2024-12-31

7Counterevidence

What cuts against this reading

Denmark is not the US; the window ends in 2024, before agentic tools; the effect could be a compositional wash across winners and losers.

8Update history

  1. 2026-09-10unmeasured to consistent with normalconf — → 80 · evaluate

    Evaluator: 0 (Humlum working-paper PDFs (author site), 2024-12-31) is inside the consistent band (lo=-0.02 hi=0.02). Auto-reason; band rationale: Normal = earnings and hours effects within two percent two years in; fast = a decline of five percent or more for exposed workers. Between is `emerging`.

    obs:dcb86d2c

9Confidence

80 / 95 — good evidence, some ambiguity

Confidence is independent of status: 90–95 multiple strong independent sources; 70–89 good evidence, some ambiguity; 50–69 mixed or hard to operationalise; below 50 limited or vague.

10Related