---
id: "lib-2026-07-00-has-ai-widened-employment-gaps-norway"
type: "library"
title: "Has AI Widened Employment Gaps Norway"
lang: "EN"
org: "RF Berlin"
source_type: "Paper"
date: "2026-07-01"
week: "2026-W27"
url: "https://nosov.ai/library/lib-2026-07-00-has-ai-widened-employment-gaps-norway/"
original_url: "https://www.rfberlin.com/wp-content/uploads/2026/07/26179.pdf"
tier: "A"
tier_name: "Strong evidence"
tier_reason: "Labour economists from the Frisch Centre and BI Norwegian Business School use the full Norwegian private-sector payroll register with event-study and within-firm designs, a strong dataset in a not-yet-peer-reviewed discussion paper."
summary_basis: "full-text"
original_minutes: "55"
tags: ["early career", "replication"]
---

# Has AI Widened Employment Gaps Norway

Tier A (Strong evidence): Labour economists from the Frisch Centre and BI Norwegian Business School use the full Norwegian private-sector payroll register with event-study and within-firm designs, a strong dataset in a not-yet-peer-reviewed discussion paper.

Summary based on the full text.

## In short

RFBerlin Discussion Paper 179/26 (July 2026) by Øystein Hernæs (Frisch Centre) and Andreas R. Kostøl (BI Norwegian Business School). It tests the US "canary in the coal mine" finding, that young workers in AI-exposed occupations lose jobs first, on the whole Norwegian private sector up to February 2026. Selected exposed occupations such as software development show a sharp early-career drop. Across the full distribution of occupations, the most exposed quintile of young workers did not fall behind the least exposed; the authors call it a relatively precise null.

## Key numbers

- 0.1% vs 0.3%: employment growth since October 2022 in the most vs least AI-exposed occupations (data to February 2026)
- -18%: per-capita employment change for software developers aged 21-30 after ChatGPT
- [-0.022, +0.065]: 95% confidence interval for the most- vs least-exposed quintile gap, ages 21-30
- about 3.1 million: employment records per month in the ARBLONN register

## What it reports

- Since October 2022, employment in the most AI-exposed occupations grew 0.1%, against 0.3% in the least exposed: an "AI index" of -0.2, small and not statistically significant.
- Software developers aged 21-30 fell about 18% per capita, between the US (-20%) and Sweden's raw figure (-44%). Customer service agents show a milder version.
- Low-exposure benchmarks show no young-worker drop: electricians aged 21-30 end near baseline; home health aides grow in every age group, the youngest most.
- For ages 21-30, the conventional 95% confidence interval for the top-vs-bottom quintile gap is [-0.022, +0.065]; it includes zero at every robustness setting.
- In a replication of the US design for ages 22-25, the average post-2022 gap is about 3 log points, but from spring 2025 it widens to about 35 log points by February 2026. The authors treat this as a pattern, not an identified effect, given failed pre-trend tests.
- Within-firm estimates for the youngest group end slightly positive, consistent with the cell-level result.
- Context: Norwegian firms using AI rose from 24% in 2023 to 55% in 2025 (4,294 firms surveyed); the policy rate went from zero in September 2021 to 4.5% by January 2024.

## How they got it

Monthly A-ordningen employer reports compiled into Statistics Norway's ARBLONN register, accessed through microdata.no, covering all formal private-sector jobs except self-employment. Occupations are ranked into quintiles using GPT-4-based exposure scores (Eloundou et al.) and Anthropic usage measures. The team runs Poisson event studies indexed to October 2022, HonestDiD robustness checks and within-firm difference-in-differences. Results feed a public monthly dashboard, kiindeksen.no.

## Limits

- The authors cannot separate AI from post-COVID correction, monetary tightening and workers shifting careers in anticipation.
- Exposure measures capture theoretical task overlap, not actual use at the employer.
- Discussion paper, not peer reviewed; the late 22-25 divergence rests on the final ten months of data.

## Why it matters for companies

Headline stories of collapsing entry-level hiring come from a few occupations, mainly software and customer service. Across Norway's whole economy the gap is not visible yet, which argues for tracking your own hiring data by role rather than reacting to national anecdotes. The late widening for 22-25 year-olds is a signal worth watching in graduate programmes. Strong employment protection and centralised wage bargaining, similar to much of continental Europe, may slow or mask effects that appear faster in the US.

Original: https://www.rfberlin.com/wp-content/uploads/2026/07/26179.pdf