
Columbia University
Ivy League university; research confirmed 75% of unemployed Americans never file UI claims.
Last refreshed: 5 April 2026
If three in four displaced workers never file claims, can we measure AI's true toll?
Latest on Columbia University
- What did Columbia University find about unemployment insurance?
- Research by Columbia and Fortune confirmed that roughly 75% of unemployed Americans never file for unemployment insurance.Source: Fortune / Columbia
- Why do most unemployed people not file for benefits?
- Severance packages delay filing, recent graduates lack work history to qualify, and contractors are categorically ineligible.Source: Fortune / Columbia
- Is AI unemployment being undercounted?
- Yes. The workers AI displaces (tech professionals with severance, contractors, graduates) are the least likely to appear in claims data.Source: Fortune / Columbia
Background
Columbia University is an Ivy League research university founded in 1754 and located in Morningside Heights, New York City. It is one of the oldest and most research-intensive universities in the United States, with particular strength in law, economics, public policy, and journalism. In the context of AI workforce coverage, Columbia's research with Fortune confirmed that approximately 75% of unemployed Americans never file for unemployment insurance — a finding with direct implications for measuring AI-driven displacement.
The 75% non-filing rate illuminates a fundamental flaw in using UI claims as a proxy for AI job losses. Workers most likely to be displaced by AI — higher-earning technology professionals receiving severance packages, recent graduates with insufficient work history to qualify, independent contractors ineligible by classification, and gig workers excluded by statutory definition — are precisely those least likely to appear in weekly initial claims data. The standard BLS data that policymakers use to monitor the labour market therefore systematically undercounts AI-specific displacement even as that displacement accelerates.
The Columbia finding connects directly to the measurement crisis running through all AI employment coverage in 2026. New York's first-in-nation AI WARN Act produced zero AI attributions from 162 companies. The nine-senator Coalition wrote to the BLS demanding better data. The Challenger Report — the most comprehensive AI-attribution tracker — is based on employer self-reporting that companies can opt out of. Against this backdrop, the Columbia research establishes a floor for the scale of invisible displacement: even if official data showed no AI-driven unemployment, three-quarters of those displaced would never appear in the record at all.