Paper: AI layoffs trap demand
Researchers at UPenn and Boston University published a paper called 'The AI Layoff Trap' arguing that widespread AI-driven layoffs can destroy consumer demand and create a Prisoner’s Dilemma where firms can't stop automating. The brief summary also cites a finding that roughly 80% of U.S. jobs contain tasks susceptible to automation. (x.com)
A new economics paper argues companies can automate themselves into a demand slump: cutting workers with artificial intelligence also cuts customers with paychecks. (arxiv.org) “The AI Layoff Trap,” posted to arXiv on March 21, 2026, was written by Brett Hemenway Falk of the University of Pennsylvania and Gerry Tsoukalas of Boston University. The paper says firms acting rationally in competition can still automate “well beyond what is collectively optimal.” (arxiv.org) The mechanism is simple: one company saves money by replacing tasks with software, but the lost wages reduce spending across the economy. The authors model that spillover as a demand externality, meaning each firm ignores part of the damage because it lands on everyone else. (arxiv.org) That produces a Prisoner’s Dilemma, the paper says, where every firm has an incentive to automate even when all firms would be better off with less automation. The authors write that more competition and “better” artificial intelligence make the excess automation worse, not better. (arxiv.org) The paper arrives as economists and companies are still arguing over whether generative artificial intelligence mainly changes tasks inside jobs or eliminates whole jobs. An earlier OpenAI-led study found about 80% of the United States workforce could have at least 10% of work tasks affected by large language models, while about 19% could see at least half their tasks affected. (openai.com) That 80% figure is often cited as a layoff forecast, but the 2023 paper did not make adoption or job-loss predictions. It measured task exposure, which is closer to saying software can do pieces of a job than saying a company will remove the job entirely. (arxiv.org) Falk and Tsoukalas say several popular fixes do not solve the trap in their model. Their abstract says wage adjustments, free entry, capital income taxes, worker equity participation, universal basic income, upskilling, and Coasian bargaining all fail to eliminate the problem. (arxiv.org) The policy result in the paper is narrower than the social-media version of the story. The authors say only a Pigouvian automation tax — a tax meant to price in the broader social cost — can stop firms from over-automating in their framework. (arxiv.org) The paper is a theoretical model, not a measurement of layoffs that have already happened. Its claim is that if artificial intelligence displaces workers faster than the economy can reabsorb them, market competition alone does not supply a brake. (arxiv.org)
