Noah Smith
· Noahpinion
· July 01, 2026 at 06:31
· ⏱ 23 min read
| Read on Substack ↗
Summary
The 'vibecession' may be an artifact of flawed University of Michigan survey methodology, not real consumer distress. AI is more likely to depolarize than polarize, and generative AI is currently complementing rather than replacing human labor. Millennials are outperforming prior generations in after-tax income, and US deindustrialization is driven more by shifting consumer demand than offshoring.
•University of Michigan consumer sentiment hit all-time low but shift to online polling and oversampling Democrats may explain the drop; adjusted data aligns with 2013 levels, per Joel Wertheimer.
•A 2026 Conlon & Schwardmann experiment with 1,500 participants found AI advice depolarizes choices despite sycophancy, due to substantive information.
•Chad Jones' 2023 paper on AI existential risk sets extinction utility to zero, which understates infinite loss implied by log utility and may distort risk-benefit calculations.
•Corinth & Larrimore (2026) show Millennials at age 36-40 have fewer incomes below $30k and more above $40k than prior generations; within-generation inequality has increased.
•Richard Baldwin's decomposition finds 70-80% of US deindustrialization since 1995 came from falling demand for manufactured goods, not outsourcing.
•Russia's oil revenue windfall from the Iran war was small due to a stronger ruble, tax deductions for fuel subsidies, and declining crude production.
•Rebecca Diamond's study on GLP-1 drugs found single women 29% more likely to marry/cohabit and jobless women 27% more likely to find work within three years.
•Kharazian, Simon, and Stevens (2026) found that companies adopting generative AI increased total hiring and entry-level hiring, contrary to substitution fears.