Martha Gimbel on the Impact of AI and the Trade War on Labor Markets

Episode Summary

Executive Summary: Martha Gimbel argues that AI’s labor-market impact is real but still early and overstated, with current evidence showing little broad disruption since ChatGPT’s release. She contrasts AI’s gradual, productivity-boosting effects with the clearer but delayed drag from tariffs, warns that weak public data hampers policymaking, and stresses that disruption is manageable only if policy, firms, and workers adapt proactively.

Main Topics: Budget Lab mission and long-horizon policy analysis (Priority: 5/5): Gimbel explains that the Budget Lab was created to provide rapid, rigorous analysis of fiscal and economic policies, often over a 30-year horizon, because near-term budget windows can miss major long-run effects of debt, demographics, and investment. Data limitations and the need for stronger statistical infrastructure (Priority: 5/5): The conversation emphasizes how missing government data leaves policymakers and analysts 'flying blind,' why private-sector data are useful but incomplete, and why statistical agencies need more funding, staff, and new measurement tools. AI and the labor market: early evidence vs hype (Priority: 5/5): Gimbel argues that despite intense commentary, there is little evidence yet of broad labor-market disruption from generative AI. She says technology adoption takes time, and observed labor shifts since ChatGPT are mostly flat or noisy. Young workers, college graduates, and exposure to AI (Priority: 4/5): The discussion focuses on whether recent college graduates and other highly exposed workers are seeing job-market harm. Gimbel says the evidence is weak and that any recent weakness likely reflects the broader macro cycle more than AI alone. Tariffs, uncertainty, and delayed economic effects (Priority: 4/5): Gimbel says tariffs slow growth and raise prices, but their effects take time to show up. The biggest immediate threat from trade policy was uncertainty, though strength in AI-related sectors has helped offset some tariff drag. Adaptation, productivity, and policy response to disruption (Priority: 4/5): She frames AI as a productivity shock that can free workers for higher-value tasks, but warns that retraining systems and unemployment policy are not well designed to handle rapid labor-market transitions. Technology, creativity, and long-run historical parallels (Priority: 3/5): The discussion uses historical analogies—like elevator operators and the Industrial Revolution—to show that technology destroys some jobs, creates others, and often benefits later generations more than those directly displaced.

Key Arguments: The Budget Lab was built because inside government, budget and policy estimates can determine which proposals survive, and long-run analysis often reveals effects that short-run windows miss. Public-sector data remain the best source for understanding the economy; private-sector data are valuable but biased by selection effects, incentives, and incomplete coverage. The current lack of official labor-market and inflation data makes it unusually hard to assess whether the economy is heading toward recession, slowdown, or soft landing. Generative AI has existed for only about three years, which is too soon for it to have transformed the labor market at scale; adoption and workflow changes take time. Gimbel’s research finds little evidence of meaningful labor-market shifts since ChatGPT: occupational mix, unemployment duration, and employment patterns are mostly flat. Some studies show small declines in employment for very young workers in highly AI-exposed jobs, but those results may reflect the broader macro slowdown rather than AI itself. Tariff effects are real but delayed; economists were especially worried by the uncertainty and policy volatility surrounding tariff announcements, not just tariff levels themselves. AI-driven growth is helping the U.S. economy and stock market more than tariffed sectors are hurting them, which helps explain why the macro impact has been less severe than some feared. The biggest downside risk from AI is not labor displacement alone but national security and biohazard concerns; if the worst effect is labor disruption, that is comparatively manageable. Governments are not especially good at picking which jobs will emerge from technological change or at scaling retraining programs, so policy should be flexible rather than overly specific.

Data Points: Budget Lab launch: April 2024 - Gimbel says the Budget Lab at Yale was launched in April 2024 by her, Danny Yagan, and Natasha Sarin. AI labor shock timeline: ~3 years - She notes that generative AI has only been around for about three years, which is too short for large labor-market effects to fully materialize. Budget window: 30 years - The Budget Lab often analyzes policies over 30 years rather than the usual 10-year budget window. Standard budget window: 10 years - She contrasts the 30-year horizon with the usual 10-year budget window used in federal analysis. Tariff impact on growth: ~0.5 percentage point lower by end of 2025 - She says Budget Lab estimated growth would be about half a percentage point lower by the end of 2025 under the liberation-day tariffs. GDP effect from arms to Ukraine: ~20 basis points - She says adding arms to Ukraine at points in 2022 added about 20 basis points to quarterly annualized GDP, though that was not the reason for the policy. Challenger data claim: AI is 8x as responsible as tariffs for layoffs - She cites Challenger layoff data but says the attribution is likely unreliable and not something economists would accept at face value. Treasury market concern: April 2024/2025 market stress - She references market turmoil in April when tariffs and policy uncertainty shook stock and Treasury markets. Recent college graduates: Small, noisy gap emerging since ChatGPT - She says recent college grads in the data show a small but noisy divergence relative to slightly older grads. Executive assistants decline: About 4 million to 500,000 - She gives this as an approximate example of how AI/automation-like technologies reduced administrative jobs over time. Data quality note: Private-sector labor data are not comprehensive - She explains that private data like Indeed or Challenger are incomplete and benchmarked to BLS data. Public data coverage: BLS/Census data are the benchmark - She argues official federal statistics are still the best basis for macroeconomic and labor-market analysis.

Pivotal Quotes: "I am a technology-paced adaptation skeptic." — Martha Gimbel: Her characterization of her position on AI: not skeptical of AI itself, but skeptical that adoption and labor-market adaptation happen instantly. "The biggest downside risk from AI is much more on the national security, biohazards, things like that." — Martha Gimbel: She distinguishes labor-market concerns from more serious AI risks that worry her most. "What people did with electricity changed the labor market." — Martha Gimbel: She uses this analogy to argue that technology matters through adoption and organizational change, not just invention.

Implications: AI is likely to boost productivity before it causes broad displacement, but the transition could still be painful for some workers. Policymakers need better data, flexible safety nets, and proactive firm-level adjustment rather than rigid retraining plans.