Demand Side I: Labor Markets, Productivity, and Automation
POLISCI 158 · AI & Politics · Session 2 · Mon, Sep 28, 2026
Fact 1: no economy-wide displacement
Nov 2022 to Jun 2026: ADP sample about +6%, most-exposed quintile about +4%. Figure 1, Brynjolfsson, Chandar & Chen (Aug 2026), ADP data; exposure from Eloundou et al. (2024).
Fact 2: young workers in exposed jobs
Ages 22–25, two most-exposed quintiles: about −11% (Nov 2022 to Jun 2026); three least-exposed: about +10%. That is 19% below where they would be had they kept pace. Figure 3 (two of six panels), Brynjolfsson, Chandar & Chen (Aug 2026).
Fact 5: automation vs. complementarity
- Where AI automates tasks: young workers’ employment falls
- Where AI complements workers: no clear ordering
Ages 22–25. Figure 5, panels B–C, Brynjolfsson, Chandar & Chen (Aug 2026); usage measures from the Anthropic Economic Index (Handa et al., 2025).
Tracker: no statewide surge
Figure 2, Hyman, von Wachter, Herrera, Moghadam, Morris, Motghare & Segal, California Policy Lab (June 2026). Claims through May 2026.
Tracker: college-educated, high-exposure claims up
Figure 3, California Policy Lab (June 2026). Claims indexed to Nov 2022 = 100; exposure from Eloundou et al. (2024).
Tracker: the Bay Area
Figure 4, California Policy Lab (June 2026). Observed exposure from the Anthropic Economic Index (Handa et al., 2025).
Measuring exposure: easier said…
Indexes disagree more about the most exposed jobs.
Figure 1, Gimbel, Kendall & Kulsakdinun, “Labor Market AI Exposure: What Do We Know?”, The Budget Lab at Yale (Feb 19, 2026).
Okay, then maybe we measure automation?
Headlines as shown in the 157 deck (summer 2026): Reuters (Jun 22, 2026); Wall Street Journal (May 5, 2026); New York Times.
But can we trust these announcements?
JobsLost.ai (jobslost.ai), screenshot as used in 157 (mid-2026). Compiled from company announcements and news reports; on Sep 27, 2026 it listed 203 layoffs and 265,086 jobs.
On the other side: will productivity pay off?
Economic Policy Institute, “The Productivity–Pay Gap” (chart as captured in 157, data to 2025; page updated Sep 14, 2026 with data to 2026). epi.org/productivity-pay-gap
The China shock: 2 million jobs in 12 years
Figure 6b, Autor, Dorn & Hanson (2016), “The China Shock,” Annual Review of Economics. Job losses 1999–2011 (2.0 million economy-wide, input-output linkages) from Acemoglu et al. (2016) as summarized there.
The China shock: little compensation
Per $1,000 of import exposure per worker:
- −$549 household wage and salary income per working-age adult
- +$58 government transfers per person
- only $3.65 of it via UI and TAA
Autor, Dorn & Hanson (2016), Table 4 cols. 7–8 and Figure 7; annual, commuting zones 1990–2007.
What does it take? Mass sacrifice = mass compensation?
Tax Policy Center, Historical Highest Marginal Income Tax Rates (1913–2023); Scheve & Stasavage (2016), Taxing the Rich, Princeton University Press.
What about organized labor?
Writers Guild of America West, “Artificial Intelligence: Know Your Rights” (updated Dec 18, 2025). wga.org
What about organized labor? Approval is high
Gallup, Megan Brenan, “Labor Union Approval Relatively Steady at 68% in U.S.” (Aug 28, 2025).
What about organized labor? Membership is low
Figure 1, Bauer & Cole, “Six Facts About Union Density in the United States,” The Hamilton Project (May 22, 2026).
California’s AI-related job loss tracker
Office of the Governor of California, press release, June 25, 2026 (as shown in 157). Tracker built by the California Policy Lab with the Employment Development Department.