Seventy percent of Americans believe AI will eliminate more jobs than it creates. That single statistic, from a Quinnipiac University poll released March 30, 2026, is the load-bearing fact in Robert Shiller's June 22, 2026 New York Times op-ed "This Doommaxxing Has Got to Stop".
The Nobel laureate's claim is not that AI is harmless. His claim is that the story we tell about AI harm is itself becoming the harm.
AI doommaxxing, in Shiller's framing, is the runaway cultural narrative that artificial intelligence will wipe out vast employment. Through the lens of narrative economics, his body of work on how shared stories drive economic behavior, that story can become self-fulfilling. When enough people defer hiring, cancel training, and hoard savings out of fear, they manufacture the slowdown they were afraid of. The Quinnipiac 70% gap is the evidence that the story has reached the cultural density required to move real decisions.
TL;DR
- 70% of Americans expect AI to destroy more jobs than it creates, per Quinnipiac's March 2026 poll. Personal worry rose from 21% (April 2025) to 30% in one year.
- Behavior is already shifting. Computer science enrollments fell 8.1% in 2025-2026 while mechanical engineering rose 11% and electrical engineering 14%.
- Aggregate job losses have not arrived. BLS May 2026 payrolls added 172,000 jobs versus expectations near 80,000.
- A credible dissenter exists. Nobel laureate Daron Acemoglu projects AI productivity gains of just 0.55% over a decade, far below Goldman Sachs' 7% forecast.
- The cultural amplifiers are running. Margaret Atwood, Jeff Bridges, and SAG-AFTRA are personalizing the anxiety for non-technical audiences.
What is the Shiller narrative-economics thesis on AI jobs?
Shiller's mechanism is a feedback loop. Widespread belief that AI will eliminate jobs leads individuals and firms to behave as if that future has already arrived. They freeze hiring, postpone training, cut consumption, and raise precautionary savings. Those aggregate behaviors compound into actual economic deceleration. The fear births the reality.
This is not a new idea for Shiller. His narrative economics research has long held that economic events are driven by stories as much as by fundamentals, and that stories "go viral" and reach "critical narrative density" before triggering real-world consequences.
He invokes the Luddites who smashed machinery in early industrial England, Karel Čapek's 1920 play R.U.R. that coined the word "robot," and the post-1929 crash where collapsing confidence reinforced decline.
What makes the AI case different, in Shiller's view, is speed. Technological unemployment fears have materialized before, during the Agricultural and Industrial Revolutions, but the potential pace of AI displacement creates a qualitatively different dynamic. He positions mid-2026 as a juncture where the narrative could tip toward productive adaptation or self-defeating panic.
The four channels he identifies are concrete: deferred hiring, cancelled training and education, precautionary savings that weakens consumer demand, and capital misallocation toward AI capabilities at the expense of human talent. Each is individually rational. Each, aggregated, is contractionary.
How widespread is AI job anxiety in 2026?
The polling is consistent across independent survey houses. Quinnipiac's March 2026 release is the headline number: 70% of Americans believe AI will eliminate more jobs than it creates, with 30% personally worried about losing their own job, up from 21% in Quinnipiac's April 2025 survey.
That 9-point jump in personal anxiety inside one year signals accelerating concern, not stable awareness.
Trust in the technology itself is low. The same Quinnipiac release found 76% of Americans trust AI only "rarely" or "sometimes" to do what it claims, and just 21% trust it "most of the time" or "all of the time."
Gen Z respondents showed the highest pessimism about AI's job impact, a notable detail given that younger workers are furthest from the labor market's incumbent protections.
Other pollsters confirm the pattern. Pew Research Center found majorities concerned about AI's effect on their personal financial situation. The Los Angeles Times reported more than half of Americans believe AI is likely to harm them personally. Gallup documented "real concerns" across multiple dimensions including job displacement.
The cross-pollster consistency rules out the artifact-of-wording critique.
Is AI anxiety already changing real decisions?
The most concrete behavioral evidence comes from higher education. National Student Clearinghouse Research Center data for the 2025-2026 academic year, reported via the Washington Post, shows computer science enrollments at four-year universities declined 8.1%, with CS as a single major dropping 11.2%. That reverses a decade-long post-2008 surge.
Simultaneously, fields perceived as AI-resilient grew. Mechanical engineering enrollment rose 11%, electrical engineering 14%, and overall engineering programs 7.3%. BSN nursing programs grew 4.9%, adding 12,434 students, per the American Association of Colleges of Nursing.
The Deloitte 2026 Higher Education Trends report frames this as a "career recalibration" toward physical trades, healthcare, and specialized engineering.
| Field (2025-2026 enrollment) | Change |
|---|---|
| Computer science (all) | -8.1% |
| Computer science (single major) | -11.2% |
| Mechanical engineering | +11.0% |
| Electrical engineering | +14.0% |
| Overall engineering | +7.3% |
| BSN nursing | +4.9% |
Whether these are rational adaptations or anxiety-driven misallocation is contested. The behavioral signal, though, is unambiguous: students are voting with their majors.
At the corporate level, the evidence is more circumstantial. Microsoft CEO Satya Nadella, in a June 21, 2026 Wall Street Journal interview, named the contradiction directly: "You can't say, hey, all white-collar jobs are gone and this could even be a weapon and we will use all the power to build data centers."
His comment, widely circulated, captures the cognitive dissonance in vendor messaging. Companies are simultaneously warning about disruption and pouring billions into the infrastructure that supposedly causes it.
A specific mid-2026 phenomenon is "entry-level compression." If AI tools can draft documents, enter data, and run basic analysis, organizations rationally reduce junior hiring while keeping senior headcount. Stanford HAI's 2026 AI Index documents the pattern, and the Stanford Digital Economy Lab's Canaries project tracks leading indicators.
Total employment can hold steady while career-entry onramps narrow, which intensifies anxiety even without aggregate job loss.
How is the AI fear narrative being amplified culturally?
Shiller's framework requires stories to achieve critical cultural density. Late June 2026 delivered several amplification events that personalized the anxiety for non-technical audiences.
Margaret Atwood, author of The Handmaid's Tale, appeared at the Babel literary festival on June 27, 2026. She recounted using Anthropic's Claude once to verify a Father Brown mystery spoiler, only to find it hallucinated the answer.
"The thing about AI is that it's garbage in, garbage out," she said, in coverage that spread from Deadline to The Verge. A respected literary authority endorsing skepticism carries cultural weight that technical critiques from computer scientists do not.
Jeff Bridges discussed Suno AI, the music generation platform, on Theo Von's podcast "This Past Weekend" around June 26-27, 2026. "AI is, it's frightening, man. It's very frightening, but it's an amalgamation of all our wisdom, our soul, our things."
His ambivalence, frightening yet representing collective human wisdom, captures a mood that is neither pure technophobia nor techno-optimism. The interview reached older Americans and creative professionals who rarely encounter AI through technical channels.
The entertainment industry keeps the narrative in circulation. SAG-AFTRA's September 30, 2025 agreement with major studios set frameworks for AI consent and compensation, but performers have continued to flag its limits. Music-industry disputes over AI voice cloning and generation, including the Sludge Pump and Slipknot controversy, keep creative-labor anxiety in the news cycle.
These moments matter for the self-fulfilling prophecy thesis because narrative economics emphasizes that stories spread through social networks. When Atwood, Bridges, musicians, and actors publicly express anxiety, the narrative achieves the personal salience that pure economic statistics cannot.
Does the labor market data actually show AI displacement?
Here is where Shiller's thesis gets uncomfortable. The Bureau of Labor Statistics Employment Situation release for May 2026 showed payrolls adding 172,000 jobs against expectations of 80,000-85,000, a significant upside surprise. OECD unemployment data updated the same month kept rates near historical lows. GDP, consumer spending, and business investment continued expanding.
That creates an obvious tension with the 70% Quinnipiac finding. A large majority believes AI will eliminate more jobs than it creates, yet current data shows continued job growth. Either the anxiety is anticipating displacement that has not yet materialized, or it is disproportionate to current reality.
Goldman Sachs offers a measured middle ground. Their widely cited analysis suggested up to 300 million jobs could be affected globally, but Jan Hatzius's research emphasizes the gap between AI's potential to automate tasks and the actual pace of implementation, which is slowed by integration costs, organizational resistance, and regulation.
Goldman HR leader Jacqueline Arthur, in MIT Sloan Management Review, described an augmentation-first approach to workforce planning for most roles.
What is the strongest counterargument to Shiller?
The most credible dissent comes from MIT economist Daron Acemoglu, who shared the 2024 Nobel Prize in Economics. Acemoglu has argued that the productivity effects of current AI technologies are likely modest, projecting potential total factor productivity gains of no more than 0.55% over a decade.
That stands in stark contrast to Goldman's 7% forecast or McKinsey's 3-4 percentage point estimate, and is documented in Brookings coverage of his work.
Acemoglu's core argument is that automation gains require not just technological capability but economic and organizational adaptation: new business models, new regulatory frameworks, new social institutions. The historical record shows that while specific jobs disappear, aggregate employment stays stable because new tasks and industries emerge to absorb displaced workers.
His position is not that AI has no effect. It is that the effect will be gradual, unevenly distributed, and dependent on institutional responses. That is a meaningful challenge to Shiller's framing of mid-2026 as a tipping point where panic could swing the outcome.
Stanford's Canaries Dashboard supports a nuanced read. Certain occupations show AI-related disruptions, but aggregate labor market signals remain consistent with expansion. The Stanford AI Index 2026 public opinion chapter notes that expressed concern about AI job losses significantly outpaces measurable job losses attributable to AI.
That gap could reflect rational anticipation, or anxiety running ahead of reality.
What this means for you
If you are hiring, building, or planning a career in mid-2026, the practical signal is this: the fear is real and measurable, the displacement is not yet, and the gap between the two is where decisions get made poorly.
For operators, three moves follow from the evidence. First, separate aggregate anxiety from sector-specific exposure. Entry-level knowledge work shows compression signals; healthcare and skilled trades show enrollment growth.
Calibrate hiring plans to your actual task profile, not the headline panic. Second, watch the Canaries indicators for your occupation rather than the quarterly BLS headline, which lags.
Third, name the narrative dynamic inside your own organization. If your leadership team is simultaneously freezing headcount and doubling the AI infrastructure budget, you are running Shiller's feedback loop on yourself.
For individuals, the enrollment data already shows the rational response: gravitate toward work that combines physical presence, regulated judgment, or specialized engineering depth. The mistake is treating the 70% consensus as a forecast. It is a story. Stories can become true, but only if enough people act as if they already are.
Conclusion
Shiller has identified a real mechanism and the polling confirms it has reached cultural scale. The 70% Quinnipiac number, the enrollment shifts, and the cultural amplification from Atwood and Bridges all show the narrative operating.
What the thesis cannot yet claim is that the narrative has produced aggregate economic damage. May 2026 payrolls grew more than double expectations, unemployment sits near lows, and Acemoglu's productivity skepticism offers a credible floor under the doom case.
The honest read is that Shiller is half-proven: the fear is real and reshaping behavior, but the self-fulfilling contraction has not arrived. Whether it does depends less on the models than on the next hiring decisions, training budgets, and savings rates that the fear itself is shaping.
Sources
- Robert Shiller, "This Doommaxxing Has Got to Stop," New York Times, June 22, 2026
- Robert J. Shiller, Yale Insights
- Quinnipiac University Poll, "The Age of Artificial Intelligence," March 30, 2026
- Quinnipiac University Poll, April 2025 release
- Pew Research Center, "Key findings about how Americans view AI," March 2026
- Los Angeles Times, "More than half of the U.S. Says AI is likely to harm them," March 31, 2026
- Gallup, "Americans Express Real Concerns About Artificial Intelligence"
- Deloitte, "2026 Higher Education Trends"
- Satya Nadella interview, Wall Street Journal / MSN, June 21, 2026
- New York Times Dealbook, "The Growing Anxiety Over AI, Jobs and the Future," May 20, 2026
- Stanford HAI, 2026 AI Index Report, Public Opinion
- Stanford Digital Economy Lab, Canaries Dashboard
- BLS Employment Situation Summary, May 2026
- OECD Unemployment Rates, Updated May 2026
- MIT Sloan Management Review, "How Goldman Sachs Stays Agile: HR Leader Jacqueline Arthur"
- Brookings, "AI and economic mobility: Opportunities and challenges" (Acemoglu)