Chicago Fed President Austan Goolsbee on why the data won, why the vibes stay bad, and what the AI build-out means for the next rate cycle.
In a wide-ranging conversation recorded at the Chicago Fed, Kyla Scanlon sat down with Chicago Fed President Austan Goolsbee1 to revisit a warning he had issued a year prior: that the economy might be heading back into a collision between what the numbers show and how people feel. One year on, his verdict is unambiguous. The data won. The vibes did not follow.
Goolsbee describes the standoff in characteristically candid terms: "The family fight continues. It never stopped. The argument began at Thanksgiving and it extended to Christmas and into the New Year." The Fed, he says, set interest rate policy over the intervening period on the basis of hard numbers. Consumer sentiment remained, in his word, "horrible." The two trajectories never converged.
Sentiment as a Forecasting Tool Is Losing Its Job
The more consequential development, Goolsbee argues, is not that people feel bad but that their feelings have become a progressively weaker predictor of what they actually do. "The correlation between sentiment and spending has been very, very weak," he says. For decades, sentiment was a standard Fed input precisely because it tended to foreshadow consumer behaviour. That relationship, he notes, "didn't just break down recently. It's been getting a little weaker, a little weaker over time."
"Vibes are important for a lot of things. But if you were using the vibes as a predictive variable for consumer spending, which was historically kind of how the Fed was using it, it makes you a little less certain. You're never going to hear me say throw any data away. But it moves it down on my priority list." — Austan Goolsbee
In its place, Goolsbee has restructured what he watches. On the labour side: the unemployment rate, the vacancy rate, the layoff rate, and the hiring rate. On the consumer side: retail spending, derived from hard numbers rather than surveys. Sentiment may still carry information about expectations and relative positioning, but as a forecasting variable for spending, its incremental value "doesn't seem to be that high" right now.
Why People Feel So Bad
Goolsbee identifies three interlocking causes. The first is salience: certain prices, particularly gasoline and groceries, are publicly visible in a way that CPI figures are not. "If you're driving down the street and you're seeing diesel is $6 a gallon, people are going to say they think the economy stinks." The second is the politicization of economic perception. Even factual questions about the unemployment rate, he observes, flip based on the respondent's politics and the politics of the sitting president. The third is episodes where real incomes fell. "When people see prices going up faster than incomes, it's going to blow back."
Scanlon, who coined the term "vibecession," pushes further, pointing to social media as an amplification mechanism. The visibility effect Goolsbee attributes to gas stations, she argues, now operates across the entire economy through algorithmic feeds. Goolsbee finds the observation pointed: "The publicizing of the price is what makes people more attuned to the cost of living."
As Scanlon's panel colleague Brian Fabes summarized: "People live in levels, not rates." CPI inflation may have moderated, but the price level itself is approximately 30 percent above January 2020. That is the number people are responding to.
One Vote, Not One Dollar
A structural tension runs beneath the sentiment divergence: GDP and the stock market are dollar-weighted measures, while consumer surveys are person-weighted. Goolsbee illustrates the point starkly. If a tiny fraction of households captured all the income gains, survey-based measures would look poor even as aggregate output grew. New York Fed data suggest the K-shaped divergence in consumption exists but is less extreme than the stock market's performance might imply, running to roughly a one percentage point gap between higher- and lower-income groups.
Low Hire, Low Fire: Year Two
The labour market continues to exhibit a configuration Goolsbee first flagged a year ago. "The hiring rate is as low as it normally is in a recession, and the layoff rate is as low as it normally is in a boom. And both of them are stable." Initially attributed to uncertainty, the persistence is becoming harder to explain that way. "If we go for five years, it can't really be uncertainty anymore." His read, for now, is stability rather than deterioration. One complicating signal: businesspeople across the Seventh District report that the AI data centre build-out is tightening local labour markets for skilled workers, driving wages above what employers had budgeted even six months ago.
The AI Build-Out: Demand Shock or Category Error?
On the question of AI's macroeconomic implications, Goolsbee draws a line between two scenarios. To the extent the build-out displaces spending from other categories, the aggregate temperature of the economy is not overheated. But to the extent it "jumps out of its lane" and generates excess demand across the broader economy, it becomes a conventional demand shock. "That's not a supply shock. That's just a demand shock. And the theory is pretty clear on what the central bank has to do."
"Nobody spends a trillion and a half dollars thinking they're going to give the product away free. So either it's going to be really expensive, or the valuations are going to take a hit, in a fundamental way." — Austan Goolsbee
His message is unambiguous: if the AI cycle produces a genuine demand overheat that bleeds into the broader economy, the central bank response is a standard one. AI is arriving at the Fed as a demand story, through construction activity and wage pressure, before it arrives as a productivity story. The question of whether the productivity gains materialize at the scale implied by current valuations he leaves open, though the analogy to 1999 to 2001 hangs in the air.
5 Key Takeaways for Advisors and Investors
1 Sentiment is no longer a reliable spending signal. The Fed has effectively downgraded consumer sentiment as a forecasting variable for consumption. Client anxiety about the economy does not reliably predict spending contraction or recession. Hard data, particularly retail figures and labour market rates, are now the more actionable inputs.
2 Price levels, not inflation rates, are what clients are feeling. The 30 percent cumulative price-level increase since January 2020 is the operative client reality. Advisors communicating about "cooling inflation" are speaking in rates; clients are hearing in levels. Closing that translation gap is a practical communication challenge.
3 The labour market is stable, not deteriorating. Goolsbee reads the current hiring-and-layoff configuration as pointing to stability. Recession risk from the labour market, in his framing, is not the current concern. Portfolios positioned for imminent labour market deterioration may be misreading the signal.
4 The AI build-out is a demand story first. Before any productivity tailwind materializes, AI capital expenditure is generating wage pressure and local economic heat. Investors in sectors exposed to data centre construction or skilled-labour competition should be tracking this dynamic. Overvaluation risk is live if productivity gains disappoint.
5 The Fed has tools and will use them. If the AI cycle produces a genuine demand overheat that bleeds into the broader economy, the central bank response is a standard one. Rate policy remains a live instrument. Advisors should not assume AI-driven growth translates to a permanently accommodative monetary environment.
Footnote:
1 Scanlon, Kyla. "Chicago Fed President Austan Goolsbee on Vibes vs Data." Kyla's Newsletter, Substack, 30 Sept. 2026, https://kyla.substack.com/p/chicago-fed-president-austan-goolsbee-b24.
