In September 2026, the University of Michigan's Index of Consumer Sentiment crashed to 47.8 — worse than at the bottom of the Great Recession. Inflation ticking, gas at nearly four dollars a gallon, the Iranian conflict still fresh in memory. It all seems to add up: people are scared — so wallets should shut. But three days later the Commerce Department reports: retail sales in August jumped 1.2% — the strongest growth since March, the best core retail reading in two years. And Goldman Sachs economist Joseph Briggs writes to clients with an almost heretical thought: sentiment no longer reflects the economy because it reflects something else. "Low levels of economic sentiment likely reflect a more fundamentally gloomy assessment of the state of the world, not the economy," he writes, pointing to the decline in plain human happiness: the share of "very happy" Americans fell from 31% in 2016 to 23% in 2024.
This is not one banker's private quirk. Behind that line lies a half-century history of economic science mixed with psychology, war, and one naive idea: that before big purchases, people first "feel." The story of how the economic barometer became a terrarium of political passions and media noise — and also about who now measures what the barometer no longer measures.
What caught me was a morning note in the digest: Goldman Sachs explained the record-sour September sentiment index not by the economy but by "declining happiness" — as a separate, independent variable. Sounds like a joke, but behind it stands a serious shift: one of the pillars of market analytics — the index of consumer sentiment — has been officially declared an unreliable predictor by Goldman Sachs itself.
Why this topic: sentiment indices were considered "leading indicators" for half a century — central banks, retailers, and marketers believed in them. If the barometer needle really has detached from the economy, this is not a financial detail but a breakdown of the instrument itself, against which retail dashboards and central bank rhetoric still calibrate forecasts. And this is a rare case where the "broken dashboard" has a precise birth date, a culprit, and even a patch.
The story begins not with financiers but with a psychologist. George Katona — a Hungarian from Budapest, doctor of experimental psychology from Göttingen (1921), a Gestaltist by training in learning and memory. In 1933, with the Nazis' arrival, he emigrates to the US, and during the war the government hires him: to fight wartime inflation not only with prices and rationing but with psychology — his book is titled exactly that, "War without Inflation" (1942). That's when he notices something economists of the time didn't see: people spend not because they "should according to the budget equation" but because they have both the ability and the willingness to buy.
After the war, Katona founds the Economic Behavior Program at Michigan — 1947, Survey Research Center. His team begins systematic household surveys: about their own finances, about business conditions, about purchase plans. In 1952, from these answers the Index of Consumer Sentiment is assembled — a "barometer" of consumer behavior that later enters the official Composite Index of Leading Indicators of the Commerce Department. The methodology quickly spreads to Europe: from 1972, the European Commission conducts a similar survey across the EEC.
And here's the drama of the index's birth. In mid-1945, press headlines predict "eight million unemployed by 1946": economists expect a repeat of the Great Depression — 12 million demobilized, factories without orders. Nobel laureate Lawrence Klein later admits where the forecast failed: "in predicting consumer spending… the magnitude of error is such that it leads to catastrophic recommendations." Reality: 3 million unemployed, and instead of deflation — wild inflation. People who had accumulated currency under the mattress during the war (well, war bonds and cash) come out to market like it's a celebration: the savings rate collapses from 11.7% at the start of 1946 to 2.2% by mid-1947 — a minimum not repeated for half a century.
Katona with his surveys explained why: consumers had not economics but psychology — accumulated savings, pent-up demand, thirst for normal life. The index that caught this "willingness" predicted a boom when models predicted a bust. That's how economic science got a fifth pillar: not income, not prices, not rates, but sentiment.
What exactly this pillar measured is amusingly minimalist. The famous Michigan index is built from answers to five questions: are your finances good now, will they be better in a year, is now a good time for major purchases, will it be better in a year, and is the economy growing. The Conference Board since 1967 surveys about current conditions and expectations six months out. On this foundation stands an entire industry: since 1952 the index is published monthly, traded in the news, moves stock prices on Friday morning-date releases and enters handbooks of "leading indicators."
Why believe it at all? Because for decades it honestly delivered. Federal Reserve banks recorded: through 2020, annual correlations of the sentiment index with growth in real personal consumption expenditures (PCE — nearly 68% of US GDP) held at an average of 0.69 for Michigan and 0.60 for Conference Board. In 2014, Boston Fed economists even showed: expanding the "five questions" to 42 gives additional predictive power beyond fundamental models and even beyond professional forecasters' adjustments. Sentiment was not fortune-telling — it was a signal.
The logic was simple and elegant in Katona's spirit: a consumer with money and good mood will buy a car, a consumer in panic will postpone. Attitudes are what turn "can buy" into "will buy," and it was precisely this distinction Katona laid in the foundation of behavioral economics decades before Kahneman: economics studies the rationality of the result, psychology — the rationality of the process, and the sentiment index was the bridge between them.
And then the bridge split apart.
Formally, the new era was christened by a meme: in summer 2022, fininfluencer Kyla Scanlon coined the word vibecession — "recession vibes, but not the actual recession yet." Inflation at a 40-year high, sentiment hitting historic lows, but employment at records, GDP growing, wages creeping up. Americans in an Affirm survey (June 2024) confidently stated the country had been in recession since March 2023 — with good growth and cooling prices. The survey showed depression, cash registers showed a party.
The Chicago Fed in a recent letter (Chicago Fed Letter #521, 2026) measured this breakdown numerically, and the numbers are grim: the 120-month rolling correlation of the Michigan index with real consumer spending growth after 2020 dropped almost to zero — from historic 0.69. For Conference Board — 0.35, worse than any moment in forty years. On the horizon of a single quarterly forecast model, the Kansas City Fed shows the same: over the last 30 years, adding sentiment to a model of fundamental data practically does not improve the spending forecast — difference within 0.5 percentage points, statistical noise.
Funniest of all, the split goes both ways. On one hand — "gloomy but spending": the consumer is selective, economical, and yet continues to buy, relying on asset value growth and accumulated savings (Reuters, September 2026: retail +1.2% for the month with sentiment at 47.8). On the other — the phenomenon of doom spending: 27% of Americans (per Credit Karma survey) spend money to cope with anxiety, and among Gen Z — 37%. Bad vibes now don't reduce but increase spending — this is an inversion of all of Katona's logic, under which sentiment was supposed to work as a brake.
And don't forget the political dimension. The partisan gap in sentiment is a veteran: Democrats and Republicans equally love the economy under "their" White House and hate it under the other's. After the 2016 election, the Michigan expectations gap reached 74.6 points in six months. But the key finding in this area is Mian, Sufi, and Khoshkhou's study (Review of Economics and Statistics): after the 2008 and 2016 elections, people's stated expectations shifted by party affiliation, while administrative spending data — not at all. That is, a significant part of "sentiment" is political projection, decoration, not behavior.
Now about what Goldman found. Briggs looks not at surveys about the economy but at surveys about life: the General Social Survey (NORC, University of Chicago), which since 1972 has been asking Americans a simple question — how happy are you. The answer is troubling: the share of "very happy" fell from 31% (2016) to 23% (2024), the share of "not too happy" rose from 13% to 20%. The pandemic knocked out happiness, and it never grew back — while financial satisfaction in the same survey fell more weakly. Happiness decoupled from the wallet.
Briggs searches for the cause and finds it in trust in institutions: precisely that, by his analysis, provided a "disproportionate share" of the decline in overall happiness. This aligns with a large literature: Brady and Kent in Dædalus show that over fifty years, confidence in almost all American institutions has collapsed (except the military and science), and the decline is polarized — Republicans trust business and police, Democrats — press and universities, and trust in "other" institutions is destroyed symmetrically. The Guardian in a fresh poll records: 95% of Americans believe the country is in an "affordability crisis" — even when macro statistics are not so grim.
The scale picture: the World Happiness Report 2026 puts the US in 23rd place out of 147 — below Saudi Arabia, behind Finland, Denmark, Costa Rica. Experts explain not by money but by everyday security, social connection, and trust — what has sagged for Americans. The result is a two-story construction: GDP and employment on paper are solid, but the "state of the world" — from wars to political crisis and gas prices ($4.16 per gallon versus $3.19 a year ago) — is read by people as chronic malfunction. Sentiment is not "lying about the economy," it honestly registers the state of the soul, which no longer converts one-to-one into spending.
And now the nasty surprise that rarely makes headlines: part of the "historic collapse of sentiment" may not be about people at all. Nate Silver in June 2026 publishes an analysis with a provocative thesis — everyone's debating the vibecession on data from a broken instrument. The Michigan index since April 2024 has been phased over from telephone interviews to online surveys, and regression analysis by Cummings and Tedeschi (compared with daily Morning Consult data on the same five questions) shows: online responses are systematically gloomier, the gap — about 8.9 points. This is a structural break: comparing the series after July 2024 with the historical series 1978–2024 directly is statistical barbarism. After sample weighting, the sentiment level corresponds more to 2013 than to the bottom of the Great Recession.
This is not the only crack. Partisan nonresponse — different political groups answer the phone with different willingness — and partisan "expressiveness" (people use the question about the economy to express opinions about the president) distorted the index even before the transition. UMich itself since 2017 admitted: "the partisan gap completely dominates rational assessment of trends." And the Chicago Fed in 2026 released a composite index C-CSI, where Michigan data is corrected by daily measurements — and on it sentiment looks "fairly high," not much below early 2024. That is, the sentiment-economy gap may partly be a methodological artifact, not a matter of soul.
And this is before accounting for the fact that the "average American" has disappeared as a figure of analysis. The lowest observed sentiment now is among non-stockholders, the sentiment of the richest tercile holds on portfolio growth; the top 10% by income provide 49.7% of all consumer spending (Moody's Analytics, record concentration). The aggregate index averages an economy where there is no longer an "average consumer" but two different consumers — stockholder and non-stockholder — with irreducible sentiments.
What's a forecaster to do when their pillar stops predicting? The market gives a practical answer: replace surveys of feelings with transaction data. The Chicago Fed since 2021 builds a weekly retail index CARTS on card processing; the Bank of Japan maintains an analogue; BEA experiments with multi-source data. High-frequency card data catches behavior directly, without the intermediate question "and how do you feel." Sentiment has moved from "harbinger" to the category of society thermometer: informative but not prognostic.
The irony is that the "barometer breakdown" has a happy historical rhyme. In 1946, economists were wrong because the formal model didn't account for people's psychology — and Katona built the index to return psychology to forecasting. In the 2020s, the tool built to capture the human factor itself became a victim of the human factor: politics, media, and methodology changes distorted the signal more strongly than the real economy. Katona in 1960, by the way, foresaw all this: aggregate sentiment shifts, he wrote, are caused not by personal experience (which mutually cancels out) but by "social learning" — widely broadcast information that people perceive synchronously, and "even people with contrary personal experience find it hard to swim against the current." He described the algorithm of the news feed half a century before social networks; it's just that now the current has the power of a billion screens, and the barometer has a crack in its housing.
Goldman's conclusion — "sentiment may not recover even if the economy continues to do well" — is actually a polite way of admitting: the sentiment index no longer measures spending. It's a thermometer of trust in institutions and society's general well-being. A useful instrument. It just measures something else now.