The Consumer Sentiment Manual

Research note. Computed at render from our committed monthly Michigan sentiment series (UMCSENT, 1952+) and SPY closes. Prints are dated by initial release. See How we checked it.

What is consumer sentiment?

The University of Michigan’s Surveys of Consumers have asked US households the same core questions since the 1940s: how are your finances versus a year ago, how will they be a year from now, what about business conditions, and is now a good time to buy big-ticket items. The published index compresses five of those answers into one number, benchmarked so 1966 equals 100. It is the longest-running feelings gauge in American economics — and the operative word is feelings: it measures what people say, which this Manual will keep separating from what they do.

The latest reading

The June 2026 survey prints 49.5 — the 0th percentile of the full 74-year record, against a long-run median of 88. The 12-month average sits at 53.6, and the all-time low — 44.8, set May 2026 — is only months old. The live chart with the release calendar is on our consumer sentiment page, with the OECD-standardized consumer confidence series as the second opinion.

How the number is built

Michigan surveys roughly 600–900 respondents a month (web-based since a 2024 transition from telephone — a methodology break worth remembering when comparing precise levels across that line), publishes a mid-month preliminary and end-of-month final, and reports the headline index plus current-conditions and expectations components. Our series is the final monthly index via FRED, dated by initial release. The reference ecosystem around it: the Michigan Surveys of Consumers publish the questionnaire and components; Numerator runs a weekly purchase-panel sentiment index when you need higher frequency; and the Fed’s Survey of Consumer Finances measures the balance sheet itself — assets, debts, net worth, every three years — which makes it the reality check on the feelings, not another feelings gauge.

The record-low era — the ten lowest readings ever

Here is the fact that reframes the whole gauge: 8 of the ten lowest readings in 74 years have come since 2022. Only the 1980 Volcker squeeze keeps the remaining spots — the 2008 crisis, with unemployment at 10%, does not make the list. Households in 2022–2026 reported feeling worse than in any recession the survey ever measured, while unemployment stayed near historic lows.

Feelings vs receipts — the vibecession divergence

The record-low era’s defining feature is that spending never confirmed it. Through the same stretch, real consumer spending and real retail sales kept growing and credit-card delinquencies spent years falling — the full receipts-side case is our vibecession study. The best current explanation is that the index increasingly registers cumulative price levels, housing affordability and political identity rather than imminent behavior. That doesn’t make it useless; it makes it a gauge of the gap between mood and action — and that gap is where the information now lives. The level alone says less than it used to.

Is low sentiment bullish? The honest test

The famous claim — buy when Main Street despairs — tests out, modestly. Since 1993 (where our daily SPY history begins), bottom-decile sentiment months preceded +13.2% average SPY returns over the following twelve months, positive in 32 of 39 cases, against a +10.1% all-months baseline. Top-decile optimism preceded +9.4% — below baseline. Two honesty notes before anyone trades that: the pessimism months cluster into a handful of episodes (2008–09, 2011, 2022, and the still-open 2025–26 stretch), so the effective sample is far smaller than 39; and the edge is a tilt measured in single percentage points, not the “insanely accurate buy signal” of the folklore.

Where it will mislead you

Four traps. Feelings are not forecasts — the survey predicts spending poorly in polarized eras; read it against receipts, never alone. The partisan lens — since roughly 2016, answers swing sharply with which party holds the White House, in both directions, which inflates swings without economic content. Methodology breaks — the 2024 move from phone to web interviews shifted levels; precise comparisons across that line overstate precision. And the record-low reflex — a 74-year low reads like a screaming buy signal, but the 2022–2026 era shows the level can stay historically depressed for years while markets set records; the contrarian tilt above is real and small, and it never timed anything to the month.

How we checked it

Every figure comes from two committed files: the monthly Michigan index with release dates, and SPY daily closes collapsed to month-ends. Percentiles and the ten-lowest table use the full 1952+ history. The decile study buckets months by full-history decile cutoffs, restricts to 1993+ so forward returns use real SPY prices, grades twelve-month windows only where complete, and treats overlapping months honestly: they cluster into episodes and are not independent samples. The record-low-era claims are counts, not characterizations — 8 of ten is arithmetic.

Frequently asked questions