Question
Consumer sentiment fell off a cliff while unemployment stayed low. What was actually happening?
The University of Michigan's sentiment index dropped from 69 in early 2023 to the low 50s by fall, right as job openings remained elevated and nominal wage growth (especially for lower earners) was accelerating. People told pollsters they were miserable. But aggregate spending didn't collapse. Real consumption growth slowed but stayed positive. So either consumers were wrong about their own situation, or the survey was measuring something other than what we usually think.
I'd bet it's both, but the measurement story is undercooked. The Michigan survey asks about current conditions and expectations. By mid-2023, inflation expectations had started to de-anchor again depending on how you asked—some measures showed people expecting 3-4% annual price growth indefinitely, which is real. But you also had a sharp inversion in the yield curve and constant financial press about recession risk. That gets into "how bad will the economy be in 12 months" territory, not "am I doing okay now." These are different questions and surveys mix them. Someone with stable employment who thinks a recession's coming in 2024 might rate their personal finances fine but overall sentiment terrible. The indices weight those together.
The other thing: sentiment surveys skew toward people with time to answer calls and internet access. Asset prices matter to those cohorts. Stock indices fell 20% peak-to-trough in 2022, and even though they'd recovered by 2023, that volatility is salient. Meanwhile wage earners without 401ks saw nominal gains that actually meant something in real terms. Hard to know if the index is tracking mood or just correlation with market swings.
I don't think sentiment was *broken*, exactly. But I'd be careful calling it a "great vibecession" without unpacking what the index is actually measuring at any given moment.
1 comment
Log in to comment.
The measurement problem you're pointing at is real, but I'd push back slightly on how clean you're drawing the line. I spent years watching something adjacent in immigration court—when people filled out intake forms about their situation versus what they said under oath versus what the record ultimately showed, you'd sometimes get three different narratives from the same person in the same afternoon. Not because they were lying exactly, but because the question itself changed what they thought they were being asked.
The Michigan survey asks "how do you think you'll be doing financially a year from now?" That's not a prediction. That's a question that activates your news consumption, your peer group, maybe your portfolio, maybe your rent bill—and all those things at once. A court interpreter doesn't get to ask follow-up questions when someone's answer is incoherent; you just render what they said. Same thing happens here. The index captures something real about anxious anticipation, but then we slot it into "consumer sentiment" as if it's measuring satisfaction with current purchasing power, which it's not.
What you said about asset exposure is the part that actually bothers me though. If the survey tilts toward people with time to answer calls and financial accounts to check, you're not measuring how the modal worker feels about their paycheck. You're measuring how people who watch their net worth fluctuate feel about the future. Those are correlated enough to look like the same thing, but the policy conclusions could be completely different.