Question
Consumer sentiment vs. labor market tightness in 2023—what's actually going on
I was on a call with a colleague last week who mentioned that everyone she knew was anxious about employment, even though unemployment was sitting at 3.8%. She half-joked that the vibecession was real. But then I looked at quit rates and wage growth data, and it occurred to me that we might be asking the wrong diagnostic question.
The standard story goes: unemployment stayed low, inflation fell, real wages ticked up, yet consumer sentiment indices like the Michigan survey kept sagging. The divergence seemed bizarre. But here's what I think we've been missing—sentiment measures are asking about prospective financial situations and economic conditions, not employment security. Those are different things. And if you're a median earner in 2023, the prospective part actually looked worse than the immediate part, even if the immediate part was fine. You'd lived through 18 months of inverted yield curves, repeated banking panic headlines, and talk of recession. The sentiment response might have been rational, not broken.
That said, the measurement problem is real but maybe narrower than claimed. The Michigan survey asks about personal finances "a year from now"—it's built to catch expected deterioration. And expectations about recession probability *were* elevated, even if realized unemployment didn't materialize. The question is whether that forecast error tells us the measure is broken or whether consumers were just wrong about the future (they were, but not crazily so given the information set). The quit rate stayed high into 2024, suggesting people felt they had options. Simultaneously, they were anxious. Both can be true.
What I'd actually want to know: did sentiment recover once inflation clearly fell and the recession fears evaporated? And did it recover at different speeds in different survey instruments? The Michigan monthly and the Conference Board survey don't always move together. If sentiment measures were capturing something about household financial expectations rather than current labor market conditions, that's not broken—it's just not the employment story. Worth unpacking what they're actually measuring before we call them noise.
3 comments
Log in to comment.
I'd push back on the "consumers were just wrong about the future" framing, because I've seen a version of this play out in disability employment where the gap between aggregate data and lived experience gets papered over with narrative about forecast error. In 2015 or so, placement rates for people with significant disabilities were climbing, employers swore they wanted disabled workers, and the policy metrics looked solid. But the people I was counseling kept reporting the same thing: they'd get hired into roles that evaporated after ninety days, or into "permanent temp" arrangements where they couldn't accrue benefits or request accommodation without jeopardizing the position. The aggregate labor market was fine. Their prospective financial situation—the thing they actually cared about—was fragile in ways the quit rate wouldn't capture.
The Michigan survey asks about "a year from now" because that's when financial brittleness shows up. It's not irrational that someone in a nominally tight labor market would expect things to worsen if they've internalized that their job security isn't what the unemployment number suggests. You can't easily quit into something better if the betterness only lasts until the next round of cuts. I'd bet if you looked at the sentiment divergence by industry or firm size or tenure, you'd see the anxiety wasn't evenly distributed—it was concentrated in sectors or cohorts that *should have been* anxious given what actually happened next. The forecast wasn't wrong so much as unevenly calibrated to real risk.
You're right that the sentiment-unemployment divergence makes less sense once you separate "I have a job now" from "I expect my finances to be stable." But I'd push back on one thing: you're treating the prospective part as a clean measurement problem, when it's actually a translation problem, and that matters.
I've watched this exact thing happen in depositions and legal proceedings. Someone gets asked "do you expect your financial situation to improve?" and the answer depends entirely on whether they're hearing "will your income go up" or "will your costs stay manageable" or "will the system I'm navigating stop being hostile." A median earner in 2023 wasn't just forecasting GDP. They were forecasting whether their rent would keep climbing, whether childcare costs would stabilize, whether a health event would crater them. Those aren't "consumer expectations about the economy"—they're survival calculations. The Michigan survey can't actually capture that because it has to ask the same question to everyone in the same words.
So when you say consumers "were just wrong about the future given their information set," I'd qualify it: they had *different* information. Not better or worse. A quit rate staying high doesn't tell you whether people felt secure; it tells you they had options in that specific moment. Doesn't mean they believed those options would still exist in twelve months, or that taking one wouldn't expose them to something worse. The sentiment measure caught something real. It was just measuring something narrower than "will unemployment stay low"—it was measuring "am I vulnerable?" And that question didn't have a reassuring answer in 2023, even if the headline rates looked fine.
The quit rate staying high is interesting, but I'd want to know what actually happened to those people who quit. Did they move into better jobs, or just churn around looking? Because there's a real difference between "I have enough confidence in the labor market to leave my current gig" and "I'm actually materially better off afterward." The first is what kept quit rates elevated. The second is what actually matters for whether sentiment should've recovered.
I ask because I spent years watching operator turnover data get cited as proof of "a competitive labor market working" when really what we had was people cycling through the same pool of bad jobs, burning out faster. High quit rates can look like labor market tightness from the outside but feel like chaos from inside—especially if you're in a field where replacement training takes six months and you end up short-staffed continuously. That shapes how you'd answer a survey about your financial prospects a year from now, regardless of whether you technically had the option to quit.
Did anyone look at whether the sentiment recovery tracked with wage growth *relative to replacements*, or just nominal wage growth? Because there's probably a slice of the labor market where nominal wages looked decent but real employment security felt worse—fewer benefits negotiations, shorter tenure before the next layoff cycle.