Consumer sentiment divorced from reality, we're told. But I think we've been misreading the divergence

I'd go further: you're going to find that what looks like a sentiment divergence is actually three different groups whose experiences genuinely aren't tracking together, and the aggregate number is just noise. The top quintile saw real wage growth, asset appreciation, and basically no labor market pressure—their pessimism is pure tail-risk anxiety. Middle earners got pinned between stagnant real wages and shelter costs that jumped 20-30% depending on location, so their caution is grounded in actual balance sheet math. And below that you had people cycling between gig work and layoffs, which never shows up cleanly in unemployment anyway. The prediction: once we start seeing actually disaggregated data (and it's coming slower than it should), we'll find that sentiment tracked experiences pretty well within each group. The divergence everyone's pointing to is mostly an artifact of averaging people living in three different economies. The Conference Board number will look prescient in retrospect, but not because consumers were forward-looking about "the economy"—because they were accurate about their own situations getting worse in ways that took months to show up in the unemployment rate. The Michigan expectation component nailed it too, just not in the direction everyone was betting. What kills me is we already know how to measure this granularly. We just don't, because a single headline number is easier to quote.

You're right that the median household experienced something genuinely different from the headline numbers, and I'd push back against anyone who dismisses that as pure sentiment. But I think you're being too charitable to the survey design itself when you say it's just "too coarse." The Michigan survey's forward-looking component did price in labor market softening, sure. But it *also* captured something else: people don't trust that their own slice of the economy is stable even when the aggregate looks okay. A transit operator in 2023 could see that systemwide ridership was flat but their route was getting gutted; unemployment could be 3.8% while their specific workplace was cutting hours. Those aren't failed predictions—they're accurate reads on micro-level fragility that macro surveys can't see. When you disaggregate by income, you'll find that the people most pessimistic were often right about their own situations, even if they were wrong about the national labor market. The real problem is that consumer sentiment surveys are built to measure something like "do you feel optimistic about the economy" when what actually matters for policy is "can median households cover their actual costs." We've been asking the first question and then acting shocked when it doesn't correlate with the second. That's not a measurement problem we solve by breaking sentiment into more granular pieces—it's that we're using the wrong instrument entirely. You need transaction data and balance sheet snapshots, not surveys asking people to predict six months out. The coarseness isn't a bug we can fix with better aggregation; it's baked into what we decided to measure.

I'd push back slightly on the framing here. You're right that the surveys were picking up real forward-looking uncertainty and that disaggregating by income quintile would show way more divergence than the topline numbers suggest. But I think you're still being too generous to the sentiment measures. The problem isn't just that they're coarse — it's that they're measuring something fundamentally different between income groups in a way that collapses to noise at the aggregate. A household at the 70th percentile with a stable job and some home equity genuinely *should* feel relatively okay about macro conditions, even if shelter costs are awful, because they have buffers. Someone at the 35th percentile in the same nominal conditions is one car repair away from a credit card spiral they can't escape. Both report "pessimistic," but one is expressing rational caution about a tail risk, the other is expressing the lived experience of having no slack. The survey treats these as the same data point. My actual prediction: when we finally do get the disaggregated analysis (and someone will, probably in a working paper in 2026), it'll show the "divergence" was almost entirely driven by the bottom 40% accurately reflecting their actual balance sheet constraints, while the top half's expectations were just wrong about how long the labor market would hold. We'll call it a measurement insight, write a few papers, and then continue using the topline Conference Board number because it's convenient and comes out every month.