Measurement broken, or measuring something real that GDP missed

I'd push back on this a bit, actually. You're right that sentiment indices catch real constraints—the mortgage rate thing is genuine. But I watched something similar play out in transit ridership during 2022-23, and the lesson I took was almost opposite to yours. We had this period where our on-time performance metrics stayed solid, service hours were up, and labor costs were stable. Sentiment was terrible anyway—riders complained constantly about crowding, reliability, the whole thing. Management kept saying "the data says we're fine." But when we finally drilled into actual trip patterns, it turned out people weren't responding to average conditions at all. They were responding to whether *their specific commute* had gotten worse. One line got 15% worse in rush hour (genuine delay spike from traffic pattern changes), and usage collapsed on that line specifically. Another line improved slightly overall but became more unpredictable, and people abandoned it anyway—they'd switched to cars. The thing is, we *had* the granular data. We just weren't looking at it. The aggregate metrics felt clean. The sentiment felt like noise. So my question back at you: when the original poster says measurement wasn't granular enough, do you mean the data literally doesn't exist, or that nobody's constructed the right index from what we have? Because in my experience it's usually the second one. The Conference Board could probably weight housing affordability against job availability if they built that index. They choose not to, probably because it's messier and less stable month-to-month. That's different from measurement failure. That's just choosing to measure something comforting instead of something true.

You're going to see this get worse before anyone builds better instrumentation, and I'd guess the lag will be longer than you think. Here's why: the institutions that could actually measure this—the Fed, BLS, Treasury—have existing incentive structures around what they measure. They've got decades of methodology tied to unemployment, CPI, wage growth. Reorienting to track debt service as a percentage of income for non-college workers in specific metros, or the actual cost of upgrading your housing situation, or what fraction of jobs in a region actually pay enough to cover rent in that region—that's not a spreadsheet update. That's designing new data collection, getting buy-in, waiting for quarterly reports, dealing with seasonal adjustment debates. Meanwhile the sentiment indices will keep looking wrong because they're actually picking up something real that the aggregate figures smooth over. You'll get more think pieces about "irrational consumer pessimism" and periodic recalibrations of the sentiment indices themselves, treating them as the problem. The working assumption will remain that unemployment is the meaningful number and sentiment is noise, partly because unemployment data is clean and fast and partly because institutions are built around it. By the time anyone builds the granular tracking you're describing, we'll be three years into the next cycle and everyone will move on.