The Number We Counted Wrong

The Minneapolis Fed published the other one on July 15. Erik Hembre, Maxine Xu, and Ben Horowitz call it the homeowners-to-population ratio — HPOP — and the arithmetic is so simple it's almost rude: count adults who own a home, divide by adults. The number everyone quotes counts housing units. If a unit is owner-occupied, the household inside it is a homeowner household, and the household is the unit of account. Everyone in it disappears into the tally.

Twelve points, nationally. Under 35, the gap opens into a canyon: the conventional rate says 37 percent, HPOP says 22. Only about one in three adults under 35 heads a household at all. The rest live inside somebody else's.

That's the whole trick, and it isn't a trick. Nobody falsified anything. About 14 percent of American adults live in owner-occupied housing without owning it — adult children back in the childhood bedroom, aging parents in the spare room, siblings, cousins, the friend on the couch since the lease ended. The traditional measure does not record them as failing to own. It records them as residents of an owner-occupied unit, which is to say: inside the American Dream, statistically, by adjacency.

The objection arrives fast, and it's a fair one: isn't multigenerational living normal — the global and historical default, a preference rather than a failure state? Look at the spread. Hawaii's gap between the two measures is 18.9 points. North Dakota's is 3.9, and not from policy genius: about 10 percent of young adults there live with a parent, against 30 percent nationally. Preferences do not sort themselves by state along a housing-cost gradient that cleanly. At this scale, co-residence is priced, not chosen. The gap is an affordability gauge wearing a methodology costume.

We have been here before, though not quite in the way it first looks. In 1983 the Bureau of Labor Statistics stopped pricing the homeownership component of the Consumer Price Index off house prices and switched to owners' equivalent rent — what your house would rent for, standing in for what it costs you to live in it. Defensible, and argued in public: a house is an asset as well as a shelter, and the CPI measures consumption. Shelter didn't leave the index; it is still the largest single component of it. What left was house appreciation, which stopped registering as inflation in 1983 and therefore did not register across the four decades it did the most work. The federal poverty threshold is a 1963 food budget times three, indexed for prices ever since, because food was roughly a third of household spending in 1963. It is not now. Housing is. The threshold has never been told. And the unemployment rate on the evening news is U-3, which requires you to have looked for work in the last four weeks — stop looking and you don't become unemployed, you exit the numerator and the denominator simultaneously. The rate improves.

These are not one mechanism, and it's worth being exact about that. One measure was actively revised. One is a failure to update a basket everyone agrees is stale. One was never designed to count people and got asked anyway. And U-3 isn't hidden at all: the BLS publishes U-6 in the same table, on the same morning, every month, and has for decades. That last case is the one that gives the game away. When the honest number is printed directly beside the flattering one and the flattering one still leads the broadcast, the instrument was never the problem. The instrument is downstream. What persists across all four is the reach — the reliable, unforced preference for the number that leaves nobody with a question to answer, and the corresponding slowness to build one where it doesn't yet exist.

A measurement is a coherence. A shared agreement about what's real that lets millions of people act together without renegotiating reality every morning — genuinely valuable, most of what makes a society possible. It's also where the leverage sits, because the agreement decides who counts as party to it. And a coherence earns its legitimacy by including the people it is built on top of, not by defining them out of the frame.

The household-based rate defines them out of the frame. Not maliciously — structurally. It was built to measure the housing stock, and it measures the housing stock beautifully. We asked it about people and it answered about buildings, confidently, for sixty years, and nobody noticed the category error because the answer was flattering.

But there's a harder reason it went unnoticed, and it isn't a mistake at all. The household is not a neutral container that statisticians happened to grab. It is the unit the economy actually settles on. Mortgages are underwritten to households. The deed, the interest deduction, the equity, the appreciation, the thing that compounds and gets inherited — all of it accrues to the household, none of it to its residents. So the household-based rate is not failing to see the 29-year-old in the basement. It is seeing him precisely: he has shelter and no claim. The measurement and the property regime agree because they are describing the same object. The number was faithful. Just not to him.

Notice what changes when you switch. Count units, and housing policy optimizes for units: starts, permits, occupancy, supply. Every one of those can improve while he stays exactly where he is — he already lives in an owner-occupied unit, so he can never appear as a shortfall. He is not a data point. He is furniture: part of the dwelling, not a party to it. Count adults, and he can't be hidden. The unit of measure is the unit of moral concern; everything downstream is bookkeeping.

The forecast isn't optimistic, and not for the reason you'd expect. Both numbers exist now, so the argument is no longer about methodology — it's about which one gets quoted. And quotation follows settlement. The 65 is referenced by mortgage markets, by tax policy, by every instrument ever written against a household. Nothing whatsoever is written against HPOP. No contract underwrites an adult. A measure that nothing settles on has no constituency, and measures without constituencies don't get quoted — which is precisely why U-6 has sat in public for decades, correct and ignored.

Which leaves the uncomfortable part. A society at scale cannot see itself directly. It sees only through instruments it agrees to maintain, which makes measurement a kind of delegated perception — and the failure mode of delegated perception isn't lying. It's resolution. The instrument cannot render the thing, so the thing is not there to be seen, and nobody ever has to decide to ignore it. Sixty years of not noticing required no one's bad faith. It required only that the number be flattering and that nothing depend on the alternative.

Twenty-two percent. Write it somewhere it can't be revised — and then go find something that settles on it.