The replacement band,
town by town.

The band is a building period, not a population. A town that filled in before the war holds very few of these homes however large it is, and a town that was built through the nineties holds a great many however small.

Among towns with at least 2,000 homes on record, the share sitting in the band runs from 0.4% to 42.4% — a spread wide enough that the town with the most houses and the town with the most work are rarely the same place.

Twenty to forty years is a wide thing to call one cohort. Split at its midpoint, the older half — built 1986–1995, now 31–40 years old — holds 128,653 homes. The younger half holds 117,297.

Behind them, 24,646 homes built 2007–2011 are not in the band today and enter it over the next five years. That is 10% of today’s band arriving on top of it, and it is not spread evenly.

None of this says a roof is due. It says when the houses were built, which is a fact about the town, and leaves the rest to you.

For each town we took the median assessor building area of the homes in the band and divided it by the median roof we measured from above. We expected a spread. What came back was two separate populations with almost nothing between them.

Between them, at 1.6–1.8×, sit 3 towns out of 300. That gap is the finding. A range of house shapes would fill it in smoothly; a choice each assessor makes about what to write in a column does not.

And 25 towns come in under 1.00×, where the assessor’s figure is smaller than the roof — garage and porch under the same roof, never reaching the finished-area column.

So a multiplier calibrated in your own town is not slightly off in the next one. If you cross that line it is out by roughly double, and nothing on the card tells you which side you are on.