Characteristics of owner-occupied housing

Discovery of the week:

Decennial census data on the value of owner-occupied housing was always based on "specified owner-occupied units," meaning that they were single-family units on parcels of less than 10 acres (thus excluding farms), and did not have office space in the buildings.

ACS uses ALL owner units as the denominator, rather than "specified units." I am told that this change was made to accommodate condos in multi-unit structures, a worthwhile goal. However, this means that value data cover owner-occupied units in small multi-unit buildings, like two-flats and three-flats.

The bureau housing staff apparently thought that this was a rare situation. As those of us who labor in the rustbelt know, it's a very common situation in old northeast and midwest central cities. This change is distorting both the value distribution and the housing cost data for owner-occupied units, since we don't know whether the costs are for the owner unit alone or for all the units in the building. Heat is unlikely to be separately metered in these buildings; electricity may or may not be.

This change was made, as is so often the case, without any user input. A PUMS analysis could sort it out, but for a small city (even one large enough to have a PUMA), the reduced sample size of PUMS data could render it useless. It needs to be run on the full ACS sample.

Needless to say, I am not pleased!