The American Community Survey and Total Housing Units

In an earlier post - The Excess Vacant Housing Supply - I mentioned that there are serious question about the Census Bureau's Housing Vacancies and Homeownership (CPS/HVS) survey, and that it is probably not useful for estimating the excess vacant housing supply.

There is another more robust annual survey - the American Community Survey (ACS) - that is based on a sample of 3 million housing units every year. Unfortunately this data doesn't jibe with the decennial Census data.

The table below shows the ACS estimates of total housing units taken every July 1st. In 2000, the ACS was benchmarked to the 2000 decennial Census (as of April 1st). I've included the total completion data for single family, multi-family, manufactured homes - and calculated the implied number of demolitions using the change in the ACS.

For most years the ACS data looks somewhat reasonable, although I'd expect the number of demolitions to have peaked in 2004 through 2006. Over the first nine years of the decade, the change in the ACS averaged about 200 thousand less than total completions - suggesting demolitions of around 200 thousand per year and that is probably reasonable.

However, in April 2010, the decennial Census showed significantly more housing units than the ACS had captured (obviously a negative 1.15 million homes weren't demolished in early 2010!) The decennial Census data itself seems a little off since it suggests only about 645 thousand housing units were demolished during the decade (that would be very low). Most estimates are demolitions are in the 200 to 300 thousand per year range (so the ACS seemed reasonable through the first 9 year of the decade).

These discrepancies really needs to be explained before the ACS can be used for estimating the excess supply of vacant housing units. It is possible the 2000 Census under counted the total number of housing units - or the 2010 Census over counted the total. Or perhaps the completion data from the Census Bureau is low. But this shows one of the reason it is very difficult to estimate the excess vacant housing supply - an error of over 1 million units is huge.

Source	Date	Period	Total Housing Units	Change	Completions, Total	Calculated Demolitions
Census	4/1/2000		115,904,473
ACS	7/1/2000	3 Months	116,300,799	396,326	468,300	71,974
ACS	7/1/2001	1 Year	117,905,005	1,604,206	1,719,600	115,394
ACS	7/1/2002	1 Year	119,456,206	1,551,201	1,771,800	220,599
ACS	7/1/2003	1 Year	121,076,837	1,620,631	1,784,700	164,069
ACS	7/1/2004	1 Year	122,824,501	1,747,664	1,866,000	118,336
ACS	7/1/2005	1 Year	124,711,041	1,886,540	1,980,900	94,360
ACS	7/1/2006	1 Year	126,500,212	1,789,171	2,068,800	279,629
ACS	7/1/2007	1 Year	128,132,164	1,631,952	1,831,600	199,648
ACS	7/1/2008	1 Year	129,313,137	1,180,973	1,370,200	189,227
ACS	7/1/2009	1 Year	129,969,653	656,516	999,700	343,184
Census	4/1/2010	9 Months	131,704,730	1,735,077	584,000	-1,151,077
Decennial Census Change:			15,800,257		16,445,600	645,343