Monday, February 07, 2011

Lawler: Housing Vacancy Survey appears to massively overstate number of vacant housing units

by Bill McBride on 2/07/2011 08:30:00 AM

CR Note: Many analysts use the quarterly Housing Vacancies and Homeownership survey from the Census Bureau to estimate the number of vacant housing units in the United States. This survey probably overstates the number of vacant units.

From economist Tom Lawler ...

Early Look at a Few States’ Housing Stock Numbers from Census 2010 Suggests that Housing Vacancy Survey Massively Overstates Number of Vacant Housing Units

The Census Bureau has starting releasing Census 2010 data for selected states, with the first “batch” being Louisiana, Mississippi, New Jersey, and Virginia. On its website, the data shown include not just population counts, but housing counts as well – occupied as well as vacant. I don’t know if these are the “final, official” housing count totals, but here are the data shown on the website for these states. Also shown are the counts from Census 2000 (NOT adjusted for subsequent “overcounts/undercounts,” as well as the gross vacancy rates (total vacant units – including seasonal – divided by total housing units) for Census 2010, Census 2000, and the gross vacancy rates from the Housing Vacancy Survey HVS) and the ACS for 2009 (2010 data have not yet been released).

 LouisianaMississippiNew JerseyVirginia
Census 2010 (4/1/10)
Housing Units1,964,9811,274,7193,553,5623,364,939
Census 2000 (4/1/00)
Housing Units1,847,1811,161,9533,310,2752,904,192
Gross Vacancy Rate
 LouisianaMississippiNew JerseyVirginia
Census 201012.00%12.50%9.50%9.20%
Census 200010.30%9.90%7.40%7.10%
HVS 200914.00%16.40%12.10%12.20%
ACS 200914.00%14.60%10.50%10.80%

What is, of course, especially noteworthy is that the “gross” vacancy rates from the HVS for 2009 (which are yearly average estimates) are massively higher than the Census 2010 estimates for April 1, 2010, as well as being materially higher than those from the ACS for 2009. As I have noted numerous times, the HVS has for at least a decade “found” materially higher gross vacancy rates for most areas (Louisiana being an exception) than the ACS, and “found” materially higher gross vacancy rates for 2000 than was the case for Census 2000. I’ve talked to folks about this, but no one seems to know why.

What is also interesting is that the gross vacancy rate from Census 2010 for the above states is substantially lower than those shown in the ACS for 2009. While given the low level of housing production it would not be surprising that the gross vacancy rate on April 1, 2010 would be lower than the average for 2009, it’s hard to believe that the “actual” gross vacancy rate in these states declined by as much as that suggested by assuming that the ACS 2009 GVR and the Census 2010 GVR were both “correct.”

Not surprisingly, the gross vacancy rates in the above states from Census 2010 were all up significantly from Census 2000. However, it would appear as the GVRs did not increase by as much as suggested either in the HVS (state data are only available back to 2005) or the ACS, and it seems clear that the HVS’ estimates for TOTAL vacant housing units is a boatload higher than the “actuals.”

Next week the Census is scheduled to release Census 2010 data for Arkansas, Indiana, Iowa, Maryland, and Vermont, and the deadline for data releases for all states is, intriguingly, April Fools’ Day.