by Calculated Risk on 11/05/2013 07:51:00 PM

CR Note: CNBC had an article today: The housing stat you need to watch

Household formation—when a person who lives with someone else (parents, roommates, etc.) moves into another housing unit on his or her own, creating a new household—has averaged around 1 million per year historically as the U.S. population grows. In the early part of this century, when housing was just beginning its boom, it jumped by nearly 2 million. In the third quarter of this year, just 380,000 new households were formed, according to the U.S. Census.

This household formation data is from the Housing Vacancies and Homeownership report and is probably not accurate.

From economist Tom Lawler: Census Residential Vacancies and Homeownership Report for Q3: Household Estimate, If True, Disturbing; Fortunately, Probably Isn’t

The Census released the Third Quarter 2013 “Residential Vacancies and Homeownership” Report this morning, which is based on the Housing Vacancy Survey supplement to the Current Population Survey. Here are some summary stats from the report, which is commonly called the Housing Vacancy Survey (HVS) report.

Select Stats, Housing Vacancy Survey (Housing Units in thousands)
	Q3/2013	Q2/2013	Q3/2012	YOY Change
Rental Vacancy Rate	8.3%	8.2%	8.6%	-0.3%
Homeowner Vacancy Rate	1.9%	1.9%	1.9%	0.0%
Gross Vacancy Rate	13.61%	13.62%	13.66%	-0.05%
Homeownership Rate: NSA	65.3%	65.0%	65.5%	-0.2%
Homeownership Rate: SA	65.1%	65.1%	65.3%	-0.2%
Total Housing Units	132,845	132,754	132,482	363
Occupied Housing Units	114,767	114,677	114,387	380
Owner Occupied	74,901	74,563	74,878	23
Renter Occupied	39,866	40,134	39,507	359
Vacant Housing Units	18,077	18,077	18,095	-18
Totals may not add up due to rounding

The surprising – and if true, disturbing – stat from the report is the estimate for occupied homes, which for last quarter was up just 380,000 from a year ago. That YOY increase is the lowest since the second quarter of 2010, and such anemic growth, if accurate, would be “most disturbing” from a “housing recovery” perspective. However, other estimates from a CPS-based survey earlier this year show significantly faster household growth from early 2012 to early 2013 than does the HVS, making it difficult to determine what, if anything, this latest report might mean.

Click on graph for larger image.

The CPS/HVS estimate for occupied housing units, or households, is “controlled” to separate estimates of the number of housing units from the Population Division, and assumes that (1) these housing units estimates are correct; and (2) the HVS estimates for the % of the housing stock that is occupied are correct. Historical estimates are “consistent” with historical estimates of the housing stock from the Population Division, but only back to 2000.

Alternative CPS estimates for the number of households are available from the CPS Annual Social and Economic Supplement for March of each year. These household estimates are “controlled” to estimates of the civilian non-institutionalized population from the Population Division, and assume that (1) the population estimates (both total and by age, sex, etc.) are correct; and (2) that the “headship” rates from the CPS/ASEC are correct. Historical estimates from the CPS/ASEC consistent with revised population estimates for last decade (based on Census 2010 results) are only available back to 2011.

The CPS/ASEC estimate for the number of US households for March, 2013 was 122.459 million, up 1.375 million from March, 2012. The HVS estimate for the number of households for March, 2013 was 114.061 million, up 373 thousand from March 2012, and the HVS estimate for the average number of households in the first half of 2013 was 114.480 million, up 575 thousand from the first half of 2012. The CPS/ASEC estimate of the number of US households increased by 2.532 million from March 2011 to March 2013, while the CPS/HVS estimate increased by only 1.306 million over that period.

Different CPS-Based Estimates of US Households (000's)
	Mar-13	Mar-12	Mar-11	Mar-13 vs. Mar-11
CPS/HVS	114,061	113,688	112,755	1,306
CPS/ASEC	122,459	121,084	119,927	2,532

These shockingly different estimates both for the number of households and for recent changes in the number of households have understandably confused, baffled, and annoyed analysts and policymakers.

While both the CPS/HVS and the CPS/ASEC estimates are “controlled” to select Census 2010 results – one for the number of housing units, and the other to the number of people -- neither survey’s results are “controlled” to critical Census 2010 estimates such as the number of households (or headship rates), homeownership rates, vacancy rates, etc. -- making both of limited value to housing analysts and policymakers. Comparisons to Census 2010 results indicate that both CPS-based surveys considerably overstate homeownership rates, especially for “younger” adults, and that the CPS/HVS considerably overstates housing vacancy rates. While it is not clear whether the “bad” CPS results are based on sampling or non-sampling error – no real research has been done on this – the “sampling frame” for the CPS is incredibly outdated, and Id guess that “sampling error” is a “big deal.”

For analysts focused on the housing outlook, which is heavily dependent on the growth in households, the latest HVS estimates are “disturbing,” but the massive disparity between CPS-based household growth estimates based on “people” counts and household growth estimates based on “house” counts earlier this year makes one, I guess, “less disturbed,” though kinda pissed off.