Wednesday, December 26, 2012

Lawler: An “Update” to the “Excess” Supply of Housing

by Bill McBride on 12/26/2012 04:42:00 PM

CR: Housing economist Tom Lawler sent me the following long piece that suggests a large number of the excess vacant housing supply has been absorbed.

Housing economist Tom Lawler writes: An “Update” to the “Excess” Supply of Housing; How Much Has the Number of Vacant Homes Fallen Since April 1, 2010 (through December 1, 2012?)

It is over 2 1/2 years since the Decennial Census 2010’s “snapshot” of the US population and housing market on April 1, 2010. While private housing analysts are still awaiting the result of research by Census analysts on the reasons for the sharply different results of Census 2010 compared to other Census surveys (e.g., the ACS and the HVS), I thought it might be useful to review some numbers since the Census was taken.

On the housing production from, Census estimates suggest that from April 2010 to November 2012, housing completions plus manufactured housing units totaled about 1.817 million (an annualized pace of about 681 thousand).

There are no data on the net loss to the housing stock over this period. Prior to the release of Census 2010 results many folks thought that the net loss to the housing stock last decade was averaging around 200 – 250 thousand units a year, but the decennial Census results suggested a much smaller number. But for fun, let’s assume that the net loss in the housing stock since the decennial Census has been about 400,000, or an annualized rate of 150,000.

Such a number would imply that the housing stock at the end of November/beginning of December increased by about 1.417 million, or an annualized rate of about 531 thousand.

Now what about the number of households (or occupied housing units)? Sadly, here there are no good, reliable data to count on. For 2012, there are two sources of “estimates” on US “households,” both based on supplement surveys of the Current Population Survey. One source is the Housing Vacancy Survey, which assumes that (1) the Census’ Population Division estimates of the US housing stock are correct; and (2) the HVS’ estimates of the % of the housing stock are correct. Census 2010 results (and to a lesser extent ACS results) strongly indicated, of course, that the latter assumption is not correct: the HVS appears to overstate significantly the share of the housing stock that is vacant, with the overstatement growing over the past few decades.

With that caveat in mind, the HVS estimates are that the number of US households averaged 114.916 million in September 2012, compared to an average of 112.633 million in March-April of 2010. The official Census 2010 household estimate for April 2010 was 116.716 million. Assuming that the HVS estimates for the last 3 months of 2012 show similar YOY growth as the previous few months, and “grossing up” the totals to be consistent with Census 2010 totals, the HVS estimates might suggest household growth from April 1, 2010 to December 1, 2012 of about 2.62 million, or an annualized rate of about 983 thousand. This is a “low” estimate.

Another source of an “estimate” of US households in 2012 is the Annual Social and Economic Supplement to the Current Population Survey. This annual survey, taken over February, March, and April with an “expanded” sample size relative to the “normal” monthly CPS and HVS surveys, purportedly produces “estimates” of the number of US households in March of each year that are consistent with (1) civilian non-institutionalized population estimates, and (2) survey results. The CPS/ASEC, in essence, is “controlled” to population estimates, as opposed to the CPS/HVS, which is “controlled” to housing stock estimates.

In the latest CPS/ASEC for March, 2012, the estimate of the number of US households was 121.084 million, which is a staggering 4.368 million higher than the official Census 2010 estimate for April 1, 2010. The CPS/ASEC revised household estimate for March, 2011, based on Census 2010 population controls, was 119.927 million, up from the previous 118.682 million in the 2011 report based on Census 2000 population controls. Census did not provide updated March 2010 estimates based on Census 2010 population controls.

It should be noted, however, that the CPS/ASEC household “estimates” are not “controlled” to Census 2010 household estimates, but instead are “controlled” to population estimates, and the CPS/ASEC survey results appear to significantly overstate US households (they also aren’t consistent with decennial Census estimates of the household, as opposed to civilian non-institutionalized, population estimates). I “guesstimate” that a CPS/ASEC household estimate consistent with Census 2010 household population estimates and recently-released 2012 household population estimates by age group for March, 2012 would be about 119.6 million, and that an estimate for December 1, 2012 using updated household population estimates would be about 120.5 million, about 3.8 million higher than the Census 2010 estimate for April 1, 2010, and an annualized increase of about 1.425 million. This is a “high” to “very high” estimate.

Another alternative would be to look at updated estimates of the household population (available through December 1, 20121), and then make certain assumptions either about household size (very crude) or make certain assumptions about “headship” rates by age group. Below is a table with some data to start with.

A few things are worth noting: first, overall population growth is estimated to have grown at an annualized rate of about 0.74% since the decennial Census was taken, and the household population is estimated to have grown at an annualized rate of 0.76%. This growth rate is significantly lower than last decade’s average, partly reflecting lower immigration levels and partly reflecting lower birth rates.

The population of adults – which is more important in terms of household growth, is estimated to have grown at a more rapid annualized rate. E.g., the 25+ year household population is estimated to have grown at an annualized rate of about 1.09%.

   Annualized % Chg
 4/1/2010 12/1/2012 
Resident Population308,747,508 314,918,6150.74%
Household Population300,758,251Average Household Size306,855,5150.76%
Households116,716,2922.577  
Household Population by Age Group    
15-2440,202,045 40,442,5540.22%
25-3440,005,898 41,371,9611.27%
35-4440,277,153 39,666,291-0.57%
45-5444,288,729 43,274,284-0.87%
55-6436,068,290 38,485,6462.46%
65-7421,429,316 24,256,9894.76%
75+17,380,962 18,231,3151.81%
25+ Years199,450,348 205,286,4861.09%
Households by Age Group Headship Rate*  
15-245,400,79913.43%  
25-3417,957,37544.89%  
35-4421,290,88052.86%  
45-5424,907,06456.24%  
55-6421,340,33859.17%  
65-7413,504,51763.02%  
75+12,315,31970.86%  
* Households divided by Household Population


[Note: the difference between the “resident” population and the “household” population is the number of people estimated to be living in “group quarters,” usually broken out between “institutionalized” (including correctional facilities for adults, juvenile facilities, and nursing/skilled nursing facilities) and “non-institutionalized” (including college/university student housing, military quarters, and other group housing).]

There are two “Q&D” ways one might “gueestimate” the number of households on December 1, 2012: one – very quick, extremely dirty – would be to assume that the average household size had remained the same. That approach, which doesn’t take into account shifts in the age distribution of the population, would lead to an estimate of 119.028 million, up 2.366 million from April 1, 2010.

A second approach would be to assume that the “headship” rates for different age groups on December 1, 2012 was about the same as on April 1, 2012. Using that “Q&D” approach, one would get an estimate of the number of households on December 1, 2012 of about 120.283 million, up 3.567 million from April 1, 2010.

So … let’s assume that a “very low” case for household growth from April 1, 2010 is around 2.4 million (annual rate of 900 thousand); a “high” case is 4.0 million (annual rate of 1.5 million), and a “base” case is around 3.2 million (annual rate of around 1.2 million). What might these numbers mean for the number of vacant homes as of December 1, 2012 compared to April 1, 2010?   Here is a table showing (rounded) what the numbers might look like.

Changes from April 1, 2010 to December 1, 2012 (millions of units)
 LowBaseHigh
Household Increase2.43.13.8
Housing Production*1.81.81.8
Net Housing Units Lost0.40.40.4
Housing Units1.41.41.4
Vacant Housing Units-1.0-1.7-2.4
*Housing Completions plus Manufactured Housing Placements (with November estimate)

Under a “very low end” estimate of household growth, the number of vacant units since April 1, 2010 would be down by about a million. Under a “very high end” estimate of household growth, the number of vacant housing units would be 2.4 million lower. And a “not too unrealistic” estimate of household growth would imply that the number of vacant housing units was down by about 1.7 million.

Now, does a 1.7 million decline in the number of vacant homes for sale since April 1, 2010 seem plausible? Well, if that were the case one would probably expect that the number of homes for sale, for rent, and held as REO would be down significantly. So, let’s take a look at some available numbers.

NAR estimates that the number of existing homes for sales declined from 3.09 million at the end of March 2010 to 2.03 million at the end of November 2012, a decline of about 1.06 million. Realtor.com’s listings numbers fell by a similar amount. Obviously not all homes listed for sale are vacant, but a significant % are vacant.

Census estimates that the number of completed new SF homes for sales declined from 92 thousand at the end of March 2010 to 40 thousand at the end of October 2012, a decline of 52 thousand.

The REO inventory of Fannie, Freddie, FHA, and private-label ABS, combined with an estimate of the REO inventory of FDIC-insured institutions (based on $ carrying amounts and estimates of the average carrying balance) declined from about 531 thousand from the end of March 2010 to about 367 thousand at the end of September 2012, a decline of about 164 thousand. Some, but probably less than 40%, of these REO properties were listed for sale.

There aren’t good, aggregate data on the number of homes for rent: HVS has estimates, but comparisons with decennial Census data indicate that HVS rental vacancy rates not only are overstated, but also that the overstatement has grown over time. Given that caveat, the HVS estimates show that the number of homes for rent declined from a first-half 2010 average (to come close to an April 1 estimate) of about 4.458 million to a third-quarter 2012 average of 3.809 million, a decline of about 649 thousand. The actual decline is probably larger.

Hmmmm…..gosh, a decline in the number of vacant homes of about 1.7 million since April 1, 2010 sure SEEMS plausible!

But wait: if the number of vacant homes since Census 2010 has been that large, then that would imply a sizable reduction in the “excess” supply of housing – enough so that, if true, one should have expected to see stability in, or even in many areas even increases in, home prices in 2012! Could that really be true? (CR note: see previous posts!)

Looking ahead to the next few years, the likely growth in population by age groups suggests that household formations should average about 1.3 million a year, with some upside if headship rates rebound in any meaningful fashion.

1 Actually, “estimates” are available through July 1, 2012, and data from August 1, 2012 through December 1, 2012, are short-term “projections.”

CR Note: This was from housing economist Tom Lawler.

Last 10 Posts