Thursday, January 12, 2017

Lawler: New “Household” Numbers, Same Old Conundrum

by Bill McBride on 1/12/2017 11:08:00 AM

From housing economist Tom Lawler: New “Household” Numbers, Same Old Conundrum

Housing Survey (AHS) for 2015, and the “Families and Living Arrangements” data from the Annual Social and Economic Supplement to the Current Population Survey (or CPS/ASEC) for 2016. Both surveys produce estimates (wildly different, of course) of – among other things – the number of US households and the homeownership rate.

Starting with the American Housing Survey, 2015 marked the first time since 1985 that the AHS was based on new national and metropolitan area “longitudinal” samples based on the latest available Master Address File. From 1985 through 2013 the AHS sample was mainly based on housing units selected from the 1980 Census as well as samples of housing units subsequently constructed in areas requiring building permits. Not surprisingly, the previous methodology was subject to sizable “sampling” issues. The 2015 “national’ estimates are based on (1) a “national case” sample of 34,769 representative of the US and nine divisions; (2) a 45,270 “over-sample” of top 15 metropolitan areas; (3) a 30,111 over-sample of 10 additional metropolitan areas; and (4) a 5,248 oversample of subsidized renter units.

The AHS household estimates are “controlled” to independent estimates of the US housing stock in much the same was as are the household estimates from the Housing Vacancy Survey, a supplement to the Current Population Survey. And while the AHS-based US household estimates for 2015 in aggregate aren’t massively different from that of the HVS, the characteristics of the AHS-based households for 2015 are vastly different, and are more in synch with those of the 2015 American Community Survey, as shown in the table below.

2015 US Household Estimates by Age Group, Various Census Surveys
  AHSHVSACS
Total118,290117,397118,209
15-24 4,3476,1254,441
25-3418,09619,10617,885
35-4420,44119,91720,576
45-5423,53422,02123,238
55-6423,63822,06823,069
65-7416,29215,99616,524
75+11,94212,16412,476

What is especially striking are the rather sizable differences in the shares of US households by age group between the the HVS estimates and the ACS or AHS estimates, with the HVS estimates suggesting a much larger “young-adult’ share of total households.

Even more striking are the differences in the homeowner estimates by age group.

2015 US Homeowner Estimates by Age Group, Various Census Surveys
  AHSHVSACS
Total74,36074,74174,506
15-24 5011,336577
25-346,4857,4936,582
35-4411,74311,64511,599
45-5415,96815,41315,853
55-6417,67416,64517,285
65-7412,91712,81913,160
75+9,0729,3909,450

As this table indicates, the HVS estimates for young-adult homeowners are vastly higher than the AHS and ACS estimates.

Census also produces household estimates based on the CPS Annual Social and Economic Supplement (CPS./ASEC) which are not controlled to independent housing stock estimates, but instead to independent estimates of the civilian non-institutionalized population. Since (1) CPS-based surveys overstate housing vacancy rates; and (2) housing stock estimates appear to be understated, CPS/ASEC household estimates are higher than CPS/HVS estimates.

While AHS, HVS, and ACS estimates are more or less annual averages, CPS/ASEC estimates are for March, and the last two estimates for the latter are shown below.

US Household Estimates by Age Group, Various Census Surveys
AHS (2015)HVS (2015)ACS (2015)CPS/ASEC (Mar 2015)CPS/ASEC (Mar 2016)
Total118,290117,397118,209124,587125,819
15-244,3476,1254,4416,3706,361
25-3418,09619,10617,88520,07520,047
35-4420,44119,91720,57621,12121,222
45-5423,53422,02123,23823,56623,295
55-6423,63822,06823,06923,50923,896
65-7416,29215,99616,52416,88617,551
75+11,94212,16412,47613,06113,448

Again, what is “most striking” about the numbers in this table is the substantially higher estimates of young adult householders from the CPS-based surveys relative to the other surveys.

(I’ll have much more on this topic later, including an assessment of the reliability of the household estimates from these surveys).