The "Household Conundrum"

And Early Read on Existing Home Sales in October

Housing economist Tom Lawler has been sending me his predictions of what the National Association of Realtors (NAR) will report for the last 12 years.  He has graciously allowed me to share his predictions with the readers of my blog - and now this newsletter. Tom really analyzes the data, and he is almost always much closer to the actual report than the “consensus” of analysts.

Tom also writes about a number of housing issues, and today he discusses the “household conundrum” - why it is unclear how many households there are in the U.S..

Early Read on Existing Home Sales in October

From Tom Lawler:

Based on publicly-available local realtor/MLS reports released across the country through today, I project that existing home sales as estimated by the National Association of Realtors ran at a seasonally adjusted annual rate of 6.34 million in October, up 0.8% from September’s preliminary pace and down 5.8% from last October’s seasonally adjusted pace. Unadjusted sales should show a larger YOY decline, reflecting this October’s lower business day count relative to last October’s.

Local realtor reports, as well as reports from national inventory trackers, suggest that the YOY decline in the inventory of existing homes for sale was slightly larger than was the case in September.

Finally, local realtor/MLS reports suggest the median existing single-family home sales price last month was up by about 12.9% from last October.

Note: The National Association of Realtors (NAR) is scheduled to release October existing home sales on Monday, November 22, 2021 at 10:00 AM ET. The consensus is for 6.20 million SAAR. 

Also, note that the NAR reported median prices were up 23.6% Year-over-year (YoY) in May, 23.4% in June, 17.8% in July, 14.9% in August and 13.3% in September. And it appears the YoY growth slowed a little more in October. I mentioned this in The Coming Deceleration in House Price Growth

The “Household Conundrum”

The following is from Tom Lawler:

One of the challenges facing housing analysts in trying to assess the current and future state of the US housing market is the lack of timely and reliable estimates of the number of and characteristics of US households.  This challenge has become even more acute since the onset of the pandemic, as significant declines in response rates to government surveys of households have adversely affected the accuracy of already imperfect household estimates. The lack of reliable household data makes it difficult to assess how much of the astonishing strength in the housing market since the middle of last year has been related to “demographics,” as opposed to behavioral and preference changes associated with the pandemic and, more recently, the surge in investor purchases of single-family homes.

The government produces several intercensal estimates of US households, but the ones most widely followed are those from the Housing Vacancy Survey (a supplement to the Current Population Survey, or CPS), the Annual and Social Economic Supplement (ASEC) of the CPS (which is a larger sample size), and the American Community Survey (ACS).  And the only timely estimates produced are quarterly numbers from the HVS (though detailed HVS estimates of the characteristics of households are only available on an annual average basis.)

While HVS household estimates have always been volatile (partly related to the small sample size of the CPS), and have not matched decennial Census results, the estimates produced since the onset of the pandemic have been … well … ridiculous.  Below is a table of the HVS estimates of total and owner-occupied households.

As the table indicates, the number of households estimated from the HVS exploded upward following the onset of the pandemic, and the number of owner-occupied households rose by an astonishing and completely unbelievable amount.  And following that surge the HVS estimates of total households fell, and the number of owner-occupied households plummeted!

The reason for these inane results almost certainly stems from a combination of survey issues following the onset of the pandemic and the methodology used to estimate the number of occupied housing units (or households). 

Here is an excerpt from the HVS report for the first quarter of 2020.

“Data are generally collected the week of the 19th. The reference period is the time of interview, in this case March 15-25. On March 20, the Census Bureau suspended in-person interviews. Further, two Census call centers that assist with interviews were closed. The Census Bureau continued to conduct the CPS/HVS by telephone and made efforts to collect telephone interviews for households and vacant units that would normally have been interviewed in person. The response rate was 73 percent in March 2020, about 10 percentage points lower than the response rates of 82 percent in January and 83 percent in February. The response rate in March 2019 was 82 percent.”

Person interviews remained suspended through June, and did not move back to 100% until September.  Overall response rates continued to decline through the second quarter, hitting a low of 64.9% in June, and remained below 70% until September.  Since then response rates have risen but have remained below historical norms.

Census did not alter its methodology in response to the pandemics, and it said that it “likely cannot fully understand or quantify the effects of the pandemic on the CPS/HVS data and estimates.”  

However, what likely happened is that the lack of personal interviews and the dramatic lower response rates produced substantially lower vacancy rates – and higher home ownership rates – than was actually the case.  And since HVS household estimates are “controlled” to independent household estimates, a lower vacancy rate estimate resulted in dramatically higher household estimates.

Ironically, one other CPS-based household estimate – that derived from the CPS/ASEC – may also have been impacted by the pandemic, but in the OPPOSITE DIRECTION.  The CPS/ASEC produces household estimates for March of each year, though the survey spans over several months.  As with the CPS, response rates in March and April of 2020 were way below normal, and those lower response rates (and other survey issues) almost certainly impacted the CPS/ASEC household estimates.

Below is a table showing the CPS/ASEC derived household estimates for March of the past several years.

As the table shows, the CPS/ASEC household estimate for March 2020 was actually BELOW that of March 2019, a stark contrast to the sizable gain in the CPS/HVS household estimate.

The reason could be twofold:  first, the lower response rate to the CPS/ASEC could have produced a higher than “actual” increase in the average household size for 2020.  And second, the CPS/ASEC household estimate is “controlled” to population estimates, not housing stock estimates.

(Note:  the decline in the CPS/ASEC household estimate from 2019 to 2020 was also related to the different population estimate “vintage” used in the two years.  CPS/ASEC historical estimates are not revised to reflect revisions in population estimates.)

Informed readers might ask “well what about the American Community Survey household estimates for 2020?”  Well, there aren’t any.  Here is a statement from Census on the ACS.

“The Census Bureau has an obligation to produce accurate, relevant statistics about the nation’s economy and people. The data collection issues experienced by the 2020 ACS severely affected the data quality of these statistics, therefore, the Census Bureau decided not to release the standard ACS 1-year data for 2020. This report describes the data collection disruptions in 2020, the modifications to standard weighting and estimation to combat the collection issues, and the resulting data quality issues that informed the decision to not release the standard 1-year ACS data products.”

Census does plan to release “experimental” ACS estimates at the end of this month, but has cautioned that it may be inappropriate to compare these estimates to previous year estimates.

And finally, of course, detailed information from Decennial Census 2020 are not yet available, are there are some concerns with the quality of that data.

Obviously, this lack of data on households, combined with other post-pandemic trends (read, e.g., Investors, Second Homes, and AMH, oh my!) has made it extremely challenging assess the root causes of or sustainability of the housing boom over the last year.