by Bill McBride on 1/20/2017 05:25:00 PM
Friday, January 20, 2017
Earlier: Lawler: The Household Conundrum, Part I: The CPS/ASEC Data
From housing economist Tom Lawler: The Household Conundrum, Part I Continued: The CPS/ASEC Data for 18-29 Year Olds
Many analysts have noted that estimates for the number of households from the CPS/ASEC appear to have overstated the “actual” number of households, with most of the overstatement being estimates of the number of “young adult” householders. If that were the case, then obviously the CPS/ASEC also understates the “other” living arrangements of young adults (e.g., young adults living with parents, other relatives, or non-relatives. Below is a table comparing the shares of 18-29 year olds by household relationship as estimated by the CPS/ASEC to those from Decennial Census 2010.
Note that the CPS/ASEC is a “household” survey, and not a survey of the entire population. Note also that the CPS/ASEC counts students living in college housing as living at home, while the Decennial Census counts students living in college housing as living in “group quarters” (and as such are not counted in the “household” population. On April 1, 2010 Decennial Census estimates show about 2.5 million people living in college housing, with the vast majority (though not quite all) of those people being between 18 and 29 years old. As such, I have added the number of 18-29 year olds living in college housing to the Decennial Census estimates of the “household” population in order to make Census data comparable in definition to the CPS/ASEC data.
|Share of 18-29 Year Olds in the "Household" Population (CPS definition) by Relationship to Householder, 2010|
As the table indicates, CPS/ASEC estimates of the share of 18-29 year olds that were householders in 2010 is almost 3 percentage points higher than estimates from the Decennial Census. CPS/ASEC estimates of the share of 18-29 year olds living with parents, in contrast, are 1.4 percentage points below Census estimates, and the share of 18-29 year olds living with “non-relative” (roomer or boarder, housemate or roommate, unmarried partner, or “other”) is 1.6 percentage points below Census estimates.
While these differences may not seem that large, the (1) are statistically significant; and (2) imply that the CPS/ASEC estimate for the number of 18-29 year old householders is about 1.6 million “too high.”
Even more startling, though not as important, are the CPS/ASEC estimates for the number of householders 15-17 years old – about 210,000 – compared to the Decennial Census estimate of 28,297.
On the next table is a comparison of the number of households by age group from Census 2010 (“official” numbers) compared to those of the CPS/ASEC (adjusted to reflect Census 2010 population count “controls).
|Household Estimates by Age Group, 2010 (000's)|
|Decennial Census||CPS/ASEC (Adjusted)||% Difference|
It is not readily apparent either (1) why the CPS/ASEC estimates for the number of “young” adult households are overstated by as much as they are (or why the overstatement is a function of age); or (2) why this overstatement showed up for the first time (albeit by not nearly as much) in 2000.
Some believe related to the significant increase in the non-response rate in the CPS that began around the same time (1994) that the CPS shifted computer-assisted interview data collection system (CATI), combined with higher non-response rates for younger adults relative to older adults. Others wonder whether the differences are somehow related to the “imperfect” sampling frame of the CPS/ASEC. Personally, I don’t know the answer.
If, in fact, the CPS is not getting the “living arrangements” of young adults “right” – overstating both the share of young adults that are householders and the share of young adults that are homeowners (see LEHC, 1/18/2016), then it would not be surprising that other statistics about young adults from the CPS (employment, income, etc.) may also be “off.”