Thursday, May 24, 2012

Lawler: Post-Census Study of Census 2010

by Calculated Risk on 5/24/2012 04:39:00 PM

From economist Tom Lawler:

Post-Census Study of Census 2010: Population, Household Count Extremely Close; Vacant Housing Unit Count Too Low

Census yesterday released some of the results from its “Census Coverage Measurement” (CCM) program for Census 2010, which is a post-enumeration exercise to assess the “accuracy” of the decennial Census numbers. While there’s a lot of “stuff” in the two CCM memoranda released yesterday, here are a couple of “highlights.”

1. The CCM (similar to the 2000 A.C.E. Revision II and the 1990 P.E.S) for the US household population (excluding remote Alaska areas) suggests that the 2010 Census had a de minimus net “over-count” of just 36,000, or 0.01%. Studies mentioned above suggested that Census 2000 had a net over-count of 0.49%, and Census 1990 had a net under-count of 1.61%. While net over/under-counts for specific race/ethnic groups or age groups were in many cases “significantly different from zero,” on balance Census 2010 seems to have been the “best” ever. (see

2. The CCM designed to provide estimates of housing unit net coverage suggest that Census 2010 under-counted the number of US housing units by 0.60%, similar to Census 2000’s estimated 0.61% under-count. The 2010 estimated under-count for occupied units was an insignificant 0.03%, below Census 2000’s 0.33%, while the estimated under-count for vacant units was 4.80%, higher than Census 2000’s 3.37%.

As with other post-Census studies, the CCM for 2010 shows “Census 2010” numbers that don’t quite jive with the previously-released Census 2010 number for total, occupied, and vacant units, which I think is because they reported Census 2010 results adjusted for “reinstated units.” (don’t ask!). However, as best as I can tell here is a table showing “official” Census housing units counts and “post-Census-study” estimates of housing units counts for Census 2010 and Census 2000.

"Official" Census Housing Unit Counts (000's)
Gross Vacancy Rate11.38%8.99%2.39%
"Adjusted" Census Housing Unit Counts (000's)
Gross Vacancy Rate11.88%9.24%2.63%

Based on the CCM results, Census 2010 understated the gross vacancy rate by about 0.5 percentage points. The CCM’s gross vacancy rate was higher than the Census 2010 GVR in all states save Alaska.

The CCM results also suggest that the US homeownership rate on April 1, 2010 was 65.2%, just a tad above the “official” estimate of 65.1%.

There’s a lot more in the report, available at

There are a couple of things worth noting. First, the higher estimate for 2010 housing units, combined with the 2000 HUCS results, suggest that the US housing stock from 4/1/2000 to 4/1/2010 increased by 15.880 million units. Other Census estimates (from surveys) suggest that housing completions plus manufactured housing put in place from April 2000 to March 2010 totaled about 16.734 million. That implies that the net loss in the housing stock to demolition, net conversions, and “other stuff” over that 10-year period by just 854,000, or 85,400 a year – an incredibly low number. I hope someone at Census plans to look at that.

Second, of course, the CCM suggests that official Census 2010 results understated gross vacancy rates, which suggests that estimates of the “excess” supply of housing on April 1, 2010 based on decennial Census results are “too low” – though, of course, adjusted estimates are still way below estimates using the obviously biased CPS/HVS data.

Adjusted Decennial Census Measures
Gross Vacancy Rate10.50%9.20%11.90%
Homeownership Rate64.20%66.10%65.20%
CPS/HVS Measures (first half average)
Gross Vacancy Rate11.40%11.70%14.50%
Homeownership Rate63.90%67.20%67.00%

CR Note: The calculated number of demolitions per year seems very low. As Lawler notes, I hope someone at the Census Bureau is looking into this. Also - this review suggests that estimates of excess vacant housing units as of April 1, 2010 were low.