by Calculated Risk on 10/29/2011 04:31:00 PM

This is a very long article from economist Tom Lawler. The introduction is below. Here is the entire article in word format (typos and all).

In looking at the “disparities” between recent Census 2010 housing data and other Census “housing” reports, many folks have focused the most on differences in the Census 2010 vacancy rates and the Housing Vacancy Survey vacancy rates, and the possible implications of these disparities on measuring the “excess” supply of housing. Indeed, Census officials, reacting to numerous stories on this issue (thank CR for this), agreed that they had “some splainin’” to do, and said that they are “actively investigating” the differences in Census 2010, ACS 2010, and HVS 2010 vacant housing unit estimates, and plan to report the results of this research at the 2012 Federal Committee on Statistical Methodological Research Conference this coming January.

In looking at the advance program for that conference, the session devoted to this topic is labeled “Evaluation of Gross Vacancy Rates from the Decennial Census Versus Current Surveys.”

Census officials have been “mum,” however, about other “honkingly big” differences between Census 2010 and both the CPS/HVS and the CPS/ASEC, including big differences in household growth by age group, and sizable differences in homeownership rates by age group. There has also been no discussion about the growth in the housing stock as measured by Census 2010 and Census 2000 compared to other Census estimates of total housing production, and how the apparent “net loss” in the housing stock related to various factors (demolitions, disasters, conversions, etc.) last decade was MASSIVELY lower than that implied by the biennial “Component of Inventory Change” (CINCH) report using American Housing Survey Data (all aspects of which are inconsistent both with decennial Census data and ACS data).

Officials have also not said much if anything about how some of these disparities are not new, but in fact have been evident as far back as at least 2000. Stated another way, the “measurement” issues are not just a “point in time,” issue, but there are time series issues as well, which make macroeconomic “analysis” of the housing market [difficult].

CR comment: The entire article is here and a must read for anyone using Census data to analyze the housing market. Hopefully we will have better data to work with in the future!

Earlier:
• Summary for Week ending Oct 28th
• Schedule for Week of Oct 30th