by Calculated Risk on 5/17/2011 03:45:00 PM

CR Note: A key piece of data for the housing market - and the U.S. economy - is the current number of excess vacant housing units.

Unfortunately it is very difficult to get a good handle on this excess supply (it is large, but how large?). Both Tom Lawler and I are hopeful that we can arrive at a more accurate estimate using the Census 2010 data to be released this month (the estimate will be as of April 1, 2010).

Please excuse Tom's punctuation - but he has been arguing for better housing data for years - and he is clearly frustrated!

By Tom Lawler: The “Excess Supply of Housing” War: Is the 3.5 Million Estimate “Gold” (Man, No!); or Can You Take the 1.2 Million Estimate to the (Deutsche) Bank?

A few weeks ago Goldman Sachs’ analysts made headlines by arguing that the “excess” supply of housing, or actually the number of US housing units sitting vacant “above and beyond normal seasonal and frictional vacancies,” was “about” 3.5 million. This week Deutsche Bank analysts estimated that at the end of 2010 there were about 1.2 million “excess” vacant housing units in the US. Both sets of analysts relied heavily on data “provided” by the US Bureau of the Census in deriving their “estimates.” And, to the best of my knowledge, neither set of analysts was comprised of imbeciles. Yet jiminy cricket, those are pretty huge differences with massively different implications about the prospects for the housing markets and home prices over the next few years!!!!! And the major reasons for these differences? You guessed it, massively disparate sets of data from different areas of the Census Bureau on US housing!!!!

As readers probably guessed, Goldman analysts’ estimates are based on what are almost certainly flawed and biased estimates of the occupied and vacant housing units from Census’ quarterly “Residential Vacancies and Homeownership” Reports, commonly referred to as the Housing Vacancy Survey (HVS). While there had already been strong evidence that the HVS dramatically overstated both homeownership rates and vacancy rates prior to the decennial Census 2010 (from the ACS), the incoming data from Census 2010 pretty much confirm that the HVS data has serious biases, probably related to serious sampling issues.

The Deutsche Bank analysts’ estimates are based on the Census 2010 gross vacancy rate versus a weighted average of the Census 1990 gross vacancy rate (weighted 75% for pretty flimsy reasons) and the Census 2000 gross vacancy rates (weighted, of course, 25%). (DB analysts “walk forward” the April 1 Census 2010 estimates to the end of 2010 using what I believe are “questionable” assumptions about household formations and net demolitions). Census 2010 has not yet released national data on vacant units by status (for rent, for sale, etc., and it was sorta weird that DB analysts didn’t just wait a few weeks to do a more “rigorous’ estimate based on more complete Census 2010 data.

DB’s piece includes a decently long and not “too” bad discussion (though with many errors and/or omissions) of the multiple and disparate sets of data on the US housing stock derived from various surveys done by different areas in the Census Bureau.

What is disturbing, of course, is not necessarily that different sets of analysts can come to different sets of conclusions when analyzing US housing data. Rather, it is that there are multiple and conflicting “official” sets of government-produced data on the US housing stock, with little or no discussion from government officials/analysts are which – if any – dataset should be used by analysts to estimate the “excess” supply of housing in the United States

Earlier:
• Housing Starts decline in April
• Industrial Production unchanged in April, Capacity Utilization declines slightly
• Multi-family Starts and Completions, and Quarterly Starts by Intent