The US housing data is all over the map. The disparate data comes as federal regulators grow increasingly concerned that banks with swelling portfolios of troubled loans tied to land and housing are struggling to unload some of their real-estate debt.

Exactly where house prices are is important information the markets need now. Goldman Sachs looked at 24 house price busts with declines of more than 15% since the '70s across 15 countries. On average, real house prices tended to fall around 30% and only bottomed out after six years. That's pretty much what the market is looking for.

Standard & Poor's widely followed S&P/Case-Shiller index says that house prices are now down 16%, in nominal terms, from the peak hit in the second quarter of 2006. Housing prices rose 132% from 1997 to peak by the end of the first half of 2006, the biggest housing boom in US history, the index says.

But according to available data from the Office of Housing Enterprise and Oversight, U.S. home prices have declined by a seasonally adjusted 3.9% after the index hit its peak in April 2007. That's right. Just 3.9%.

What gives? There are wide disparities in how each housing price index is calculated.

The S&P/Case-Shiller home price index is thought to be intrinsically gloomier than most housing indices because it only uses purchase prices from sales of homes. It doesn't factor in new appraisals from refinancing deals or remodeling values.

The S&P/Case-Shiller indexes are value-weighted, meaning that price trends for more expensive homes have greater influence on estimated price changes than other homes. It's notably pessimistic as it's now heavy with distressed transactions of homes backed by jumbo loans along with those backed by subprime loans. Case-Shiller also lags the market by about four months, so we're seeing prices for January.

Meanwhile, the government home-price data from OFHEO do not include jumbo and subprime mortgages as Case-Shiller does. And the geographic coverage of the indexes differs.

OFHEO's U.S. index is calculated using data from all states. OFHEO offers data for over 400 different census, state and metropolitan statistical areas versus only 20 major metro areas for the CSI. However, critics say OFHEO data is thin in hard-hit markets like Florida and California.

The quirks in housing data don't stop there. The Commerce Dept. recently reported that sales of newly constructed single family homes rose in April for the first time in half a year. Sounds great right? Wrong. Buried in the data is the fact that the prior six month's data were revised downward, so it made the comparison look a lot better.

The housing data matter in particular for banks trying to set up the right loan loss reserves. Wachovia in mid-April cited the milder OFHEO data to forecast that the U.S. housing crisis would end in "mid-2009" and that prices would fall an additional 6.8% before then, or 12.9% from their peak, the Wall Street Journal reports. Mid-April was when the Charlotte, N.C., bank reported a first-quarter loss and raised $8b.

Similarly, when Seattle-based Washington Mutual posted a $1.14b net loss for the first quarter, chief executive Kerry Killinger used Case-Shiller data to show investors the magnitude of declines in housing prices. But when WaMu assessed the current value of properties backing its mortgages, WaMu used OFHEO data, the presentation's footnotes suggest, the Wall Street Journal reports.

That's just a sample of some of the dubious data that's been on my mind. Here's more:

  • A critically important market for derivatives, the credit swaps market, now approaches an estimated $65T. But nobody in this galaxy knows what the exact figure is, instead, computerized black boxes loaded with algorithms and ‘binomial trees' conduct these trades away from the major exchanges, leaving market regulators in the dark;
  • Oil supply and demand data tracking systems in emerging markets such as Russia, India and China are either as transparent as a vat of molasses or simply don't exist, causing oil prices to whipsaw as speculators try to fill in the blanks (see prior blog "Why Suing OPEC Won't Work ").
  • Similarly, OPEC's output data is unreliable, often because member countries don't want to admit they're producing above OPEC's quotas. The markets must then rely on data from unofficial "tanker trackers" like Lloyds Maritime Information Services or Petro-Logistics SA, a teensy company that operates upstairs from a grocery store in Geneva, Switzerland.
  • Watch China's poor oil data. There are thousands of so-called teapot refineries all over China" that are left out of China's official statistics, Business Week reports. China appears to be creating a strategic stockpile of oil, but has never acknowledged it, says Eduardo Lopez, senior demand analyst for the Paris-based International Energy Agency, an affiliate of the OECD. Lopez says Russia produces "awful data" and reliable demand statistics are scarce in countries like India and Indonesia.
  • The scarcity of good global data is a key reason why it's impossible to know for sure whether the next "super-spike" in oil in the coming three or four years will be up to $200 or more...or down to $80 or less.
  • Over the past 25 years, actual federal tax revenues in any given year have averaged $150 billion (in today's terms) above or below the revenue levels that the Congressional Budget Office predicted a year in advance, analysts note, due to static, not dynamic, estimates.