The National Association of Realtors said Wednesday that, since 2007, it had overstated sales of previously owned homes by about three million. Between 2007 and 2010, actual sales were 14.3 percent less than reported, the group said, while in 2010, there were 14.6 percent fewer sales.

The association was quick to say, however, that the discrepancy - created primarily by outdated methodology and by an unanticipated drop in the number of Americans trying to sell their homes themselves - did not affect local numbers or prices over the last four years.

The group's chief economist, Lawrence Yun, attributed the discrepancy to divergence that developed over time between sales reported by multiple listing services throughout the country and those determined by a Census Bureau benchmark.

About half the revision is the "result solely from a decline in for-sale-by-owners, with more sellers turning to real estate agents when the market softened," Yun said.

Some listings appeared on more than one multiple listing service, he said, adding that "issues related to house-flipping also contributed to the downward revisions."

"From a consumer's perspective, only the local market information matters, and there are no changes to local multiple-listing-service data or local supply-and-demand balance, or to local home prices," Yun said.

Prudential Fox & Roach senior vice president Steve Storti said data used to compile its monthly HomExpert Market Report for the eight-county Philadelphia region are "as accurate as the input directly into the multiple listing service from the brokers in the market."

"There is no chance of mass inaccuracy with this kind of direct analysis, in which you can drill down to the individual transactions that make up a data point," Storti said. (Figures supplied by HomExpert formed the basis of The Inquirer's Home Price Survey in October.)

The Realtors' national data are based on estimates of all preowned-home sales, even when real estate agents are not involved. HomExpert limits data to brokers' transactions and uses exact numbers, not estimates.

For November, HomExpert reported sales in the eight-county region totaling 3,118, 9.8 percent higher than the 2,800 sold in the same month in 2010. There were 3,022 sales in October, an increase of about 3 percent month to month, HomExpert said.

Nationally, with the revisions in place, the Realtors group reported a 4 percent increase in sales in November above October's levels, 12.2 percent higher than the same month in 2010.

In a news conference Wednesday, Yun said that the process by which the association measures sales was based on 2000 census data, and that "it was time for a 'rebenchmarking.' "

Retooling the methodology has been under way for a year and a half, he said, and will be done annually.

Why did the Realtors wait so long?

Philadelphia economist Kevin Gillen said an economic model such as this one is typically modified only if its "performance has relatively degraded" compared with the way it worked previously.

"One of the primary reasons that so many models failed to predict not only the housing bubble, but the performance of the housing market after the bubble burst, is that such an event has never happened before," Gillen said.

Forecast models are built from past data, he said, and "the model will only work very well if the future looks a lot like the past."

Although observers outside the housing industry had questioned the sales numbers, it was not until February that real estate data provider CoreLogic said the Realtors' group was overestimating sales of previously owned homes by 15 percent to 20 percent.

For example, in its February report on sales, CoreLogic said unsold inventory on the market in November 2010 represented a 16-month supply, while the Realtors group said it was only 9.5 months' worth.

IHS Global Insight economist Patrick Newport emphasized that "the revisions did not alter trends," just adjusted the levels of sales.

"This does not mean that the new numbers have accurately captured the trends," he said. "We can only hope they do."

Holland, Bucks County-based economist Joel L. Naroff said the Realtors' sales revision was "in sync with the larger decline in gross domestic product during the recession that was reported by the Bureau of Economic Affairs."

"In other words, the meteor that cratered the housing market was a lot larger than initially estimated," Naroff said, adding, "And you thought the dinosaurs had problems?"