Price discovery

Do real estate agents accurately report home sale prices?

Three professors at Florida Gulf Coast University's Lutgert College of Business in Fort Myers sought an answer by comparing home-sale prices reported by real estate agents on a multiple-listing service, better known by its acronym MLS, with final settlement statements called HUD-1 at closing.

Professors Tim Allen, Kenneth Lusht and Shelton Weeks found that the average of all price errors by the MLS data they studied was $10,277, or a difference of 4.9% from the prices reported on the settlement statements at closing. In their findings published recently in the Appraisal Journal, the professors found price differences in nearly 9% of their sample ranging in size from -1.1% to 21.4%.

This research is important because residential appraisers often rely on MLS for property information and transaction prices. In addition, Realtor organizations publish monthly and annual reports of prices that many businesses in related industries rely on to make decisions. “To my knowledge this is the first paper that measures the magnitude,” says Lusht.

“I was working with MLS data and found some things that were fishy,” says Allen. “I was seeing prices that were much higher and lower than asking prices.”

Allen won't say which MLS data he was using except to say that it was somewhere in the Southeast U.S. “MLS data is proprietary,” he says. (Both the National Association of Realtors and Florida Association of Realtors declined to discuss the findings.)

Meanwhile, Allen persuaded two undisclosed banks with multiple branches to share private HUD settlement statements so he could compare the actual transaction prices with those reported by agents. Banks had an interest in finding this out, too: “They want to know if the appraisals are valid,” Allen says.

The work was tedious because Allen had to manually match the settlement statements to each transaction reported by the MLS. In the end, he was able to match 400 sales prices with HUD and MLS information from 115 listing brokers over a four-year period from 2004 to 2008. “I worked six to eight months before I got enough to use,” says Allen.

The professors found that the errors were not random because prices were more likely to be overstated than understated. The average overstated price was $14,038, or 6.7% more than the prices stated on the HUD statement. The average understated price was $877, or 0.42% of the HUD-reported prices. The largest overstated error they found was by 21.4% and the largest understated error was 1.09% of the HUD price.

The price errors in the MLS were not limited to the price levels of the homes or other characteristics such as property age, number of bedrooms or bathrooms. The research covered properties that cost as little as $51,500 and as much as $2.15 million.

However, the professors found that pricing discrepancies between the HUD statements and MLS were much more likely to occur during a market downturn. For example, an MLS error was 19.2% more likely in 2007, the first year of a sharply declining market, compared with 2006.

The professors speculate two possible reasons why these errors occur. Some brokers may intentionally inflate the prices for which they helped sell homes, making it appear that they negotiated a higher price for their clients. But a more benign reason is that MLS-reported prices may reflect contract prices before adjustments made prior to final contract settlement.
For example, in a falling market the prices may have been lowered to keep deals from collapsing or to account for property damage or broken appliances.

Fact is, there's no incentive for agents to change the sold price in the MLS database after adjustments have been made at settlement. “They're not necessarily lying about those results,” Allen says. Anyhow, real estate commissions are calculated from the HUD statement, he notes.

The professors caution the data from a Southeast MLS preclude making more general assumptions nationally. “I couldn't say it holds true in Kansas,” Allen says.

And the study's relatively small sample of 400 transactions limits the conclusions. “I'd rather have 40,000,” Allen says.

Still, the study serves as a cautionary note to appraisers to consider other sources of information for appraisals, the professors say. “It's a caveat, a heads-up,” Allen says. “Verify the information before you rely on it.”

Follow Jean Gruss on Twitter @JeanGruss