Editor’s note: Many real estate markets have slowed, and the effects are now being felt by brokers and agents who are struggling through tougher negotiations, and buyers and sellers who are aware of changing times. Among consumers, the slowdown has shaken up expectations and attitudes toward home buying, causing them to worry about future price decreases. In this three-part report, Inman News examines three areas of impact: how the slowdown is expected to influence economic growth; problems created in the property appraisal process; and consumer attitude toward home buying. (Read Part 1 and Part 3.)
They aren’t about to replace experienced appraisers, but proponents of automated valuation models say they can give their human counterparts a run for their money as lenders look for numbers they can trust in a down market.
AVMs — computer programs that appraise homes by crunching numbers from public records and multiple listing service data — are used for mortgage loan pre-qualification, loan underwriting and quality control, and to value loan portfolios and mortgage-backed securities.
Both human appraisers and AVMs can have difficulty producing accurate valuations in markets where home prices are rapidly appreciating or depreciating. But there’s evidence that AVMs are getting better at coping with swings in the market — perhaps even better than humans.
That could be bad news for appraisers who are already coping with the slowdown in housing sales, but a boon for lenders and consumers who may save time and money in the loan approval process.
Dearborn, Mich-based appraiser Terry Hanning said he believes AVMs already “take the cream of the crop, the top 20 or 30 percent” of appraisals that are the least complicated to perform.
Hanning, who has worked as an appraiser since 1986 and has owned his own firm, EDI Appraisal Services Inc., for the last 10 years, said 95 percent of his work is residential appraisals.
“We’re doing a lot of nonconforming type work, where they can’t use those AVM models,” Hanning said. EDI Appraisal Services, which employs five appraisers and covers six counties in southeastern Michigan, “gets the ones with deferred maintenance, the ones that are harder to do.”
Not only are the properties tougher to appraise, but loan officers “are trying to push you on value,” Hanning said. “But that’s true for every appraiser.”
In recent years, appraisers have gained better access to public records, Hanning said. At the same time, regulations have become tougher and more information is required for an appraisal, he said.
“An appraisal takes longer to do, and there is a constant push to lower fees,” Hanning said. “But the information is out there. Usually, if an appraiser does their job, they can find 99 percent of the sales in the area.”
Despite the proliferation of AVMs, Hanning doesn’t think they’ll ever put human appraisers out of work.
“There’s always going to be the need for communication with loan officers, banks and mortgage companies,” he said. “They want to talk to someone they can trust that can get out to the properties.”
Home sales have actually picked up a little in the last three months, Hanning said, but “it’s been a slow ride for a year and a half in my business. We’re just fighting to keep our doors open, and a lot of people are leaving the appraisal business.”
Analysts at Fitch Ratings who evaluate the credit worthiness of mortgage-backed securities issued a report in June expressing confidence that AVMs, when used appropriately, can be relied on.
Fitch’s Residential Mortgage Criteria Report explained the company’s decision to end a policy of “haircutting” — discounting AVM and other “non-full” appraisals by 10 percent to 15 percent in about two dozen regions where it considered the housing market soft or weak.
“Fitch conducted research on AVMs and concluded that the data supporting them do not significantly lag market conditions to render the data ‘dated,’ which was the basis for Fitch’s soft or weak market approach,” the report said. “Data providers that cull property information from public sources have become much faster at collecting and entering the data. Also, because home prices do not fall precipitously within very short time intervals, comparable sales data a few weeks or months old are unlikely to generate overinflated values in a soft market.”
The report acknowledged that AVMs don’t capture information about a property’s condition, and have limited access to data in “nondisclosure” states — Arkansas, Indiana and Texas — where public information about real estate transactions is limited. Another potential problem is that lenders can consult a “cascade” of AVMs, querying several to shop around for the best appraisal. But the benefits of AVMs outweigh their risks, analysts at Fitch concluded.
According to Fitch, “AVMs are most reliable when valuing typical properties in stable neighborhoods and less reliable when valuing unique properties and rural neighborhoods or disaster areas.”
The company said it will continue to review lenders’ policies for risk layering, and expects them to use AVMs selectively. Lenders who don’t pass Fitch’s review could still see AVM and broker price opinions discounted by 5 percent or more, the company said.
Robert Walker, executive vice president of collateral solutions for First American Real Estate Solutions, welcomed Fitch’s conclusions.
“I would like to dispel the myth of AVMs having data lags greater than appraisers,” Walker said.
As an aggregator of real estate data that it collects in 2,800 counties nationwide, First American RES has the ability to update its AVMs daily. To save money, Walker said, many appraisers subscribe to data on CD-ROM, which may be months old.
First American not only updates its four AVMs on a daily basis but runs blind tests in which the valuations produced by the models are checked against actual sales prices. The tests conceal the sales price of recently sold properties from the models, asking them to estimate value based on the other data used in automated appraisals.
“We subject our AVMs to blind tests on maybe 12,000 transactions a night,” Walker said. “Every morning I get a report, as does everyone on my team, on how we performed last night.”
The results are used to fine-tune the models to the peculiarities of local markets.
“We’re not applying the brakes or hitting the gas every day” but making selective adjustments on a monthly basis, Walker said. The goal is to maximize the number of predictions that fall within 10 percent of the blind sales price, and eliminate instances where the model overvalues homes by 25 percent or more.
Each AVM “has its own personality, reflective of the area of expertise of its developer, and therefore has a sort of signature,” Walker said. By maintaining four AVMs, First American can tailor products to individual markets. The company’s Value Point 4 AVM, for example, is being optimized for Florida, he said.
This month First American reported it has provided more than 250 million automated property valuations to lenders and investors since it began tracking customer usage of AVMs three years ago.
If AVMs are picking up converts, Mark A. Cannon, South Florida residential division director for Integra Realty Resources, is not one of them.
An appraiser who is often called to testify as an expert when disputes over property valuations end up in court, Cannon said AVMs are easily thrown off by details or circumstances that human appraisers are trained to spot.
“I specialize in litigation, high-end residential valuations and mortgage fraud,” Cannon said. “When the market slows down, our firm gets very, very busy with pre-foreclosures and fraud, going after other appraisals.”
When it comes to AVMs, Cannon believes the old expression, “Garbage in, garbage out,” applies.
Although they may be fine for loans with low loan-to-value ratios like home equity lines of credit, “If somebody is looking for 70, 80 or 90 percent loan-to-value ratio, you need to use other data sources” to verify that comparable sales are valid, Cannon said.
“An AVM, you push a button and pull up all these sales,” he said. “They all look like arms-length transactions. You come up with a price per square foot, multiply by square footage, and conclude.
“But what happens if you have an investor going in and purchasing condos in bulk? Lets say there is mortgage fraud and they are selling them to themselves, to family and friends at 25 to 30 percent over market. An AVM is not going to be able to tell you that.”
It takes a human appraiser to determine that sales are truly arms-length transactions, he said.
“We require every appraiser to pull up a warranty deed on all of our comps, and cross reference the buyer and seller to make sure they are not related parties,” Cannon said.
That kind of research can expose other, more mundane instances of non-arms-length transactions that can skew comparable sales data, such as the transfer of a home owned by a corporation to one of its executives.
“The loan processor, the mortgage brokers, the loan originators are the ones who get paid the big bucks, because they are the ones charging all these major fees,” Cannon said. “When the appraiser asks for his few hundred dollars, they say, ‘I can get somebody to do it for less.’ That’s why these AVMs have come about.”
But Cannon said that lenders who skimped on appraisals during the boom years set themselves up for a fall.
“When things are good, nobody questions (the quality of appraisals),” he said. “When people stop making the mortgage payments, and we have an increase in foreclosures, lenders are going to say what did we do? Why did we go to the AVMs? We should have been in control of this.”
Proponents of AVMs say they do include tools that can expose the potential for mortgage fraud.
Sacramento, Calif.-based CoreLogic offers appraisals through a nationwide network and automated tools designed to detect collateral risk and fraud issues.
CoreLogic chief economist Mark Fleming said not all properties are ideally suited for appraisals by AVMs, but that CoreLogic’s collateral risk and fraud detection tools can help flag loans that need more scrutiny.
Fleming said research shows that when AVMs err, they tend to overvalue low-priced homes and undervalue high-end homes.
“That makes sense from an intuitive sense, because low-priced homes will have (problem) conditions you don’t know about, and on the high end, AVMs won’t pick up some value-adding details,” Fleming said.
Knowing that, the question then becomes what constitutes a low-priced or high-end home.
“That’s geographically specific,” Fleming said. “Where I live in the Washington, D.C., area, a starter home is probably $300,000 to $500,000. In other parts of the country, that’s a high-end home.
Because AVMs are built to analyze specific geographic areas, their performance will vary according to those markets, Fleming said.
The challenge for both AVMs and human appraisers in markets where prices are headed up or down is that they must look backwards, to recent sales, to infer current home values.
Fleming said a recent study of subprime loans that closed in the spring of 2006 showed AVMs outperformed human appraisers in rapidly changing markets.
The study looked at about 30,000 refinance loans in California and Florida — areas which had seen rapid price appreciation until this spring, when “the ride stopped,” Fleming said.
While AVMs tended to overvalue the properties studied by about 3 percent, human appraisers were off by 12 percent.
Fleming said the study documents that “the appraiser is being pressured by the process. There are these exuberant expectations the consumer, the loan officer and broker all have,” that home values will continue to rise. “The appraiser is being asked to go in and justify the number that’s on the table with the loan officer.”
The AVM, on the other hand, “doesn’t care whether (the loan) is a purchase or a refi — you can’t hold the arm behind the AVM and start twisting,” Fleming said.
CoreLogic is planning a more comprehensive study involving hundreds of thousands of loans dating to the mid-1990s to get a better handle on how well AVMs performed during the run-up in housing prices.
In markets where prices are appreciating rapidly, there’s often an escalation of sales that provides AVMs with lots of data. In a declining market, however, sellers may refuse to come down in price, leaving homes sitting on the market longer and providing less sales data for AVMs — and human appraisers — to use in estimating property value.
The question becomes, “Do you know where it’s going to stop? I think the models don’t have enough information to know where the bottom is going to be,” said Marianne Angarola, vice president of business development at Valocity. The company, headquartered in Chicago, performs conventional, desktop and FHA residential appraisals. “I think in those markets, that’s where an appraiser’s knowledge is very useful, because not only can they utilize whatever sales have closed, but also pending sales.”
If AVMs are vulnerable to faulty or old data, there are those who say human appraisers are too susceptible to influence by their clients, who may persuade them to value properties according to a predetermined loan size or sales price rather than market value.
Last year, AVM developers put forward a report noting a close correlation between the valuation of a home subjected to a full appraisal and its subsequent sales price, suggesting that appraisers were biased by negotiations between buyers and sellers.
In examining nearly 3 million purchase loans dating to 1977, the report found appraisals were equal to the negotiated price 40 percent of the time, within 1 percent of the purchase price 60 percent of the time, and equal to or greater than the negotiated purchase price more than 97 percent of the time.
“The frequency with which full appraisal values equal or exceed sales prices, coupled with the infrequency with which they are less than corresponding sales price, is remarkable,” the report noted in an understated fashion.
The report, “Systemic Risks in Residential Property Valuations: Perceptions and Reality,” also found that appraisers were more likely to overvalue homes than AVMs.
Funded by AVM companies including Fidelity National Financial, First American RES, Fiserv CSW, Real Info Inc. and TransUnion Settlement Solutions, the report stopped short of declaring victory.
The authors said they hoped the study would “dispel the hyped controversy that AVMs render traditional appraisals worthless. Without human intervention, AVMs are incapable of assessing the current condition of subject homes.”
TransUnion offers a semi-automated property valuation tool, Appraiser Assisted Collateral Market Value, that the company touts as a solution when traditional AVMs are inadequate but a physical property inspection is not required.
“With appraiser-assisted AVM, we try to strike a balance,” said Bruce Schulkins, director of automated valuation services for TransUnion’s Real Estate Services division. “We allow appraisers to make some adjustments to the model within certain constraints.”
The tool, Schulkins said, “performs a lot of the calculations, and presents the comparable properties to the appraiser, who is someone who has local market knowledge. It lets you visualize where the comparables are on the map, and the appraiser can make the determination — should this comp take precedence over another comp, therefore affecting the valuation.”
Adjustments must be within 15 percent of the base value returned by the AVM, although exceptions can be made for decisions documented by supporting information and which pass a second review.
Although TransUnion gives lenders more flexibility in working with ARMs, the company is still working as an intermediary “to remove that fear of a bias” in an appraisal.
To ensure that appraisals don’t return a value needed by a loan originator to move a deal forward, “Lenders have been challenged by regulators to put distance between person making the loan decision and the person making the appraisal,” Schulkins said. “Whether the bias is real or perceived … that fear is removed going through an intermediary.”
TransUnion’s clients include “top tier” home equity lenders, who expect the company to drive down the time, cost, and risk of each loan, Schulkins said. Consumers may also benefit from lower fees and faster loan approvals.
Send tips or a Letter to the Editor to email@example.com or call (510) 658-9252, ext. 150.