As published in LiveValuation Magazine
It used to be that Automated Valuation Models (AVMs) were entirely driven by public record data, but they have increasingly been made more accurate by incorporating MLS and other data sources into the valuation calculation. AVMs have been used by banks for more than a decade, but recently there has been new interest among other parties. Even brokers who were formerly skeptical of AVM performance now have an interest in them, especially as AVMs incorporate listing data and become more accurate.
What has the impact been of leveraging listing data? When CoreLogic began incorporating listing data into their AVMs three years ago, they described the impact in a press release: “Adding supplementary data has made a tremendous difference in our model performance,” said Susan Allen, Vice President of Collateral Solutions for CoreLogic. “Our models now produce results on over 85% of US homes, and we have reduced our error rates by more than 50% in the last two years.” Banks and their regulators are obviously happy with the improvement. Most recently, FFIEC regulators released a 70-page document updating guidance on valuations, and included a section affirming a role for AVMs in the lending practices of regulated banks: “An AVM may be used for a transaction provided the resulting evaluation … is consistent with safe and sound banking practices, and produces a credible market value conclusion.” (http://www.occ.gov/news-issuances/news-releases/2010/nr-ia-2010-138a.pdf)
Of course, vendors that supply AVMs to banks must be careful to ensure that the AVM is not inappropriately influenced by active listing data. The new guidance makes more than 50 references to “independence” in property valuations. A list price is typically determined by a seller and listing agent, both interested parties in the real estate transaction. It follows that an AVM based too heavily on list prices may run afoul of banking regulations. But not everyone interested in property values is a regulated bank. Perhaps the more significant recent development for AVMs is broader reach to new real estate participants.
For many years, most Realtors were hostile to AVMs and did not engage with AVM providers in any way, even as consumers were starting to access AVMs through sites such as realquest.com. Then, Zillow made AVMs ubiquitous and whether professionals liked it or not, consumers were coming to real estate professionals with Zillow’s “Zestimate” valuation. For several years, many professionals ignored this and other AVMs, discounting them as inaccurate, but now they are engaging with AVM consumers more productively.
One of the best ways to engage in a discussion regarding the AVM is to show how local market knowledge provides context regarding the accuracy of the valuation. One of the best tools for real estate professionals to do this is a product called Cloud CMA(cloudcma.com). This product has a feature called the Online Valuation Analysis, illustrated below, which compares the AVM value of comparable sold properties against the subsequent selling prices of sold properties from the MLS and to active listing list prices.
Cloud CMA Online Valuation Analysis – Sample screen
Armed with this type of report, when an AVM consumer (including buyers and sellers) approaches the professional with an AVM, the professional can show how the AVMs for comparable homes compared with both actual sold prices and list prices. Perhaps, as in one of the comparable examples above, the AVM is 34 percent above the likely selling price – or perhaps it is entirely accurate – but this type of report provides the basis to begin that discussion and immediately overcomes any assertion that AVMs provide a figure that should be taken as gospel.
Realtors can also provide more accurate context for “freeware” AVMs by leveraging professional-grade AVMs, and two players are battling for a share of the Realtor AVM market.
The largest AVM provider in the industry, CoreLogic, offers RealAVM™, which aggressively leverages MLS data in the models and analytics. RealAVM is built on the same data set and infrastructure used to provide AVMs to 18 of the top 20 mortgage lenders. CoreLogic also uses MLS data in the Listing and Market Activity Reports (LMAR). The LMAR report includes a subject property profile including both MLS listing and public record information; the MLS photo of the subject property; recently sold comparables similar to the subject property; and recent and active listings that are similar to the subject property. It also accounts for neighborhood trend information, including median sales; listing prices; sale velocity; number of transactions of properties over the past three, six and 12 months; variation of sale and listing prices for the past 12 months; foreclosure activity; and average days on market trends.
CoreLogic’s market-specific data covers 98 percent of all U.S. ZIP codes and 3,085 counties in all 50 states and the District of Columbia, representing 99 percent of the U.S. population, 97 percent of all properties (145 million), more than 50 million active mortgages, and 96 percent of loan-level, non-agency mortgage securities. In addition to the public record data, the AVM models have data detailing which properties are currently for sale, the characteristics of the property, and current asking price for the property. CoreLogic has several additional AVM products. For more information on these, visitcorelogic.com/Products/Automated_Valuation_Models.aspx.
The other player in the AVM battle is a company created by the National Association of Realtors called Realtors Property Resource (RPR™). One of RPR’s key strategic goals is for Realtor data to become central to the collateral valuation and risk process. To that end, RPR has created a new product, the RVM, or Realtors Valuation Model. The RVM launched at the MBA (Mortgage Bankers Association) conference in October of 2010. The traditional root cause of inaccurate AVMs is bad property data from public records, so the RVM improves AVM accuracy by comparing for missing or possibly inaccurate property characteristics or sales prices in public records. The RVM then uses licensed MLS listings of all statuses – including active, pending, and sold listings – to improve the accuracy of its AVM calculation. Similar to how that data can be manually evaluated in the Cloud CMA example, the RVM software automatically adjusts its valuation calculations based on list prices and sales prices of comparable properties found in MLS data. The RVM also accounts for other data in the valuation model, including local foreclosure, pre-foreclosure, auction and REO data, and other data “behind the scenes” in establishing a real-time property valuation. Realtors and Realtor appraisers can also interact with the AVM model in the RPR application, creating a feedback loop on data and model accuracy.
The RVM is available for more than 200 MLS markets, constituting approximately 36 percent of U.S. listings, and coverage is expanding rapidly. A map showing market availability based on MLS partnerships can be viewed at blog.narrpr.com/rpr-partners. Realtors, including Realtor appraisers, can access the RVM via the RPR system at no cost. Others can purchase the RVM from RPR by calling 888.914.7771.
RPR Sample Screen
There are many other providers of automated valuation models, and one can expect that MLS and other data sources will become increasingly important to all of them going forward as they strive for improved accuracy.
Two aspects of AVM development will be important to the direction the market ultimately takes: data coverage and patent infringement. CoreLogic has aggressively put up some barriers to competitors achieving listing coverage in their AVMs by signing exclusive agreements with a number of local and regional multiple listing services for using the listings for AVMs. CoreLogic also has a patent covering the most ubiquitous modeling approaches underlying industry AVMs (U.S. Patent #5,361,201, issued Nov. 1, 1994, for an automated real estate appraisal system) and announced patent infringement claims against several industry participants last year. That said, the competition is really just starting, and it will be exciting to watch how the contest for the listing-enhanced AVM market evolves.
Share this post: