Though many people don’t realize it, billions of automated valuation models (AVMs) are used each year to value properties. In fact, AVMs are the most frequently used tool for valuing residential properties. In comparison, roughly 2.5 million traditional appraisals are completed each year. Yet despite their popularity, AVMs still get a bum rap in some quarters. It’s an undeserved reputation, in my opinion. And I’ll tell you why.
Platinum Data is the largest independent reseller of AVMs in the industry. To give you an idea just how substantial the volume of AVMs is industry-wide, consider that one of our vendors reports volume of over one billion AVMs each month. That’s over 12 billion AVMs each year for that vendor alone. Add that to all of the AVMs from other vendors and you’ll see: this is probably the largest, least talked-about market in the industry.
Just who is using AVMs? I estimate that at the very least, hundreds of millions of the billions of AVMs run each year are performed on behalf of lenders and servicers. The remaining are used by other financial service providers, such as auto lenders and other creditors. In the mortgage industry, AVMs are used to estimate the value of a residential property for everything from home equity lending to pre-qualifications for first trust deeds; from appraisal due diligence to valuing servicing portfolios.
The truth is, AVMs are used so frequently because they work. They’re good at what they do, which is to provide a reasonable estimation of a property’s value, quickly and at a very low price. Yet despite the fact that AVMs are used millions of times every business day, very few in the mortgage industry are discussing them. They’re often treated like the industry’s dirty little secret. But why?
For one thing, there are a number of misconceptions about AVMs. A lot of conclusions about AVMs were drawn at a time when they were being misused as replacements for appraisals in securing a first trust deed. Information on AVMs was presented out of context by organizations that incorrectly believed AVMs might threaten their business. This in turn perpetuated some myths about AVMs. I’m not interested in pointing the finger or figuring out why certain organizations wanted to undermine AVMs. My objective is to shed light on the most relied-upon product in its category, and to uncover how AVMs are being used–or in some cases, misused–so we as an industry can improve, continue recovering, and originate and service healthy loans.
It’s time to re-evaluate the industry’s position on AVMs. If we don’t acknowledge AVMs as a critical and often-used component of certain mortgage processes and transactions, we are missing an opportunity to optimize them. The more we know about AVMs, the more we can bring them into the conversation, learn about them, improve them and leverage them to further the industry—not to mention improve our bottom lines.
At a time of shrinking volumes and constricting margins, it’s high time for us to separate AVM fact from fiction. Here, I’ll evaluate the five most prevalent AVM myths and get to the truth of the matter.
Myth #1: AVMs Are Unreliable
Several experts, including renowned valuation expert Bill King, Platinum’s senior vice president of valuation solutions, have estimated that AVMs as a whole are likely to have a standard error of about eight percent. That means an AVM will return a value that is within 16 percent of the actual value of a given property, 95 percent of the time. Before we jump to a conclusion on whether or not we can deem eight percent reliable or unreliable, let’s put that percentage into proper perspective. Studies have found traditional appraisals can have a standard error of roughly 13 percent, which means appraisals, on average, are within 26 percent of the actual value of the property, 95 percent of the time. Again, this isn’t to point fingers at one form of valuation or to judge one product as superior over another. Appraisals are considered to be the gold standard in the industry. It’s simply to put that eight percent in context. There are other factors to consider. Let’s take a closer look at how standard error is determined.
A standard error rating is an average. Each AVM and each appraiser will achieve a better rating, like two to three percent, on certain properties, and a worse rating, like 15 to 18 percent (or higher), on others. That’s because each AVM performs better within certain geographies and on certain properties. There’s no such thing as a one-size-fits-all AVM, just as there is no such thing as a one-size-fits-all appraiser. AVMs have a two to three percent sweet spot, just as appraisers do. Regulations mandate that we utilize appraisers with geographic competency. We’re required to use the most suitable appraiser for the property. The question is: why aren’t we required to use the most suitable AVM for a given property, as well?
Myth #2: AVMs Shouldn’t Be Used Because They Cannot Inspect a Property
An AVM cannot conduct a physical inspection of a property. Most in the industry understand this limitation, and regulators have taken it into account when developing AVM usage guidelines. But this doesn’t necessarily mean that the mortgage industry shouldn’t use AVMs. It simply means that AVMs shouldn’t be used exclusively to evaluate collateral in transactions such as first trust deed mortgages, but instead for servicing portfolios, home equity lending, appraisal underwriting and quality control, and prequalifying first trust deeds. It’s a matter of common sense. You’ll want an inspection on higher risk transactions. The risk associated with a brand new, high loan-to-value first trust deed is greater than the risk, for example, on one of 10,000 loans in a servicing portfolio of first mortgages, particularly when those mortgages are secured by collateral that was recently valued with a traditional appraisal.
An AVM’s inability to inspect a property doesn’t impair its ability to perform any of its core functions within the mortgage sector. We simply need to remember where it makes good sense to use an AVM and where it doesn’t.
Myth #3: AVMs Don’t Do a Good Job of Taking Neighborhoods into Account
A lot of people have criticized AVMs, saying they can’t fully characterize, differentiate and account for neighborhoods. While this may have been true in the past, times have changed. The AVM of today isn’t exactly your father’s AVM.
AVM technology has progressed quite a bit since the days when AVMs were severely misused in our industry, when they were used to value collateral for first trust deeds with loan-to-value ratios that exceeded 100 percent. Thanks to the work of the industry’s top minds in valuation analytics, AVM providers are now able to construct more and more meaningful geographical boundaries within their databases. In fact, they’re actually starting to come out with neighborhood-specific and community-specific models. They’re not simply drawing boundaries as a circle or a box. They’re establishing what are called “neighborhood polygons,” which essentially means they’re able to create neighborhoods that have jagged lines and specific boundaries.
While an AVM isn’t a person who’s lived in a neighborhood for the past two decades, it is a technology that can be programmed to learn rules and execute those rules flawlessly, 100 percent of the time. And that technology is improving every day.
Myth #4: AVMs Rely on County Data, and County Data Isn’t Reliable
AVMs do rely on county data. However, in aggregate, county data is generally reliable. A lot of folks who say that county data is unreliable argue for the use of Multiple Listing Service (MLS) data. My issue with using MLS data is that listing agents are working for their clients, not for lenders. That’s not to say that Realtors are dishonest, by any stretch. Realtors are qualified professionals who play an essential role in the housing industry, as well as in the American economy. While Realtors as a whole have zero interest or intention of deceiving lenders, their interests are focused on keeping their clients satisfied. That sometimes means listing homes for as high a marketable price as possible, not necessarily at a price that reflects the true market value of the property.
If Realtors made a habit of marketing properties using the bare bones, hand-to-heart truth, they’d probably be doing their clients an injustice. While they may not hide the ugly truth, they probably aren’t baiting the hook with it, in the same way that Coca-Cola’s ad agency won’t recommend an ad campaign that leads with the number of calories or lack of nutritional content in a can of Coke. It’s going to put Coke’s best foot forward, just as a Realtor will do with the client’s property. You’ll find out about the 140 calories soon enough. The ad agency wants you to know how Coke tastes or how it will make you feel, before you start visualizing those extra pounds. It’s not dishonest. It’s just common sense marketing, and that goes the same for marketing homes.
While there’s no such thing as 100 percent reliable data from the county recorder or any other source, I certainly don’t believe that marketing copy provides a more accurate or complete depiction than historical record. The fact is, county data is pretty darn reliable, and getting more so every year.
Myth #5: AVMs Rely on Outdated Comps
Over the years, I’ve heard it argued that AVMs aren’t credible because they use outdated comparables. It’s an old argument, and one without much merit.
Of course it’s possible that AVMs may sometimes need to rely on dated comps, but then again, so do appraisers. If an AVM doesn’t have access to current comps, neither will any other form of property valuation, and that includes a full appraisal by an appraiser. Everyone is fishing from the same pool. In any given neighborhood at any given time, comps may be abundant and current, or sparse and dated. Just like most appraisers, quality AVMs will not choose outdated comps if relevant, current comps are available. The alternative is to use comparable sales that are outside of an acceptable area, which is, of course, the same challenge faced by appraisers.
It’s certainly not ideal to use dated comps. But in my opinion, it’s even less desirable to use comparable sales outside of the subject neighborhood. However, when these scenarios occur, they are industry-wide challenges that are not unique to AVMs.
AVMs are practical, useful tools when used in the correct situations
The bottom line is that when used correctly, AVMs are viable tools for providing a fast, accurate, and very economical estimation of collateral value. Given the fact that the mortgage industry is using hundreds of millions of AVMs each year, we need to come together and open the lines of communication. Lenders and servicers need to share information on how they are using AVMs. And technology experts need to move forward, innovate, and introduce new tools that will bring positive change into this segment. AVMs aren’t going anywhere. They will continue to impact the industry, whether we acknowledge them or not. If we want to control the impact AVMs have on our industry and on our individual bottom lines, we need to start by acknowledging their use, their prevalence and their role in our operations.
In next month’s installment, I’ll share the information I’ve gathered on how lenders and servicers are using AVMs in the current market, provide concrete specifics on how big an impact AVMs are having on the bottom lines of lenders and servicers, and list specific steps that lenders and servicers can take to save tens and hundreds of thousands of dollars—perhaps even millions of dollars—simply by choosing and using the right AVM for the scenario.
About The Author
Phil Huff is CEO at Platinum Data Solutions. Phil is a CEO with a history of growing companies whose technologies revolutionize manual mortgage processes. As co-founder and CEO of eLynx, Phil built the management team, grew recurring revenue to $15 million, and orchestrated the company’s sale to American Capital for $40 million in 2004, five years after the company’s launch.