PERSON OF THE WEEK: Arturo Garcia is chief operating officer and co-founder of Platinum Data Solutions, a provider of valuation technologies to the mortgage banking industry. MortgageOrb recently interviewed Garcia to learn more about how automated valuation model (AVM) technology has advanced over the years and why it has gotten a “bad rap” in the past.
Q: You’ve been in the AVM space for nearly 15 years. Not too many people can say that. Have AVMs changed during that time?
Garcia: (Laughing) I’m not an old guy, but I’m definitely an old-timer where AVMs are concerned. I feel like I’m being asked, “Grandpa, what was it like when you first started dealing with AVMs?”
But, seriously, the AVM segment is completely different today than it was prior to 2008. If you haven’t learned anything about AVMs since before the crisis, I can almost guarantee that you’re either spending more money than you need to or exposing your company to unnecessary risk, or both.
Advanced technology, regulations and the quality of available data have all had a huge impact on the AVM segment. It’s more advanced and nuanced than it ever was, and we know more about AVMs than we ever did. The structure of the models themselves hasn’t changed a lot, but the environment is drastically different – from the way AVMs are being used, to the attention regulators pay to them, to the data they use to derive values.
Q: What lessons has the industry learned about AVMs since the early 2000s?
Garcia: We’ve learned that AVMs are a viable, cost-effective way to obtain a value for collateral. There’s a reason that billions of AVMs are used each year. They fill a need that other collateral evaluation methods don’t. We’ve also learned how to use them responsibly and the consequences of increased risk and financial loss that follow if you don’t.
The industry learned that it needs guidelines on how to use AVMs safely and responsibly. In the early 2000s, lenders’ risk departments created their own parameters for using AVMs without much regulatory guidance. There was no consistency, very few – not to mention vague – guidelines, and even fewer requirements or standards. That changed in 2010 when the Interagency Appraisal and Evaluation Guidelines started putting parameters around how to use AVMs, when it was okay to use them and when it wasn’t.
It’s still not set in stone – I don’t think any regulatory guidelines will ever be – but lenders can no longer say to an auditor that they had no idea what to do. Smart lenders are employing best practices. They don’t want their profits to suffer because their AVM is returning unreliable values or is otherwise raising red flags and putting their institutions at risk.
Q: You say the structure of the AVM models hasn’t changed much. What has changed?
Garcia: Pretty much everything else has changed. For one, the quality and breadth of the data have increased significantly. Today we have access to data that is much cleaner and more detailed, plus there’s much more of it. For example, in the early 2000s, AVM vendors didn’t have access to MLS data, and now it’s readily available.
Another change is that today we can determine the optimal AVM for a given transaction. The difference between the right and wrong AVM can be a 20% or even 30% to 40% difference in value. Before, lenders never knew if an AVM was going to “hit” until they ran the model. Today, we can test and validate the models before they are deployed to a lender’s working environment to better understand their accuracy and expected coverage.
Platinum Data has, we feel, the most unique AVM suitability testing technology, which identifies the most suitable AVM for the task. The impact of this type of insight isn’t just about higher accuracy and coverage. It’s also about compliance. The Office of the Comptroller of the Currency and other regulators require companies to have a formal methodology for selecting AVMs. Our technology makes it very easy for literally any size lender to comply with that requirement.
A third major change is in the number of AVMs. In 2007, we probably had 10 or so models. Today, around 22 AVMs are willingly participating in Platinum’s AVM testing and validation program. We’ve got all of the major players, plus some newer and regional models.
Q: Why have AVMs gotten a bad rap?
Garcia: Before the crash, AVMs were used by subprime lenders in risky transactions. But if you think about it, it wasn’t the AVMs that were the problem. It was the way they were used and the lending decisions that were made, which had nothing to do with the valuation methodology. In the years leading to the mortgage meltdown, sure, some lenders were using AVMs when they should have been using appraisals, but we need to remember that a lot of lenders were also lending at 100% to 125% loan to value. Although they may have been using an unsuitable AVM, I don’t think there’s anyone who would disagree that the bigger issue was lending more than the property was worth.
Q: What is the future of AVMs? Will AVMs replace appraisals?
Garcia: AVMs are going to get more and more mainstream, particularly as data and analytics get more sophisticated. AVMs won’t take the place of an appraisal. There will always be a need for local knowledge and expertise, not to mention an on-site evaluation of the physical property.
We can expect to see AVMs getting more and more accurate as models get refined and modelers start taking advantage of big data. Modelers are going to start asking what other data they can use to value a property. For example, they might ask what a new stadium might mean to a city’s home values. It could mean more jobs, more people going to that stadium and perhaps a revitalized area. In the future, I expect that we will see greater use of data, such as the price of oil, natural disasters, large companies moving into the area and other factors like these, which ultimately do impact property values at some level.