What’s My House Worth: How to Estimate Property Value

A Conversation With Ann Regan

Browsing for a new home is a hobbyist activity, even for people not looking to immediately purchase. The home buying experience tends to start online, and shoppers are seeing estimates from automated valuation models, or AVMs. 

As homeowners and prospective buyers increasingly rely on online platforms for property valuation, the role of AVMs has become pivotal in shaping the real estate landscape. So how do AVMs work? What is their significance in mortgage underwriting? And how is their expanding role in marketing and consumer display?

MAICLAIRE BOLTON SMITH: Welcome back to Core Conversations, a CoreLogic podcast. I am your host, Maiclaire Bolton Smith, and I’m the Senior Leader of Research and Content Strategy with CoreLogic, in this podcast we’ll have conversations with industry experts about key topics from housing affordability, to the impacts of natural disasters on property. Browsing for homes has long been a hobbyist activity, even for people not looking to immediately purchase. I know my husband and I spent a whole year just recreationally showing up at houses before we were ready to buy, equal parts to check out neighborhoods and homes and maybe for the cookies. But we all know plenty of people that enjoy exploring, model homes complete with every upgrade, whether it be just a new kitchen or bathroom or the saunas, pools or movie theaters.

So today part of the fun involves the online shopping experience, exploring sites like Zillow, thinking of what could be. So a huge part of this experience either for the casual purveyor or for the dedicated shopper is seeing estimates from AVMs or automated valuation models. These models will assign, based on a variety of inputs, the current market value of a home. In episode 15 of this podcast, we had Sherry Clevenger on the show to explain different types of valuations, and today we’re going to continue that conversation with Ann Regan, Executive Product Management for Collateral Solutions. Ann, welcome to Core Conversations.

ANN REGAN: Well, thank you, Maiclaire. It’s a pleasure to be here. I’m a big fan of Core Conversations and yours, and I love to talk about AVM, so I’m happy to be here.

MBS:  Well, that is fantastic. And that’s what we want to hear about today. So it’s great to have you here. Why don’t we get started by telling our listeners a little bit about yourself and what you do here at CoreLogic?

AR: Sure. So I manage a portfolio of analytic products for CoreLogic. This includes things like our mortgage fraud solutions, our home price indices and forecasts, some solutions for the secondary market, and pertinent today, our AVM solutions. Prior to becoming a product manager many, many, many years ago, I actually started my career as a software engineer and worked on one of the first commercial EVMs back in the nineties. And as I’m sure we’ll get into today Maiclair, much has changed over the last 30 years, I hate to think, but yes.

MBS: Wow. Okay. Well, let’s dive in. I mean, we hate using acronyms on here. I did just define this, but AVMs, let’s talk about this. What is an AVM? How does it work? What is it used for?

AR: So sure. It’s Automated Valuation Model, as you said, one of my favorite sort of misstatements, and I do it myself, is to say an AVM model, which is a little bit redundant because that would be automated valuation model model. But as I noted, they’ve been around for quite a while. There are a number of them on the market today, but there’s a couple of things that they all have in common. So at the core of it, it is an analytical or statistical model, or usually in most cases, a series of models kind of under the hood there, using various techniques that may sound familiar, things like regression, neural networks, and more recently artificial intelligence is coming into play in machine learning. But the sort of what we would say in the modeling world, that they all have the same target variable, they all try to predict the same thing, which is to estimate the current market value of the subject property.

So by prior conversation with Sherry, this is the market value part of it. It’s not the replacement. Any of those values is like, what would an arm’s like buyer and seller pay for property.

MBS: Okay. So the other thing I want to talk about, because we’ve talked about a number of things on this podcast, we recently had the Appraisal Coach, we’ve also had our Chief Appraiser, Sean Telford from here at CoreLogic, and AVMs aren’t replacing the work that appraisers do, it’s just a tool that augments the analysis, is that correct?

AR: Correct. So I mean, AVMs are definitely, their most traditional and longest standing use has been in the financial industry for risk management purposes. And AVMs we’re heavily relied upon prior to the housing crisis in the mid two thousands. And while they didn’t really cause the crisis, they were certainly identified as a contributing factor. So as a result, today lenders will still use AVMs in underwriting loans, but it’s a heavily regulated, a lot of regulation coming out of the last housing crisis. So their use of those AVMs for that purpose is heavily regulated, they’re required to do extensive testing of the AVMS to understand how they perform, something that was not happening in the early two thousands.

And also they’re required to get an actual property condition inspection. So not a full appraisal necessarily, but someone’s got to sort of do a dry buy, make sure that the house is standing and there’s not a hole in the roof. And if they sort of combine that AVM with that property inspection, they can use it for things like home equity lending, refi where the risk is lower, but really in most purchase transactions, the full appraisal is still warranted and used by most lenders.

MBS: Okay. Okay. So just kind of streamlines the process in certain situations where the risk is lower.

AR: Exactly. I would also note too that there’s sort of this interdependence between AVMs and appraisers and appraisals. One of the inputs to AVMs is appraisal data, along with property data, MOS data. So, they’ll use things like property characteristics and whatnot. Appraisals can de used to train the models themselves as we move into this machine learning. Because the appraiser knows the condition of the property, they’re going to have a really expert opinion about what property, the market value properties. And we try to encode that opinion in the AVMs. And then on the other side of the coin is, appraisal management companies will use AVMs to determine the complexity of a property to assign an appraiser or to do quality control on an appraisal. So as I said, they’re sort of, there’s this mutual dependence here that will result in neither one supplanting the other, I think anytime soon.

MBS: Well that’s great. And I appreciate that background and that helps us dive into what we’re going to talk about today. And if we look a little bit more at use cases, you mentioned a little bit that things go beyond mortgage underwriting. What are some of those other expanded use cases and what does that mean for consumers? And you mentioned refi and how that can streamline things a little bit, from the consumer perspective, how are these useful?

AR: Well, and you mentioned this in your introduction Maiclaire, I think what I’ve certainly observed in the last five, eight years here is one of the fastest growing, what I would say, alternative use cases for AVMS outside that traditional lender use case, is marketing and consumer display. Consumers are savvier than they have ever been in terms of wanting to understand what the value of their home is. It’s a huge asset, your biggest asset. And I think, again, maybe is a result of the housing crisis, 15, 20 years ago, people understand. And even what we’re seeing today with market values, changing so quickly that I think consumers are much more attuned and wanting to have a better understanding of what’s going on with their property. So they’re spending more time as you noted looking at consumer sites, real estate sites and really trying to get an understanding of what their properties are worth.

But yet there is a difference between traditionally these values that you may see on one of those consumer-facing sites, which we would term as more of a marketing AVM and they tend to sometimes overvalue the value of a property. They send to overstate it a little bit. Again, it’s a big asset, people like to think their home is going up in value. And that they’ve got a great asset there. Now, lending quality AVMs, which are used by the lenders, in contrast, tend to be a bit more conservative. The lenders are lending money, so they tend to be a bit more conservative. And they’re subject to, as I mentioned earlier, a lot more testing and auditing for consistency.

So what you end up with then is this sort of valuation dichotomy where you’ve got the consumer seeing one value, thinking a home is worth a certain value when they’re going to sell their home or buying a new one, and then as you work your way down the loan pipeline, now you’ve got, the lender potentially using a different AVM, and now you’ve got this conflict that leads to some consternation on the part of all parties.

Consumers may be familiar with a similar issue with credit scores, things like Credit Karma and other places now, I know I get my credit score on my credit card statement every month. And I think, “oh, good,” if I want to go buy something, that’s the score the lender will use, nope, they’re using a different version of that score. And again, you can have these similar issues where they don’t always match and then you’ve got problem.

MBS: That’s great. And I’m glad you made that analogy too because I think credit scores are something everybody’s familiar with, we all can see them every time we log into our bank accounts these days, we see them on our credit card statements. But when you have something, doing something like applying for a loan, they have this other magical number that you can’t ever see, and why is it different? So I think that’s a really, it’s an analogy that people can really relate to. I know I am not the only person on the planet that logs into Zillow on a regular basis and sees the value of my home. And that value of my home is probably a little inflated from what the actual value of my home would be if I were to purchase my home today or if I were to sell it. And I know I often go, “well, why are they different?” And I think from, again, from that consumer perspective, and you talked about this a little bit too, how can these conflicting values which serve different purposes, how can it be reconciled?

AR: It’s an excellent question, Maiclaire, and it’s actually one we recognized a few years ago here at CoreLogic, as we would get lenders would come to us, consumers, we sort of observed this mismatch between these two values. So we actually embarked on a huge initiative to build a completely new AVM from the ground up to address this need that we know is candidly, I think this is where the industry is going. We’re going to see more and more of these, what I would call non lending use cases. And therefore, we really felt it was important to have an analytic that could serve all those use cases. So what we did was build this model, using things like artificial intelligence, machine learning, all our CoreLogic data assets, and what we’ve done is we’ve built a single model that we can basically configure or tune for those different use cases.

So for the more traditional lending and risk management use case, we’ll optimize the AVM for accuracy, because that is the paramount objective for lenders. But for consumer and marketing use cases, we’ll optimize it more for coverage, trading off a little bit of the accuracy, because it is also frustrating as a consumer to log on to Zillow and not to be able to get a value for your property. So AVMs don’t work on a hundred percent of the US housing stock for the most part. So I think it’s important to be able to provide values that give the consumer some idea of the quality and current valuation of their property without sacrificing that accuracy for the lender.

MBS: So I hear that too, as a bit as the accuracy is super important for the lender for obvious reasons, because it’s a risk management issue, they’re lending money, but accuracy is viewed as maybe not quite as important for the average consumer. It’s a number that’s a guide. It’s something to give them a high-level overview of what the market value of their home is versus the “you can lay your money on this, that this is the value of your home.” Is that a simplistic way of looking at it?

AR: Correct. It is. And in our case where the AVM provides a value in both use cases, it’s the same. So if you’ve got a house that the lending model says it’s worth 250 K, and then you want to display that to consumer, it’ll still say 250 K, the difference could be, there’ll be another house, maybe a harder to value house that the lending AVM would say, “Hm, this one, I don’t have enough data or I have conflicting data and I can’t render a value for this” because AVMs will do that, if it feels like it just doesn’t have enough information to give you an answer, it doesn’t give you a wrong answer. It’ll just give you no answer.

But for the marketing and in a consumer display use case, it’s like, “well, I can still display this answer and it’s going to still be within the realm of reasonableness in terms of the value of that property.” It’s probably not going to be quite as accurate as that lending quality use case because that’s what when you were measuring the accuracy for AVMs, we’re looking at how close the AVM value comes to that actual sale price when the house is sold. That’s the term. And we just measure, is it 5% away is a 10% away. So again, that band, the valuation band is going to be a lot narrower for that lending use case, than that consumer marketing use case.

MBS: And I think it’s pretty easy to appreciate that too, because it’s the complexity of it, and the importance of why it needs to be so precise and accurate for the lending purpose. So thank you for diving into that with me. So you talked a little bit about what we’ve done here at CoreLogic, I know there’s something called Total Home Value to the Power of X, and I think that’s what you were started to allude into, what is the benefit of this over other AVMs, either to the consumer or the lender and this whole topic that we’ve been talking about, how it’s important to both, but maybe a little different to both. Can you talk a little bit about Total Home Value X?

AR: Sure. And I would say there’s a number of benefits, but in terms of this benefit of this single model, there’s benefit to both the consumer and the lender. As I mentioned, you’re going to end up getting valuations for a higher percentage of homes, which is beneficial to the consumer, gives you the chance to monitor the value of your home. Candidly, it’s also beneficial for the lenders, who are usually the ones that are providing these values ,or other real estate sites, they’re trying to create leads, usually for mortgage loans, but sometimes for other purposes, loan consolidation or other things. So the benefit to the lender of that higher, what we would call coverage, is they get more leads, because they are able to value more properties. So you get that benefit in terms of actually being able to see the value, but probably the bigger benefit comes for both parties when you get this consistency evaluation.

So, if I’m a lender ABC and I serve up a marketing AVM to a consumer on my website, because I want to draw a consumer into apply for a mortgage loan, they’ll be delivered a value. And then when that same lender, now pulls a lending quality AVM potentially to do maybe a pre-approval of a loan for that consumer to buy or sell a property, they’re going to be using that same model, and get a more consistent valuation. So you don’t run into these issues on sort of the end of the underwriting life cycle, where you might have a consumer thinking a home is worth $500K and now either that lending AVM says, “oh, it’s only worth $450K” or an appraiser, again, as we talked to a lot of times, appraiser will come into play here and they’ll look around at the comps and go, “yeah, no, this is only worth 450 and we can’t justify that.” So it sort of smooths over, if you will, that life cycle all the way from the marketing consumer at the starting point, right through underwriting, and allows a consistency evaluation.

MBS: And the consistency, I think is key, so does it eliminate or make up for the appraisal gap that we hear about sometimes too, is that potential? We talked about that a little bit on some of our previous podcasts.

AR: It can certainly help there. I would say in this current market, offers are coming in at 10, 20% over the listing price, so that gap for appraisals is being caused by the fact that appraisers look around at comps and what they’re being listed for and saying, “I’m not sure I can support that price.” Now, AVMs, are in terms of being able to keep up with those market changes.

For example, we use our home price forecast as an input to our total home value. So that allows us to look ahead to where we think those prices are going, and allows us to do a pretty good job of keeping up with those market changes. So I think it can help a little bit there, but in these kind of markets where prices are going up so steeply, it’s a tough challenge to overcome because I think the lenders are a little bit concerned about to lend you money based on that. And if there is some default in the future, their recourse is to sell that asset, and if that asset was overvalued by 10 or 20%, there’s a possibility that then they’re going to lose that money, so that’s a tough challenge, the appraisal gap, maybe a little but probably don’t solve it.

MBS: And we’ve talked about this a number of times too on this podcast, and I’ve talked about examples of how we bid 150 to $200,000 over asking price on a house, just because the market is, and this was three years ago when we bought our house, it’s just the market drives that, so I can understand how it can’t make up for all of that, because there are market fluctuations and demands that happen, that I don’t think any model could ever account for given it’s human behavior.

AR: And you want to have a place to live. You almost are forced into it, if you want to be able to buy a house and have a place to live. Yep.

MBS: Yeah, no, definitely. You’ve talked a little bit about machine learning and some of the techie side of adding to the world today. I like to sometimes finish off these podcasts by saying, “if you had a crystal ball” and everybody tells me they don’t have a crystal ball, but use your imagination, if we had to think of what does the future hold for AVMs, is it machine learning and kind of those tech driven world value adds, is that where we’re going for AVMs? Where do you see the future going?

AR: Well, it’s a great question, Maiclaire. And I think it’s been really fascinating for me over the last, I don’t know, 10 years or so of managing our AVM products at CoreLogic, to watch the improvement in, and actually just even since the housing crisis, to see the significant improvement both in terms of that accuracy measure, as well as coverage. When I built the first AVM, 30 years ago, it was an Excel macro that I hand delivered quarterly on floppy disks just to give you a feel for it. But now as I look ahead, we’ve got vast data assets that are increasing every day, we’re adding things like image processing of photos of the property to be able to extract condition.

And maybe we’ll get to the point where consumers can take their own photos, and we’ll save them the cost and inconvenience potentially of having someone come in the house, which has not been popular during the pandemic. So we can see more data, will always be a race to get more data and more input, which is what can make the models more predictive. But the other key aspect here, and we certainly leverage it ourselves, is cloud computing, because when you get more data and you have more statistical models and more sophisticated models, you need more CPU to be able to process all that in any sort of reasonable time and cost. So, as we’ve made the move to the cloud, we’re able to process whatever petabytes, terabytes of data in a reasonable amount of time.

So I think we certainly see that happening. And then as AVMs improving that performance, I think we may see broader adoption of them. We’re expecting some more regulation around AVM use to come out at the end of this year. And we’ll see what that says in terms of where and how lenders are able to use AVMs to support that underwriting process. But I think, as I said, as the performance continues to improve much like credit scores where, if you think about it, 30 years ago nobody knew what it was or what it meant to them. But I think we’re in the midst of that transformation as consumers have a better understanding of AVMs and what the impact on their financial situation is, and lenders get more comfortable with their performance, we’ll see them more broadly used.

MBS: Well, that’s really helpful. And that’s really interesting. And I think good to hear that regulations are coming into place as well too. And I know we’re seeing that on the appraisal side, we’re seeing that on every part of the property ecosystem. And I think it’s just something we need to keep our eye on as the future continues. And maybe we’ll have you back in the future as well, to talk about progress on things, and I think that would be great. But this has been great. It’s been so interesting and I’m sure relevant to so many of the people out there listening, because I would challenge anybody to tell me they haven’t looked at a real estate site in the last couple of months to look at either the value of their home or another potential home, just as they may be window shopping or internet shopping as it may be because I think this is something that a lot of people do quite often. So it’s good to have just that little bit of background information. So thank you for that. And this has been great.

AR: My pleasure, Maiclaire, and yes, and as you know, not only is it fun to look at your own home, it’s fun to look at your neighbor’s home.

MBS: Indeed.

AR: It’s a popular past time, “what’s going on over there next door.”

MBS: That’s very true. Very true. So, well, thank you so much for being here on Core Conversations at CoreLogic podcast. For more information on the property market and the housing economy, please visit us at corelogic.com/intelligence.

The CoreLogic statements and information in this blog post may not be reproduced or used in any form without express written permission. While all the CoreLogic statements and information are believed to be accurate, CoreLogic makes no representation or warranty as to the completeness or accuracy of the statements and information and assumes no responsibility whatsoever for the information and statements or any reliance thereon. CoreLogic® is the registered trademark of CoreLogic Solutions, LLC.