RegFi Episode 39: Technological Innovation in the Mortgage Industry

Orrick RegFi Podcast | Technological Innovation n the Mortgage Industry

RegFi Episode 39: Technological Innovation in the Mortgage Industry
32 min listen

Bob Broeksmit, President and CEO of the Mortgage Bankers Association, joins RegFi co-hosts Jerry Buckley and Caroline Stapleton to discuss the transformative impact of technology on the mortgage industry. Bob highlights the industry’s tradition of innovation, such as algorithmic underwriting and credit scoring, which has streamlined loan origination and reduced bias in lending decisions. The conversation also delves into the current interest in generative AI and how the Mortgage Industry Standards Maintenance Organization (MISMO) is creating a forum for market participants to discuss AI’s potential applications, guardrails for implementation, and the associated regulatory challenges.

Links:

Episode Transcript

Jerry Buckley:	Hello. This is Jerry Buckley, and I am here with RegFi co-host Caroline Stapleton. Our guest today is the president and CEO of the Mortgage Bankers Association, Bob Broeksmit. Bob has had a distinguished career in mortgage finance. Immediately prior to assuming the leadership of the MBA, Bob served as president and chief operating officer of Treliant Risk Advisors, a consulting company based here in Washington, where we both were involved. Bob, we're so glad to have you with us. And as you know, the premise of this RegFi podcast series is that financial regulation driven by technology, will change more over the next 10 years than over the last 50. There has been no sector of our financial services industry that is more important than the mortgage banking industry. Mortgage banking, measured by outstanding loan amounts, is by far the largest US financial sector, accounting for about $12.5 trillion in outstanding consumer borrowings. By comparison, credit card borrowings are about $1.1 trillion. Student loans account for $1.6 trillion in consumer debt. The mortgage banking industry has a long history of using technology, including AI, to facilitate loan origination and open up home ownership opportunities for American families. Powered by AI, the underwriting engines at Fanny Mae and Freddie Mac, and at mortgage banking companies themselves, are among the earliest instances where technology was harnessed to increase consumer options and to streamline the origination process. Bob, perhaps you could share with our listeners some of the benefits that technology has brought to consumers served by your members.
Bob Broeksmit:	Sure. Thank you, Jerry. Thanks for having me on and thanks to Orrick and its predecessor firms with whom MBA has had a long and very fruitful relationship. We really appreciate that. As to your question on technology, I've been at this since 1985. So, my bio says 35+ years because I got sick of adding the year. But pretty soon I'm going to have to say 40 years. And in that time, of course, a lot has changed dramatically. Think about how one can shop for a home, shop for a mortgage, close a mortgage with RON, Remote Online Notary, and frankly, if you wanted to, do it all from your couch. It's really remarkable with the advances, and COVID of course accelerated it. Because when it was unsafe, or people were nervous about having people traipse through their house, they just did a virtual tour. And the digital quality of the photographs and all of the things that one can do from one’s armchair is really startling. But the thing that I think was perhaps the biggest change from when I entered the business to today is that I started by typing verifications of employment, verifications of mortgage, and verifications of deposit, and putting them in self-addressed stamped envelopes and waiting for them to come back in the mail. And that was all done with a three ring binder about 4 inches thick with all the guidelines from Fannie Mae and Freddie Mac. And when automated underwriting came in, the first iterations were simply ways of automating the three-ring binder. And if you met the four corners of the guidelines, that would maybe help you a little bit to be sure that you did. But it didn't really automate the process of making a loan decision. It just said here's the manual, now here’s the manual on a computer. But when they started saying, "Wait a minute. Let's go back and look at what variables are predictive of loan performance, and what combination of variables, even if one of them is outside the four corners of the guidelines, are okay," getting into the compensating factors right? And remember when we used to have 28 and 36 front-and back-end ratios? Well, nobody ever talks about a front-end ratio anymore. And 36 can be 44 if you've got more cash in the bank or a higher credit score or whatever it is. So, credit scoring and automated underwriting to me have been the biggest changes, and have helped consumers because it takes the subjectivity out of the process. You are not going to have an underwriter who has a point of view about a borrower's background, race, gender, location, occupation, any of that, when a machine is saying, "Look. I'm looking at the borrower's attributes; I don't anything his or her personal characteristics." And that's taken a lot of the bias out of the process. Some will say that credit scoring is in itself relying on data that has been biased over the years. That's a whole another– we could do a whole another segment on that. But I do think that the benefits of taking out any of the underwriter's discretion as it relates to the characteristics of the borrower have been a real advance in our industry.
Jerry:	Caroline, why don't we bring you into the conversation here?
Caroline Stapleton:	Thanks so much for being with us today, Bob. And I think it's so interesting about the evolution of technology in the mortgage industry. It's something that's been particularly of interest to Jerry and me and the other folks at Orrick who think a lot about AI because regulators have lately been talking so much about kind of the perils of baked-in, you mentioned bias in how humans interact. But a potential for baked-in bias in these algorithms. But these algorithms have been used frequently for much longer than the past couple of years that they have been getting so much attention from the regulators. So, in that way, it's not really a new phenomenon in the mortgage industry, and in some ways you could say the mortgage industry has really led the way in this area. One thing that is though, particularly new in my view is the advent and the implementation of generative AI. And I think that generative AI has really quickened the pace of AI adoption, or at least the conversations about AI adoption in financial services. I know that we at Orrick are experiencing a huge uptick in demand for advice about how companies, in all sectors really, should fashion their AI policies. Not just financial services. We're helping some of the largest tech clients that not only are developing large language models themselves, but are also trying to craft the strategy and policies that they use internally around generative AI to mitigate their legal and compliance risk. So it's not just mortgage companies, but I'd be very curious for your perspective on MBA: One missions is to serve as a forum for discussing regulatory policy, providing your members, through committees and conferences, guidance about where regulatory policy is moving. And then of course there's the lobbying arm to represent the interest of your members. So, are you at MBA picking up on some of these same questions that we have about how members should be approaching this “brave new world” of generative AI, not moving too fast, but not falling behind? Staying within the lines. Thinking about the regulatory issuances and guidance in this area. So all that to say, what role do you see for MBA in this emerging area where technology and regulation are really meeting up?
Bob:	Yes, Caroline. It's a really important and timely question. And it does go a little beyond my direct expertise. So I phoned a friend, David Coleman, who is the president of MISMO, which is Mortgage Industry Standards Maintenance Organization, which is a subsidiary of MBA, and he's right in the thick of it. He let me know that interest in AI among our members has exploded over the past year and a half, and that we started getting questions about ChatGPT shortly after OpenAI released the public version at the end of 2022. Since that time, our members have quickly gone from learning to implementation. We responded by holding MISMO- and MBA-sponsored AI forums in September of 2023, which was a really cool one because it was at Amazon's HQ2 right across from Reagan airport. A spot that was kind of fun to go and see. And reception on the roof deck and all that. I didn't get in to all of the conversations, but I sure enjoyed that view of Washington from across the river in Virginia. And we just did another one in June of 2024. And speakers, you mentioned that you work with a lot of the big tech firms. Speakers included AI providers like Amazon web service, Google, OpenAI, and industry consultants such as KPMG and PWC, and a laundry list of big lenders. As well as importantly, Fannie Mae and Freddie Mac. And there were a couple of big shifts between those two events. The first one was that last year there was a lot of curiosity about it. What is it? What can it do? Who provides it? How do I get started? But this year, there's much more interest in what is everyone doing. What are my peers doing? What works? What doesn't work? And what we are seeing is more of a tiptoe into implementing AI versus a rush. One of the GSCs has approved one implementation of generative AI to assist in software development, and other implementations have been for chatbot assistants. We've certainly seen that a lot on the mortgage servicing side as you try to get your call centers efficient but also responsive to consumer's concerns. Document scraping certainly, moving text into data. Quality control, process monitoring, fraud detection, and cyber security are some of the other areas of interest. We have not seen any implementations of decision making to date. And the human being in the loop is still prevalent. But that’s certainly in the area we expect a lot of development. So, put succinctly, it is an area of almost unimaginable potential, but also a lot of risk, both in terms of compliance and making sure that borrowers are treated better as a result of AI as opposed to disadvantaged. So that's kind of the state of play as we see it at MBA.
Caroline:	Yeah. I think it's so interesting that the decisioning piece is kind of the last one, it sounds like, for your members to tread into that water. And I think that it's probably not a coincidence that that's the place where regulators have, in my view, expressed some of the greatest concerns about the things that you can't control if you lose the human in the loop for some or all of the model’s operation. That “black box” concern about how decisions are made and how applicants would understand why they received a particular output. But it does at the same time offer so much promise to be able to look for ways to enhance performance of these models. So, it's definitely– I guess we'll all stay tuned.
Jerry:	The interaction of the human and the technology in this space is pretty crucial. And we've had discussions before on podcasts about explainability and how the agencies are going to expect that the consumer will get an understandable result that isn't just a reference to the black box. So, that's evolving. And the industry and the regulators are evolving with it. And as you noted Bob, MBA has for many years taken a lead in promoting tech advances through the Mortgage Industry Standards of Maintenance Organization, or MISMO, as it's commonly known. Now on June 25, MISMO invited industry professionals to join in the newly formed Community of Practice, or COP, focused on artificial intelligence. This initiative, of course, is in its earliest days, but could you share with our listeners what this initiative is designed to accomplish? It seems to have huge potential as a collective effort to adjust the kinds of issues related to AI strategies and policies that many of our tech clients and other clients have asked for. And as MBA members, we applaud MBA and MISMO for taking this initiative.
Bob:	Yes, Jerry. The Community of Practice, which I guess is a bit of a term of art that we have used in MISMO over the past few years to try to get people in the industry who have a similar goal, whether it's, let's say, getting servicing transfers right. What are the data requirements? Let's get people who really do this all day, every day as opposed to just some people who think big thoughts, and let's really dig in, roll up our sleeves, and solve the problem. And so that's what we're trying to do here with this AI Community of Practice, which was launched as a follow-up to the forums I mentioned before, and a series of working discussions held with MISMO membership. Coming from these discussions, it was determined that MISMO was well-positioned and indeed should provide the forum that the industry needs to begin to discuss AI: risks, guardrails, the regulatory environments, and the scope of how and where we use AI, and specifically generative AI. This forum also provides membership with an opportunity to learn, keep up with the innovation and the changes underway. And AI is one of the technologies that will transform the way we lend, the way we do business. We just don't know exactly how yet. But MISMO is uniquely positioned to assist our members and the industry in implementing the technology. As you can imagine, this AI Community of Practice has been well received by MISMO membership. And the list of proposed activities and responsibilities continues to grow. But do you know what the first deliverable that was agreed to by this new group was? A glossary. Sounds fairly basic, but they quickly recognized the need to be speaking the same language. So, there's a subgroup curating a glossary for use by the Community of Practice. And it is being used by FHFA for their AI Tech sprint in July. The Community of Practice will continue to prioritize topics, activities, and work products , and their focus will on how to keep AI safe with the appropriate guidelines and guardrails necessary to protect the consumer, the developer of the AI, and the user of the AI. So, lots going on, as you said, in its nascent stages, but I suspect that this initiative will advance much more quickly because of the business imperatives. And we talk about cost to produce all the time. All the money that our members have spent on technology, the cost to produce has been stubbornly high. We just have not seen it come down meaningfully, and I think a lot of our members see AI as a potential breakthrough in that regard.
Jerry:	Well, it is a tremendous initiative. You know, I can remember, you and I go back a way, but 20 years ago when the E-Sign Act was passed was when MISMO was set up in the first place. Because there was a new way of doing business. And it's been a contributor in so many ways over the years, but I think this may be its shining moment. It really is a tremendous opportunity for the industry. Feeling its way as everybody is, to come together and develop a glossary because, yes, we have to know- have a common language here. And we’re even at the stage where common language isn't necessarily taken for granted. We actually have to come to some understanding about what are we dealing with, and how are we going to denominate these issues.
Bob:	And I think Jerry, that we're fortunate that MISMO has a several-years track record now where the industry and all the ancillary groups trust it to bring together all players from the industry and agree on a standard and a way forward. So, we've toiled in the vineyards and now I think it's going to pay off as AI comes into play.
Caroline:	You know, one place, Bob, you mentioned AI coming in to play, we talked about chatbots, some other ways that internal uses are helping make operations more efficient. But another place where we're seeing tech really take the lead is marketing, and I think the regulators are responding to that too. HUD recently released some guidance on how the Fair Housing Act, the FHA, applies in advertising housing and credit and other related transactions through digital platforms, and these guidelines are very interesting. It appears their objective is just to ensure that companies make sure that when they're when directing marketing campaigns to particular market segments that they're still complying with the Fair Housing Act, which is a known obligation. But this guidance kind of goes into detail about HUD’s perception about the specific risks in digital marketing. And, for example, audience characterization, right? Which HUD says has the potential for certain categories in audience segmentation to be proxies for protected class membership. So I'm curious, had the MBA gotten any feedback from its members on this guidance, or what your thoughts are about AI and algorithms in marketing? It does seem to be challenging, to be sure, that HUD’s not going to read subtle messaging in the wording of ads that were not intended. There's also the challenge of actually selecting the audiences themselves.
Bob:	Yes. I think this notion of how lenders making their product offerings known to the general public as opposed to perhaps through referral sources like realtors and builders, and the Fair Housing and Fair Lending consequences. And you can look at the CRA requirements as well. And when you see, for instance, a bank lender that has a concentration where its branches are located, but then perhaps has snowbirds that go from the Midwest or Northwest down to Arizona or Florida for the winter, and then, while they don't have branches down there, they want to serve that community, perhaps in second home lending, you start to do a fair number of loans down there and the government regulators are going say well that's part of your assessment area now and you have to reach out to every borrower. And it gets really complicated that the nexus between serving your consumer wherever that consumer goes and then being obligated to serve every consumer in a market where you don't have a physical footprint. So, I did on this one phone another friend, Justin Wiseman, our crack regulatory council here at MBA. And he tells me that we actually haven't heard nearly as much about this from HUD as we have about the CFPB’s guidance on digital advertising in the RESPA Section 8 and UDAB context. This HUD guidance in some ways states more settled legal obligations, and ones that aren't necessarily surprising in light of HUD’s suit against Facebook. I do note that they relied on the HUD 2013 Disparate Impact Rule in places. Which is obviously on somewhat murky legal footing, given all the litigation that has taken place. And the Supreme Court's subsequent decision on how to interpret disparate impact liability. But your question on the wording of advertising is interesting. It seems like that risk has been present for a long time, and this guidance seems much more focused on who you are targeting with your advertisements, and how you're segmenting your audience online or on social media platforms. Regulation of the content of advertising is less new and something that our members are acutely aware of. So, clearly social media platforms that have targeted advertising capability are becoming more and more significant forums for people to get news and information. Online targeted advertising is also ubiquitous. And these technologies are here and will be in the future. We continue to engage with our membership and with regulators to help them find ways to use advertising online and on social media platforms in a manner that complies with their legal obligations. But clearly, when the very way that Americans or human beings communicate with each other is changing so quickly and so many generational differences, this is top-of-mind for our members in area where there's a lot of risk if you get it wrong.
Caroline:	Yeah. And you raised a good point about Community Reinvestment Act obligations too. And in some ways, audience targeting had the potential to expand inclusiveness. Leveraging technology to find in roads into populations and geographies that a lender might need, or want to serve for internal purposes, but also for regulatory compliance. And then at the same time those tools do present the risk that HUD's identified in the guidance about, potentially, this would – the machine being able to identify proxies that we ourselves wouldn't see. So, yeah, it's definitely something I think a lot of folks are grappling with.
Jerry:	Well, one last question. The mortgage industry has many regulators as you know. The state banking agencies license the majority of mortgage bankers. Federal banking agencies charter banks that make mortgages and of course an FHA Eagle is essential to any mortgage banker and then being approved as a seller/servicer by Freddie Mae and Freddie Mac is a credential needed by many in the industry. And then of course there is the CFPB, that administers RESPA, Truth in Lending, ECOA and a whole range of consumer protection statutes that have accumulated over the years. Recently, the CFPB launched what it characterized as an inquiry into how junk fees are pushing up the cost of buying a home. When you spoke to the Exchequer Club in Washington recently, you commented on the issue that the CFPB is seeking to address with this initiative. Maybe you'd be willing, on a high level, to share your thoughts with our listeners.
Bob:	So, I don't need to phone a friend for this one. I know this one intimately and it has really animated me because I'm really disappointed with how the Bureau is characterizing this and let me explain why. The CFPB went through a very long process to develop TRID, which is one of those “only in Washington” things where it's two acronyms which form another: TILA-RESPA Integrated Disclosures. So it's acronyms within acronyms. But our members spent hundreds of millions of dollars and a lot of time implementing this massive change to how fees are disclosed to mortgage borrowers. And we have to be honest and say that before TRID, there really were surprises at closing. You would think that you get a certified check to go to the closing for $15,000 and then at the closing table they say, "Oh yeah, well a few things changed, you owe me another twelve-hundred bucks." And, you know, nobody liked that, I get that. There was a need to do something. But there are no surprises at closing anymore. TRID has done a really good job of ensuring that and lenders spend all sorts of compliance dollars making sure that the closing disclosure matches the loan estimate and any variations the lender has to eat. This is a very strict rule. And the CFPB promulgated it and enforces it. So, for the CFPB then to say that there's all kinds of junk fees in the mortgage business when those fees are disclosed way before closing on a form promulgated from that very bureau and they did a look-back on their rule and patted themselves on the back and said, "this is working great." So how in the world can you have the director of the CFPB then come out and say that things that are absolutely essential to prudent underwriting and sustainable home ownership like credit report fees, appraisal fees, flood certification fees and title insurance are junk fees that don't benefit the borrower? It's just baffling and clearly has a very political overtone. It was no coincidence that the Bureau's screed on junk fees came out the day after the State of the Union Address where President Biden mentioned junk fees because he's in a tough reelection fight and fighting against junk fees polls well and people at every campaign stop are jaw-boning at him about the lack of affordability in housing. So, there's a natural political reason to do something but when it goes from a vague notion that closing costs are too high to an absurd blog post from the assistant director of the Bureau saying they are “shocked that the use of discount fees has grown dramatically from 2021 to 2022.” I mean come on, what else grew from 2021 to 2022? How about interest rates by 300 basis points? So if you can get a 3% loan with no points in 2021, why in the world would you pay a point and buy it down to 2¾? But in 2022, if rates are 6, yeah, you might pay a point and get it down to 5¾. That's natural behavior and the benefit to the consumer is a choice to say, "I'll pay less in exchange for paying a little bit in cash up front." This is the way the mortgage industry has worked from time immemorial and for the Bureau to put out something like that emboldened us to respond by saying, “It's a little frustrating when the very agency that is meant to regulate your industry is ignorant of the workings of said industry.” It was really infuriating and now the Bureau is following up with a request for information on fees. And there's one area I'd like your listeners to pay particular attention to, because when you get through all the BS and the politics, one thing that the Bureau does seem to be genuinely interested in — and if President Biden is reelected, I'm quite sure they would pursue some version of this — is that their view, not incorrectly, is that the average consumer gets a mortgage every five or seven years or whatever the number is. The average lender does hundreds or thousands of loans every month. So who might be in a better position to negotiate on, let's say, a bundle of closing costs like title charges? Well, the Bureau thinks the lender might be able to do that and offer a bundled rate to the consumer. So that's one of the questions in this RFI. I think that the industry would do well to pay particular attention to that one and remind the Bureau that such a bundling process has been tried before. We believe it doesn't work. We believe it would advantage the largest lenders at the expense of small- and mid-size lenders. But we're still developing our responses to that but those comments are due in August. And in private conversations with Director Chopra, he has certainly acknowledged that things like fretting about discount point usage are silly. He took all of my arguments for why that was a dumb thing to focus on, but I do think that the issue of who's in a better position to negotiate for better costs is one that will persist.
Jerry:	Bob, that's interesting and, you know, you weren't even around but I was there when RESPA was passed and Senator Proxmire was an advocate for bundling and having the lender work the whole thing out and just give the consumer one number.
Bob:	He of the “Golden Fleece Awards”, right?
Jerry:	He of the Golden Fleece Award. He the author of so many of the statutes that we work with today. A remarkable man, actually, and an interesting Senator who has influenced consumer finance in a bigger way than almost any person in history. But that was his thought at the time. He advanced the idea. So it's back to the future.
Bob:	There's nothing new under the sun, right?
Jerry:	Right. Bob, thank you so much for joining us. It's really been great. You know, we've known each other for a long time and I've been an admirer of yours and the things you've accomplished and what you're doing at the MBA, at not the easiest time for the industry, is really important. So , thank you very much for joining us.
Bob:	Well, it's been my pleasure. Thanks again for everything Orrick does for the industry and really good to be with you both.
Caroline:	Thanks so much, Bob.