top of page

Fair Lending With Mike Taliefero Of Compliance Tech



Mike Taliefero co-founded ComplianceTech and together with his business partner was able to develop software that is capable of combining HMDA data with loan origination data. Their suite includes Lending Patterns, CRA Check, Fair Lending Magic, Fair Servicing, and more. To learn more about ComplianceTech and fair lending, join Mark Treichel as he talks to the co-founder of ComplianceTech Mike Taliefero. Discover how and why ComplianceTech started. Learn how a fair lending compliance program should look in terms of staffing, organization, and scope of review. It's all about due diligence and Mike believes in that. Learn more about what ComplianceTech offers to assist credit unions with Fair Lending.

---

Listen to the podcast here


Fair Lending With Mike Taliefero Of Compliance Tech

I'm excited because I have one of the Co-Owners of ComplianceTech, Mike Taliefero here today. How are you doing?

I'm doing great, and it's a pleasure to be with you on the show.

I'm excited to have you on the show. I have had some conversations with some members of your staff who demonstrated the software that you have. You and I had a conversation after that relative to the things that ComplianceTech does for credit unions and other financial institutions. If you could give us a little bit of background on how you got into forming this company and the journey that you've had that got you to this point, that would be great.

It's an interesting journey. When I graduated from law school, I never would have thought that I would end up doing what I was doing because I went to law school in the days when personal computers had not come about. I started my professional career in Washington, DC like many folks who are in compliance and regulations, working as an attorney for the US Department of Housing and Urban Development. I was in the finance division of the Office of the General Council. The finance division primarily represented Ginnie Mae and also did Fannie Mae oversight. That was my responsibility.

I also got involved in bond financing or real estate tax syndications back in the heyday. I moved up the ladder to Section Chief for Ginnie Mae after six years or so. After HUD, I left to go to the Mortgage Bankers Association, and my first job there was as the Director for Financial Affiliates. That was to represent the mortgage banking industry's interests for those companies that were depositories.

It included interacting with all of the bank regulatory agencies, including NCUA, on issues important to the mortgage banking industry. I moved up to the ladder there. I was a Vice President when I left and had also taken over secondary market activities that included interactions with Fannie Mae and Freddie Mac. While I was at the MBA, I met a gentleman in Chicago named Maurice Jourdain-Earl, who is my partner in this business.

I met him. I was being introduced as the new Director of secondary market activities. This is one of those in-between promotions. Maurice is in the room. There were about 1,000 people in the room but there were not very many minority people in the room, and Maurice came up to me afterward and struck up a conversation. Maurice is African-American, and we both happened to be from the Chicago area. He's from Chicago. I'm from the suburbs of Chicago. We started talking. He worked for Citibank then. Citibank had a downturn a few years later, and he came to me. He wanted to start this business to do due diligence. This was during the thrift crisis, and he thought there was a business opportunity.

To make a long story short, because it is a long story, we started this company to do due diligence. We organized assets for securitization in the secondary market. We did underwriting reviews of all types of loans, business loans, and mortgage loans. There were more mortgage loans than anything else. Any type of asset we help to securitize by doing the re-underwriting.

That business started to die out as things started to come to an end with the thrift crisis. When that happened, we had to think of a second act. At that time, we had a California office. It turns out that one of our employees in California was chit-chatting with one of her mortgage banking friends

and he said, “I have this HMDA data over here.” He had it in paper format. He says, “I don't know what to do with this. We filed it but didn’t know what to do with it.”

In our due diligence underwriting background, we took that information and said, “Let's try to make sense of this data.” We married the HMDA data with his loan origination system data, and a new thing occurred that we call Fair Lending Analysis. We call this the HMDA monitoring and intervention report. That's what we called the typical government-sounding name. That began things then we also started to do fair lending file reviews for lenders, and then we saw that was a very inefficient process. We started to have the idea, “Why don't we use technology to make this process better?”

That gave rise to our first desktop Fair Lending Analysis tool, which was called Shamus, which now is Fair Lending Magic. We also had another tool desktop. We called it HMDA-Ware. It sounds too much like something else, but now, that has become LendingPatterns. We have added to that suite of products with an HMDA and a CRA reporting tool. All along the way, we have done consulting. We consult in all areas for all types of lenders, credit unions, banks, non-bank mortgages, and other types of lenders. It might be an auto or small business. We covered the gamut.

That's quite a journey. We are the journey that we take. One of my favorite books is called Blink, and it talks about how you can make a decision in a blink of an eye. It goes to explain that it's not a blink of an eye. It is the journey that you took and all the decisions that you made and the things that you’ve seen that leads that computer in your head to go, “I can make that decision quick.”

Your journey builds a strong foundation, all the different things that you have been involved with there. That person saying that they had the HMDA data and going over to using something with that data on the front end, could you describe that again, the discussions that you had relative to that? As a link to that, when was it that you first went to the personal computer step? How many years ago was that?

That was back in 1995. That was the period there. We took that HMDA data. He gave it to us with no direction and said, “Can you guys tell me what this means to me?” With our experience in the industry and working with all types of lenders, we said, “In order to tell this story, let's make it more complete.” Let's append to the HMDA data loan origination system data.

That marriage allowed us to tell a deeper story. A story by channel. A story by a loan officer, underwriter or by a specific product. In those days, HMDA data didn't tell you very much. Since 2018 we have gotten to the point now where we have a data set that is almost as good as what we did when we put those two data sets together, HMDA and LOS, Loan Origination System.

If someone is reading, we have a reader who's new to their credit union, they are looking and understanding what good goals would be for a fair lending compliance program, and were chatting with you, what would you say to them?

A good and great goal is number one, to be able to detect discrimination if it is occurring. Detection involves research. It involves investigating on your own. It evolves by establishing or using key performance indicators that indicate inefficiencies. It involves tracking. It involves looking at inconsistencies and exceptions. That first goal would be detection.

The second goal would be developing controls, and these controls would be to prevent discrimination. When I think of controls, I think of guardrails. You can impose parameters on discretion where you know there is discretion, and that's part of that control process. You could also impose restrictions on exceptions. You can control the magnitude of the exception as well. Those are a few on controls.

The other thing that would be a good goal would also be able to implement corrective action if you did find something. I used to always tell people I still do that part of the job of your Fair Lending committee is to decide what you are going to do if you find something. It's better to decide what type of correct action you will take, in general, before you find a problem. The reason why is it's so hard to make that decision on the fly. It's better to make it when it's not an issue. If we find X, Y, and Z, we are going to do A, B, and C.


It's better to decide what kind of correct action you will take before you even find the problem.

If a scenario like A, B, and C shows up, we are going to give our loan officers or our loan supervisors more discretion above and beyond the policy or vice versa. Am I on the right track and trying to think of an example there if that's not a good example or if you have a better one?

That could be an example but it wouldn't be more discretion. It's always will be less or we are going to monitor the discretion. We are going to track it and give those people who are using it feedback. That would be more of the type. Usually, when we talk about corrective action, we are talking about making things well and good with the people that may have been harmed. That's what we are talking about usually. If they have been overcharged, there might be an opportunity to refinance at a lower rate or some other rebate. Let me say that should be part of the plan but that happens rarely. I don't want to give the impression that that's something that you could experience.

This is a good example. It's not frequent but it's a good example.

I would like to give you one more thing on goals, which is to have a goal to improve the process on an ongoing basis. All of the things with it, detection, controls, corrective action, looking at the recommendations, and then recycling that in the next time around, improving the process where you can.


Keep improving the process on an ongoing basis.

That can be through the building into your policy and process that annually you are going to look at the policies and procedures, take a look at what it is you learned, and then build that in moving forward. This same person that we are talking about in a credit union, if they are looking at their Fair Lending Compliance program, could you give me some context as far as what staffing levels? Maybe that varies by size and volume. What type of organization and, in the perfect world, what their scope of review would be?

It's going to vary greatly by the size of the institution. I would start with asset size and also add the issue of the types of loans that the credit union makes. If credit union makes mortgage loans, consumer loans, auto loans, small business loans or home improvement loans, the more different types they have, the greater the complexity.

The other thing at play is where they lend. The more different places, the credit union lends that adds complexity. It's going to add complexity if we talk about redlining, that issue could surface there but that also means you've distributed your resources as well. The branches and maybe third-party broker activities. It gets channeled. There are more things to look at.

We have talked about loans. The number of loan types. We have talked about the number of how you are spread out geographically. The other thing is the volume. The volume will drive whether or not you need to employ statistical analysis. If you are small or low asset size, you do about 1,000 loans or less than 1,000 loans, mortgage loans, and less than 1,000 credit union loans. You are probably not going to need to do statistical analysis in a rigorous way.

WFC 43 | Fair Lending
Fair Lending: The volume will drive whether or not you need to employ statistical analysis. If you have a low asset size or do about less than a thousand loans, you're not going to need to do this.

I said all of that to get back to your question about staffing. If you are less than the thousands, I would say you probably can get by with one person, and that person might not necessarily be doing fair lending full-time, depending on how many different things you have. You can get by with one person. You are very small. Now, I'm going to go to the other extreme.

If the other extreme is a credit union that's over $10 billion, first of all, you are probably going to be doing all of the lendings that I'm talking about. You are going to be doing it all over the country. You are going to have a lot of branch offices. You are going to have a lot of people making decisions. You may have decentralized underwriting to some degree, and that's an important factor as well. You are going to need a lot more people. You are going to be more or less like the banks. You are going to be more or less like the big banks, and you may have 30 people in your department. You may have that many.

It can run the extreme. You can maybe figure out from the lowest to the highest where that credit union might fit in terms of personnel midsize. What is more typical that we see is a compliance officer, and it's always good to have one, a fair lending officer and a fair lending analyst. You can go a long way with just those three people. That could take you up to be a $3 billion organization. You can get a lot done with that. I hope that's helpful.


If you have a compliance officer, a fair lending officer, and a fair lending analyst. You can go a long way with just those people.

That's very helpful. the bigger you are, the more markets you are in the more people you have needing the decisions, the more loans you make, and the more likely you need to do some statistically valid sampling. As there are so many variables there, you could have an area where a branch has misinterpreted policy, and that's leading to decisions different than what the board's expectation is or you could have people who are putting their thumb on the scale or whatever. It’s because that, that's why you want to have a bigger department and do more testing and sampling. Am I on track there?

Yeah. You are on track. I will add to that another scenario is with acquisitions. You may have acquired another institution. That's merging systems and culture, and that can be difficult as well for fair lending.

I have seen that and heard stories about organizations and credit unions that have merged and had to merge two cultures. It takes a lot of effort from the top to have the old group start looking through the lens, policies, and procedures of the new group. That makes a lot of sense.

I will add it as an aside too. One of the things that you can do when you are faced with that is you can merge the HMDA data, and our LendingPatterns tools allows you to do that in seconds and then analyze and see what you look like together historically. You can see whether or not it supports fair lending or whether you got a tougher row to hoe.

You mentioned the word redlining, and if we get to that. Let's speak a little bit about the concept of redlining, either what it is you've seen out there, what it is that credit unions can do to make sure it's not happening or any relevant thoughts. When I say the word red line or redlining, what pops into your head relative to what people can do in that regard and how a system like what you offer can assist in that regard?

I can answer that question. This is one area where for credit unions, while it's extremely important, I wouldn't put that necessarily at the top of the list of the things to do for the credit union. I would like to answer that question because I don't want to leave the audience with the idea that it's more important than some other basic things that I would like to tell.

Answer that question, and then let's pivot into what the basic things are that are more important to understand.

Redlining is discrimination by race or ethnicity. It could be national origin as well on a prohibited basis based on geography where the people live. The way the Justice Department has interpreted the Equal Credit Opportunity Act and the Fair Housing Act in their complaints they have shown it to be a facts and circumstances situation.

WFC 43 | Fair Lending
Fair Lending: Redlining is discrimination by race, ethnicity, or national origin on a prohibited basis based on geography.

They look at whether or not there is a difference, among other things, in the market share in minority communities versus majority communities. That's one analysis. There are other analyses as well, and they look to see whether or not there is a statistically significant difference in that market share. That's one framework. Whether or not there's a statistically significant difference in the concentration with one group versus another in those geographies.

With that analysis, and by the way, this is an analysis that can be done with HMDA data and for mortgage loans. You don't need to look at loan qualifications because it could be done with applications. It could be done with originations too but you don't have to look at the credit score of the applicant or any of those things that don't come into play. In a sense, it's easier to determine.

However, I can't speak for the Justice Department but it appears from their allegations in the various cases that have been brought that they like to have a little more than that. They want to see something as well that shows either intent, although it's not required or negligence in the activities of the institutions. They will look at the opening and closing of branches and whether the opening and closing of branches have been done in a disparate manner.

They will look at also complaints in the community. They will look at maybe complaints by individuals or former employees. Those things color the motivation even though the motive is not required but it colors it. That is combined with the statistics. Those things without the statistics are not strong, and the statistics can somewhat stand on their own but from a regulator's perspective, you want to have some of that other stuff to show that there's a serious mindset problem here that needs to change. That's on the redlining.

What I would think is a bigger issue for credit unions is that they have come to the game later than banks and mortgage companies from a scrutiny point of view. I read the day that the fair lending examination process only started several years ago for the credit unions, and it has been around longer for banks. What I find in my experience with them on our consulting is that there's more to do with respect to getting started, finding the right personnel, and designing a program.

I will start with designing a program. The design of a program begins with what I would call Fair Lending Risk Assessment. That's where the credit union looks inward and asks itself a lot of questions about its operation. Those questions come from the inter-agency exam procedures. Some people call it FlexPro. They have different names for it.

WFC 43 | Fair Lending
Fair Lending: The design of a fair lending program begins with a fair lending risk assessment. That's where the credit union looks inward and asks itself a lot of questions about its operation.

The inter-agency exam procedures are where each banking agency, including NCUA. They have agreed on a set of parameters for compliance risk, redlining, underwriting, pricing, and marketing risk factors as they relate to fair lending. One can take those risk factors and turn them into questions. We have done that in our software, as well as adding risk factors from our experience as consultants and then asking the questions. Do we have loan types that have features that could have disparate outcomes?

Ask yourself the question, “Do we have subprime lending? Do we have disparities in underwriting outcomes and so forth?” You go through that process and rate yourself low, medium or high on these things. As I said, we do that in our software so that you get an assessment and sum all of that up. Another thing you might look at is whether you are dealing with third parties. Dealing with dealers, brokers or third-party originators would be a higher risk.

Using an automated underwriting system in itself is at higher risk. I would say that in itself might be a lower risk but only if that system is generating fair outcomes. You will have to test the AUS for bias. The extent to which you are doing judgmental decisions, underwriting, and as well as pricing, whether you allow pricing exceptions, under what circumstances you allow them, and the tracking of those exceptions.

The other thing is, do you have policies, plans, and procedures that could have a disparate impact? That's always another one of those provocative topics of disparate impact. That needs to be analyzed statistically usually. Lastly, mortgage, automobile or consumer loan servicing, particularly where the collateral is involved and whether you are going to have foreclosures and handling of REO. There have been some lawsuits brought on the mortgage side for that. Credit unions have a significant risk they do such tremendous automobile lending is their repossessions, whether or not that could be happening on a prohibited basis, and complaints that might surface. Those are a few of the things.

There's a lot there. One of the words we both said quite a bit here is statistical methods and statistical sampling. Can you think of some scenarios or maybe give some examples of whether or not when to use statistics and why, and then what statistics are good to utilize?

Let me give you a thought process for that. Random pricing, underwriting, and marketing inconsistencies are okay. However, patterns of prohibited basis disparities in pricing, underwriting, and marketing are not. They are especially not when the applicants are similarly situated. I said a few things. I talked about three areas, pricing, underwriting, and marketing. Those are the three areas you want to focus on. When we talk about marketing, that's that redlining thing.

WFC 43 | Fair Lending
Fair Lending: Random pricing, underwriting, and marketing inconsistencies are okay. But, patterns of prohibited basis disparities in pricing, underwriting, and marketing are not.

I mentioned prohibited basis disparities. I am only talking about those things that prohibited basis disparities that you can get from either HMDA data. I don't know if we are going to have time to get to this or proxies that for what would otherwise be in HMDA data if there was an HMDA data collection for it. You can use surnames and demographics to estimate the race, ethnicity or gender of a person.

As I said, randomness is okay. When it's not random, that's saying there is a pattern. Statistical analysis tells us that the pattern of a disparity is real. The economists say not different from zero but that's not how normal people talk but it's a real difference. It's not a random difference. It's unlikely to happen by chance when it's statistically significant. When you are doing these analyses, if it's pricing, you are measuring based on APR, and that would show your output in terms of basis points. If you are doing approval vs denial, it would be based on odds ratios and the probability of denial.

When those differences show a race effect in the models when you controlled for qualifications, then that's when we have a problem. One more thing I like to say in consulting, we like to do an analysis without controls first and then with controls. Without controls, we call that bivariate analysis. We would be looking at whether or not, let's say, non-Hispanic White and Hispanics. The difference in APR might be 25 basis points, and if that's statistically significant. We do a T-test that would tell us a very simple thing. You can even do it in Excel or you can do it in software or our software. It would tell you whether or not there's a statistically significant difference.

That gives you a focal point. That tells you, you ought to investigate further but it doesn't tell you why because there may be a justifiable reason for that difference. The difference might be because one group on average at a higher LTV or one group on average had a lower credit score, one group on average had a higher DTI or all of the above.

All of those things are taken together to do the statistical analysis, that is explained why we have to use regression analysis, and so you have to create a regression model. This particular model that we are talking about for pricing or APR would be an Ordinary Lease Squares model, often referred to as OLS. We would run that model using APR as a dependent variable, and then we would work with the institution to find the things that they say according to their policy, explain pricing outcomes, and hopefully, we have data on those and put that in the model.

We then see whether or not the model explains the disparity or whether we also put the prohibited basis in the model, whether the disparity is explained by the race, ethnicity or age effect. That's in a nutshell. We do a different model when we are talking about underwriting. It's usually a logistic regression model because we have a dichotomous dependent variable but the idea and the concept are the same.

This is a fascinating topic. We could sit here and talk for hours and go down all of these different paths. We might need to plan on doing the second episode in on some of these topics. There's one last question I want to ask before I let you see if there's anything else you want to talk about. If someone is a $750 million credit union, or a $3 billion credit union, if someone is looking at wanting to comply and have good policies and procedures and serve their members well. What's the thought process that they should be going through and deciding whether or not they have the ability to do these things in-house or that they should seek professional help outside of their walls?

That's a very good question, and I run into it all the time in our consulting practice because this is a unique area. Like I said before, I never thought I would be blending law and statistics and scientific evidence. I never thought I would be doing that, and then technology. I don't want to be self-serving here but software helps dumb down the process. That's probably not a nice way to say it but it does make the process easier because software can give you the answers, and it can guide your approach. It tells you what to do. You want to consider maybe a tool but you can do the same thing with talent but the problem is talent is hard to find and keep interested.


Talent can do the same thing software can do but they're harder to find and keep interested.

You can find somebody at any university, particularly in the Social Science realm, that can do this stuff. If this is all they did, you would have a hard time retaining them because it would be interesting for a couple of years but maybe not forever. The ideal candidate is to find someone that has the background that you could groom that's computer literate, and they could use the tools that are out there. That's probably the lowest-cost way to do it as well.

That would be the advise that I would give. I would say assess your abilities honestly. If you don't have that person internally, one of the things you could do is use a consultant. We are not the only ones that are out there but use a consultant to model the whole process for you. That's what we do frequently. People have the software but they want to see how we use it so that they can design their fair lending program around it and know what reports to run, and then replicate what we would be delivering. Those are some shortcuts in getting to the place you want to get to the fastest.

As you were explaining the utilizing of software, how challenging it could be, and how intense this process could be, a Confucius quote popped into my head that I'm going to state here before we wrap up. Confucius said, “It's a simple task to make things complex and a complex task to make them simple.” Seeing your software where it's taking the complex task of all this data and all the data that's out there. HMDA for other institutions and looking at a particular city and seeing what how your landscape looks and looking at that and taking that data.

All the complexity that went into building all of the tools that get it to that and shows it in a simple way that I could understand as a non-fair lending expert. I was fascinated by it, which is what led me to reach out to chat with you. Before we wrap up, is there anything else that you would like to say other than how somebody gets in touch with you? Any last thoughts here before we wrap up?

One last thought is that a lot of the consulting that we get is like emergency calls. They are like, the exam is coming, and it's 2 or 3 months away, and we haven't done what we said we were going to do after the last exam or we haven't done anything. The person who was doing it left or retired. When you are in those emergency-type situations, we have organized our practice around being able to help people in that situation. We are certainly willing to even talk to people without a free consultation.

If people find themselves in a difficult situation to see whether or not there's a good fit. We are not a good fit for everybody but we are certainly willing to talk and have a conversation. We are not stingy about giving information and being helpful. You can reach me at my email, MTaliefero@ComplianceTech.com. Our website is ComplianceTech.com. I will even give my telephone number, (703) 801-1285. Thank you for having me.

I enjoyed this. My readers are going to as well. That summary at the end is perfect. Thank you for your time.


Important Links

コメント


bottom of page