top of page

Mike Taliefero On The Intersection Of Indirect Auto Lending & Fair Lending

In the journey towards achieving fair lending in auto loans, we bring together data-driven compliance and a culture of equality. This powerful combination guides the industry towards a future that is brighter and more inclusive for both lenders and borrowers. In this episode, we have Mike Taliefero, Co-Owner of Compliance Tech, to discuss auto lending and fair lending. Mike kicks off by discussing the significance of fair lending analysis in the context of auto loans, especially within the indirect lending market. He sheds light on the key factors that can give rise to potential fair lending issues in auto lending. He also examines how credit unions usually evaluate fair lending compliance during the car loan process and the important roles that data analysis and technology play in this evaluation. Mike shares real-life examples, showcasing the types of data that are analyzed to ensure compliance and foster transparency. Mike shares the best practices for lenders to comply with fair lending regulations while remaining competitive with loan options for consumers. The road to fair lending is not without challenges, but with the right data and strategies, companies have a high chance of thriving. Tune in and learn how to create a future where fairness is a fundamental principle of the lending industry, benefiting both lenders and consumers.

everybody at the table

Listen to the podcast here

Mike Taliefero On The Intersection Of Indirect Auto Lending & Fair Lending

I am excited to have Mike Taliefero, the Co-owner of ComplianceTech. Mike, how are you doing?

I'm doing fine, Mark. How about you?

I'm doing great. It's sunny where I'm at and I got a nice walk in. I'm looking forward to chatting with you for the second time. You were a guest last summer and I'm glad you were available so we could chat about fair lending analysis of auto lending. Also, as a subcategory of that, indirect lending. I'm excited to let you share your wisdom on this topic with my audience. If somebody isn't aware of what your company is and if you want to give a little bit of a background on ComplianceTech, feel free as we start off here, and a little bit of your background in ComplianceTech also.

It’s a pleasure to be here. The company is about a little over 30 years old. We started out as a due diligence company getting loans ready for sale into the secondary market. That's mortgage loans, small business loans, apartment building loans, and a variety of different types of loans. That business tailed off and we had to develop a second act.

That second act basically said, “What if we took this underwriting capability that we have?” We overlaid it with the HMDA data. Would that be of interest to lenders? That gave birth to our company. I can say we were the first company to develop fair lending software. We started there with HMDA software, then we started analyzing auto loans way back probably twenty years ago now. We were involved as experts in the first series of auto lending cases that developed.

Some of which got one of the studies that one of our economists did that got reported on ABC 2020. We have deep knowledge and history in this type of lending. We developed software to report to the government in the case of HMDA and in the case of CRA, then we developed tools to analyze those data. We provide consulting to help lenders navigate through the compliance rules with respect to fair lending.

Your pivot to a second act seems to have been amazingly successful. I pivoted. I'm doing a podcast and I'm consulting. It's nice being able to take what you learned in one way and apply it in another way. You guys are doing an amazing job out there. It's good for credit unions to have you there. Let's jump into some questions I've got here. Mike, can you provide an overview of fair lending analysis and its significance in the context of auto lending, particularly drilling down into the indirect lending market?

I'll start with the mortgage. The mortgage analysis has set the parameters and the structure for what fair lending means. Generally, that means that you are required as a lender to develop a framework and an organization that will monitor the lending activities and determine whether or not there are disparities in that lending activity by prohibited basis groups.

We started that in lending. Lending has set the rules for redlining, disparate treatment, and disparate impact. All of those things have been enforced mostly in mortgage lending. More recently and somewhat historically, they have been ported over to the auto lending market. It's the same rules and equally qualified applicants that were treated differently by a prohibited basis group. It's the same framework but applies to a different kind of lending. It'll be that way as well for small business loans. I know we might talk about that toward the end of this conversation.

That's the parameters. I like how you compared real estate creating the parameters for all of that, and that it's now appropriately being looked at as it relates to auto lending and indirect lending. What would the key factors be that contribute to the potential fair lending issues that we're seeing? Also, why are we seeing indirect lending particularly under scrutiny now?

Why do we call indirect lending indirect? We call indirect lending indirect because it's not direct, for one thing, but more clearly, because the third party is involved. The third party is a dealer or perhaps another intermediary. The analog for indirect lending on the mortgage side is wholesale. It's like retail wholesale distinction. They just don't use the term wholesale in auto lending. They say direct and indirect. The third parties are involved.

Because third parties are involved, there is a lack of control over the third party. You don't know what's going on with the third party. Nevertheless, the way the rules have evolved, you are responsible for the acts of the third party. There will probably maybe be some litigation in the future on that but that is the general rule that you're responsible for the acts of the third party.

You've engaged them. You've hadn't hired them. They are a substitute for your direct interactions with the customer. They are effective agents for you maybe not in a strict legal sense, but they are working for you. There is compensation. That lack of control is where it all starts. You're responsible at least to monitor that third party's business and what they are doing, communicating with the third party as to what your requirements are as far as compliance is concerned, and checking to see whether or not they are complying. That's it in a nutshell.

I'm thinking of when I was an examiner and some of the things. I’m anything but an indirect lending expert, but I know that NCUA will want the credit unions to have their loan policies and procedures. It’s not necessarily procedures but the policies and what the rates are, what the length of employment requirements may be, and what the debt to income requirements should be driven by what the credit union wants, not necessarily the indirect lenders. If a credit union was tuning in, maybe they'd say it's not lack of control but less control. There are some things that there might be a lack of control because the credit union has no idea who's sitting in front of them. Other than it’s just pure statistical. Any thoughts on that?

I would agree that there is some control, but it's limited also in a competitive sense. Because it's a sales-oriented business and you want the loan, there's always that temptation not to be too strict because the business may go down the street. You have that tension as well.

The auto dealer is looking at, “Which one can I make the most off of as well?” It’s not necessarily what's in the best interest of my member. From the credit union side, it’s not necessarily what's in the best interest of the buyer.

Also, there's less control, but in some cases, there is no control. For example, in how the finance person or the dealer is being compensated by their own dealership, and what incentives they have to behave in a way that increases fair lending risk.

That puts a good point on it. How do credit unions typically assess fair lending compliance in an auto lending process?

Typically, the best way is by looking at data in a similar way that you would do with the HMDA data. Unfortunately or fortunately, depending on how you view HMDA, there is no Federal data collection standard dataset. The mere fact that there isn't is what makes it more complex because depending on what platform you use and what version of the platform you have, you may have different data than the lender down the street.

Also, you may not have the discipline to manage the data in a way that makes sure you create the right data for the analysis. Let me give you an example of that. While we have it in mortgage lending, it's more prominent in auto lending. You have this counteroffer interaction. Being able to capture that is very important in figuring out at the end of the day whether or not there's been disparate treatment or even disparate impact because you want to capture that activity.

Many of the platforms that I have seen have counteroffer captured but they don't have it captured in a way that indicates that there was counteroffer activity on this application. In other words, they may have a counteroffer as one of the actions versus origination versus denial. They would all be mutually exclusive. In reality, they need to capture whether or not there was ever any counteroffer activity in order to properly model it.

If I'm thinking about that right, it's because if I didn't get the deal I was asked for and I was giving a counteroffer, and I was given that offer for the wrong reason because of the color of my skin or because I'm a woman instead of a man and that's not captured, you don't have the data that you'd need to make sure that was being done fairly.

I would try to say it more straightforwardly that you want to see whether or not there's also different treatment in how counter-offers conclude. If that varies by a prohibited basis, it may be revealed that there are harder negotiations with one group versus another. There’s a very important work that was done probably fifteen years ago. It was reported in One Law. I'll look it up and give it to you after this session.

This research indicated that a lot of the discrimination in auto lending results from where the auto dealer, the finance person, or the salesperson begins the discussion or the negotiation. In this study, they sent testers out to assess where was the starting point in the negotiations.

What they found was that with the Black and Hispanics, the dealer started at a higher price with them or a higher rate because they were making assumptions about the information they had. It's that asymmetric information. That resulted in disparate outcomes. That is an important study and that's one of the things you want to get a handle on. That's why it's important to look at those counter offers.

If you send me that study, I can put a link when this episode goes live. That'd be good follow-up information for the audience to have here. Fair lending regulations can be complex. I know you know them like the back of your hand. How can lenders ensure that they are in compliance with these regulations while also providing competitive loan offerings to their consumers and to their members or their potential members?

In the auto arena, it’s hard. The biggest hurdle is the data issues. Let me back up a second. You got this origination platform. The first thing you have to do is extract from that platform the data that are relevant to decisions. There are at least two sets of decisions at the origination stage that we have to look at.

There are the underwriting decisions and the pricing decisions. There is a third set, which is decisions after close in the servicing or repossession or things of that nature or collecting activities. That can happen too, but I want to concentrate on the underwriting and pricing primarily. Those are the biggies. If you talk about emerging issues, the servicing part of it is more of the emerging issues that CFPB is getting its hands around and delving into.

Getting the right data and defining the right data to analyze is critical. That takes knowledge about, first of all, your underwriting guidelines. Your underwriting guidelines should map to your data. When you develop a regression model to determine whether or not, there has been evidence of a statistically significant difference by prohibited basis group. When you develop that model, you'll see whether or not the outcomes have been fair or not.

WFC 127 | Indirect Auto Lending
Indirect Auto Lending: First of all, your underwriting guidelines should map to your data.

Getting the data ready is critical and there are some hurdles to that. There are also hurdles in auto lending. Lenders often may change and tweak their underwriting guidelines more frequently than in mortgage lending because mortgage lending is driven primarily by secondary market agencies. You have to be able to control for your tweaks in your guidelines. You don't want to compare your underwriting decisions for the first half of this year when you changed the underwriting policies on January 1 to outcomes in the previous year when they were different. All of that is important and capturing that.

The other thing I would say is unlike mortgages, where if you are a relatively small mortgage lender, the modeling aspect of it is not going to be as important. You may not even have to do it. Comparative file analysis may suffice with a small mortgage lender. When it comes to auto lending, you don't have to be a large lender to have a large enough sample of loans to benefit from regression analysis. There would be more regression analysis to happen in auto lending as compared to mortgage lending.

That example makes sense because you can easily change your policies because you're not going off the standard secondary market. You're changing variables. It's like if I started eating better and I started exercising, which one of those two variables is leading to my weight loss? If you're changing variables, it's hard to say what's causing what.

You can't get to the underlying causal influence.

Along the signs of regression analysis, are there other proactive steps that lenders can take to address potential fair lending disparities in their policies and procedures, and in their institutions?

It's important to have the right resources together. It's important to have everybody at the table. One of the things I've seen and I've seen it in all kinds of companies but I tend to see it a little more in credit unions that perhaps the compliance department is not empowered to get to gain resources. I mentioned that data issue.

It's important to have the right resources together. It's important to have everybody at the table.

It's not necessarily a simple thing for someone in the compliance or fair lending department to get the data from the IT department unless they have a dedicated resource. It’s important to get the dedicated resources. It's important to have everybody who's involved in every aspect sitting at the table with respect to fair lending, and fair lending gets the attention that it deserves.

You need your marketing people, the people that manage the indirect lending at the table, your underwriting folks at the table, and your credit risk people at the table as well. When things come up that the fair lending perspective is discussed and analyzed, that’s how you best head off problems before they occur. It’s having that conversation and not being a fair lending operation operating in isolation without resources.

That's a fabulous point. When I think of that, I wrote down get the right people at the table, whether it's at NCUA or a credit union. If you have stove pipes and you have the individual department responsible for what they're responsible for, it's easier to make errors, whether it's fair lending or otherwise. The other thing I wrote down is enterprise risk management. This sounds like it would fit well into an enterprise risk management program.

While NCUA doesn't require enterprise risk management, the bigger a credit union is, the more appropriate it is for NCUA to have one. While they don't have one, their priorities and their policies and procedures, and all the risks, including fair lending risks that they want credit unions to deal with. This would be a perfect way to make sure that people have a seat at the table. It is inviting them to be part of the enterprise risk management development. They're at the table and looking at the risks. It's a real risk that institutions should be taking a look at. That makes a lot of sense.

One last point on that too is the most dangerous thing. The term that comes to mind is renegades. Somebody going off, doing their own thing without anyone knowing what's going on. I'm thinking more in terms of the sales side of the business and the marketing side of the business create risk. They're improving sales maybe but they're creating risk. That risk isn't being managed or monitored.

That's a great point as well. In recent years, I know there have been some notable legal cases. You mentioned the one earlier or regulatory actions related to fair lending issues in indirect lending. Any others you want to mention in what lessons can other lenders potentially learn from these cases that have come down?

There's been a couple of FTC and CFPB. Not against some credit unions though but they focused on the finance charge markup. The way that works is that the credit department looks at credit and they render a buy rate but the dealer then marks up the buy rate. The dealer comes back to the customer and says, "Great news. You're approved. I got you approved for 7.5%.” The buy rate was 5%. That additional amount, the dealer may participate in.

Those cases or those class actions that were brought 15 or 20 years ago that we were involved in as experts, that was the central focus. That's still around. It cooled off for a long time but then it has come back, not to the same degree as it was fifteen years ago, but it has come back. The other aspect that is not specifically fair lending but the CFPBB talks about it in terms of discrimination is the UDAP or Unfair, Deceptive, Actions and Practices. These involve dealer add-ons.

I bought a car in January for my wife and they try really hard to add stuff on. When a consumer is not given accurate information disclosed and they seem to have no choice, they are adversely affected, and there's not much they can do about it, which could result in a deceptive UDAP-type practice. It could be by a prohibited basis group. It may not be. It could be by income or it could be some other category. The deceptive practices are probably, if I had to mention something that's more emerging, it's been around. It's not new but there's more focus on that lately.

You can't seem to watch the news or go on YouTube if you don't watch the news or get emails. Everybody's got an email account and artificial intelligence and chatGPT. I learned of a new one. It's chatCSV. CSV is a data file, so you can convert it to an Excel file. You can put it up and say, "I want you to analyze these." I go to dig into that one because I like spreadsheets more than I like words some days.

My wife would love that.

Artificial intelligence or this machine learning can be used in good ways and in bad ways. It's determining how to use it and then refine it. How do you see this potentially impacting fair lending, whether it's related to mortgages, indirect loans, or in-house loans? What are your thoughts on artificial intelligence, where we're at with that now, and where it might help or hurt down the road?

Artificial intelligence can be used in good ways and bad ways.

That's a complicated question and my answer is probably going to be all over the place. Let me start by saying, at some point down the line, it's going to be beneficial. Where we are now is somewhat of a problematic situation. We have many of the companies that offer these decision engines. Selling them based on we're taking the human out of the scenario so it can't discriminate.


You only have to think for a couple of minutes to realize that’s not the case. The first question I would ask is, how do you know it doesn't discriminate? That becomes the primary task of the fair lending officer. It’s to determine whether or not this decision engine discriminates. It's a sophisticated process to figure that out. Usually, what you have to do is take the inputs of the decision engine. First of all, you need to review all of the inputs.

All of the variables that go into the engine to see whether or not there is anything in there that would likely cause an impact. To some degree, you can look at things judgmentally and say, “That’s a squishy requirement. I don't see how it relates to credit risk.” The proper way to do it is to look to see whether you have an adverse outcome when you look at that particular variable alone by prohibited basis group.

That's one variable at a time and looking at each variable. In the end, what you will do is after you do that and report on each variable, you would then also look at the system as a whole or the decisions as a whole, whether or not there is disparate impact. You accept that perhaps the system is a neutral decision engine but if it's giving out a disparate impact and there's no legitimate business justification for that impact or for that outcome, then that's a problem. That's considered discrimination.

That's what you have to do. You have to do that. Every time there's an update, you have to do an updated analysis. Another wrinkle associated with that is many of these systems use the word discriminate. I wish I had another word because right now I'm using discrimination in terms of slicing the risk.

Differentiate maybe.

I'll say differentiate. Differentiate the risk into clearly approved and clearly denied. There's this area in the middle that may go to underwriters to analyze, then you have automated decisioning but you also have judgmental underwriters. Now you have two things to analyze.

That's the case. The other thing too is CFPB’s pronouncements in the past year harped on the fact that your automated decisioning can't be a block box where you can't explain why. You can't tell the consumer why they were denied specifically, then you can't use essentially what they've said. That's the other issue that has to be sorted out.

It has to be almost auditable if you will. How did we get to that path?

You can't use a block box or say, “It's too complex for us to understand. All we know is it's making the right decision.”

It shouldn't. That makes a lot of sense. That’s fascinating. Maybe next year, we can do a follow-up on this on what's happened in AI from today to 2024. Maybe every year thereafter, because there will be a lot of growth in this area but with growth, comes growing pains.

I do believe that in time, we will get to a place where auto-decisioning is a perfect substitute for a human decision but that's years away.

In time, we will get to a place where auto-decisioning will be a perfect substitute for human decisions.

Excellent point. Fair lending is a balance between being a legal requirement but also, it's the right thing to do. Leaders are responsible for doing the right thing for their organization. How can lenders foster a culture of fairness and equality within their organizations as it relates to all these things that we're talking about?

It starts with how you present yourself to the public. Increasingly, now you present yourself in digital format. Your online presence is almost the first thing people see. That used to be the ad in the newspaper but now, it's online. People are going to search online before they do anything. That homepage is first impression, how it looks, whether or not it looks like it's an organization that supports diversity and inclusion.

The other thing that it should portray is clarity. They want the consumer to be informed. It shouldn't be hard to get information. This is my personal preference here. I hate to go to a web website and before I can get any information, I have to give information. The first thing pops up, I can't get rates or I can't get anything but pop up. They want to know who I am. When you're searching for cars, it is like that.

It doesn't just pop up once. You minimize it. You close it and it keeps coming back.

As a consumer, you have to have an alternative email and number that you give. I'm sure that's what everybody does.

Yes, indeed.

It’s conveying that sense of fair dealing and honesty. There's a shortage of personnel in compliance departments in general. Credit unions may even have a tad bit greater challenge in getting enough people to find out that it's not easy and they're not growing the talent rapidly. The notion is that you are recruiting people to feel some of these spots so that they can offer that perspective. They're sitting at the ta. They can offer that perspective about discrimination because they know something about it. I think that's helpful.

WFC 127 | Indirect Auto Lending
Indirect Auto Lending: Credit unions may even have a tad bit greater challenge in getting enough people to find out. It's just not easy, and they're not growing the talent rapidly.

That's great advice. Mike, if someone was wanting to talk about the digital world and that's the first place you look, are there resources and guidances available out there for lenders to stay up to date on fair lending regulations and best practices, particularly in the auto industry? Anything you can share relative to that?

I would say the CFPB's website has a ton of stuff because they do this annual report to Congress and they also issue supervisory highlights. They also cover the other agencies. They will also say how many and what type of referrals were sent to Justice by NCUA or by FDIC. They'll have great footnotes that will take you in and give you the detail on any of the cases. That's the best one-stop shop, quite frankly.

Looking ahead, how would you envision fair lending analysis evolving in the context of indirect auto lending? What impact could it have on the industry as a whole?

I alluded to it earlier. There was going to be a greater emphasis on deceptive practices or UDAP-type practices. There is going to be more attention paid to the servicing, the backend after the loans have closed the CFPB. The Federal Trade Commission came out with an opinion clarifying that accounts, credit extensions, and servicing of accounts. COA applies to all of that activity on the backend too. You're going to see more attention paid to that. In our experience, I haven't seen a lot of activity there by our customer base, but I do think that's what's ahead.

Let me add a broadening of the whole discrimination concept to more UDAP unfair and abusive practices and acts. You're going to see more of that. It's everywhere. It happened to me when I tried to cancel my American Home Shield. It took me 45 minutes to talk to three people to cancel. Finally, I canceled it. They sent me an email saying, “I didn't owe him anything.” Five minutes later, they sent me an email saying they charged me for one month for canceling.

Unbelievable. I've got a couple of subscriptions that I need to cancel and it's like, “I don't have the energy for that.” You get the email you got charged for another month.

Those practices and things like that, you're going to see more activity.

Mike, I learned something every time I chat with you. This has been a lot of fun. If there was one question I didn't ask today, what would that question be? What's the answer for our audience?

You didn't ask about small business lending and what that's going to mean.

Let's chat about that a little bit. There’s some new guidance and some new rules out there. Give me the 10,000-foot level from your perspective.

Essentially, Section 1071 of Dodd-Frank requires lenders who make small business loans will report HMDA-like information. It's a final rule now. That's in general. There are more specifics around that. The goal of it, it’s not necessarily the final reg but the statute points, is to determine whether or not there's been discrimination against minority-owned companies or woman-owned companies.

The regulation goes way beyond that. It gets into the race of the applicants. It gets into gender, which will be hard to analyze and it may be sex. It'll be hard to analyze but it goes way beyond like what the statute is. It's going to be far-reaching. To my knowledge, credit unions aren't big in small business lending and they have this rule that loans $50,000 and above are considered small business loans.

I don't know what restrictions credit unions have on mirroring the small business lending that banks do. Nevertheless, there will be some reporting requirements there. You'll have to do the same analysis. Now, it gets a lot of attention because, like mortgages, this will be data that you can compare. You'll have peer data on small business lending that we've never had before. I believe it's going to be a big thing for the lending market as a whole. Maybe less important for credit unions than banks but still, it's going to be something else that credit unions will have to look at and prepare for if they're doing any of those loans.

I did an episode on that topic when the rule came out with Joe Goldberg, formerly of NCUA. He used to run the fair lending exams. Maybe down the road, we can chat about that in greater detail. Credit unions, if I remember right, there's a long runway toward when they'll have to start reporting. The other thing was, as you're saying, the volume of them needs to be pretty substantial so that there will be more banks that it applies to than credit unions based on the number of credit unions that will hit that volume.

That's something also to compare to HMDA. If you have standardized reporting, you don't have that issue we're talking about with indirect lending where one credit union has these 10 data fields and the next credit union has these 14 but not those other 10. It becomes more standardized than when you have that standardized data. It's easier to analyze and maybe do that regression analysis and other things.

I expect less sophisticated regression with small business lending because the sample sizes are going to be smaller. Even if you did a thousand loans, you would be a big small business lender. That doesn't sound like a lot of observations.

WFC 127 | Indirect Auto Lending
Indirect Auto Lending: There could be less sophisticated regression with small business lending because the sample sizes are going to be smaller.

That's a great point. Mike, as always, it's been entertaining and educational. I appreciate your time here. If my audience wants to chat with you and your team, what would be the best way for them to reach you?

Go to the website at They could also email me at Finally, they could call us at (202) 842-3800.

Mike, thanks for giving all your wisdom and all those ways to be reached. I want to thank you for your time.

Thanks for having me, Mark. It’s my pleasure.

For our audience, I want to thank you as always. Hopefully, you'll tune in again soon. This is Mark Treichel signing off, With Flying Colors.

Important Links


bottom of page