Online Event Archive Recorded: July 26, 2025

Watch a member-focused webinar to showcase our new report-generation capabilities. NCRC's Research team demonstrates how they create customized reports for NCRC members. Using data curated for NCRC members' unique needs, we work with each member to produce bespoke reports that target the specific places and issues that matter to them. This webinar demonstrates how we can produce reports that are perfect for working with funders, banks, elected officials and other stakeholders or to form the basis for your own publications. This is an exciting new capability that we are excited to offer.

Speakers

Ralph Cyrus, Membership Engagement Specialist
Jason Richardson, Senior Director of Research
Jacelyn Matthews, Director of the National Training Academy
Devin Thompson , Director of Health Equity & Impact

Transcript:

NCRC video transcripts are produced by a third-party transcription service and may contain errors. They are lightly edited for style and clarity.

Richardson 0:09
Great and thank you very much. Hi everybody. My name is Jason Richardson. I'm the director of research for the National Community Reinvestment Coalition. In just a minute, I'm going to go through a couple of slides on the screen, and then I'm going to stop sharing so you can you'll be able to see everybody. But I want to thank you first of all for joining us today. We're gonna be talking about reports that we generate for members of NCRC, and we're going to give you an example of one that I'm going to do for Jacelyn today. And then we're going to answer some questions about them, but I'm going to let you watch through our screens here as I pull together a report that is exclusively for Jacelyn. It's going to be based on information that she needs to do her job. And I think it's a kind of an interesting process, and I'm hoping that you all agree. Couple of quick notes. First of all, I want to make sure I reintroduce Jacelyn Matthews, who's our National Training Academy director, and she has been gracious enough to be my co-host, organizer and and test member for the day. Ralph Cyrus is also joining us. He's our member engagement specialist. He's going to be here to answer any questions you have about NCRC membership. And Devin Thompson is also available. He can help with Q and A or questions in the chat, and he himself does some interesting work on health or healthcare access and health accessibility.

Here at NCRC, we are recording this, and if you have any questions, put them in the chat. We will have a Q amp a as soon as I'm done. And of course, you can always redirect and just email me if you have any questions after the call. Jacelyn, I think that's it. Are you ready to see… Actually, I should have gone through a couple of slides, which kind of repeats what I just said here. But, and of course, Jacelyn and Ralph are here as well. Go ahead and just shut down the screen sharing, and there we go.

So ever since this I took over this research team about 10 years ago, we have focused on supporting our members with data and analysis and the back end part of that, we build data sets and we build databases. We, you know, we've done for years, but often the challenge was translating that into something that was actionable. Our members are over 700 groups across the country, many of whom don't have their own data and research people on staff, and that means that translating the data that we have into something that makes sense is always been a key area of concern for us. The report we're going to do today is just a sample of the kind of thing that we can do for members. But you know, if you're, if you're ever not sure if you're, if you're a member of NCRC and you're not sure if Jason can help you with something, the safest bet is always just to shoot me an email and ask, more often than not I can find a way to help if we have data to support it. So Jacelyn, with that, I was just going to jump right into it and talk a little bit. All right. So Jacelyn, we're going to I know you had talked to me earlier about doing looking at mortgage lending in your county. Why don't you remind me what it was Essex County, New Jersey. I believe you said, right?

Matthews 3:48
Yes, Essex County, born and raised in New Jersey, so I'm really interested in seeing what's happening in the county now.

Richardson
Right, exactly. So in we're looking at, you said you want to look at mortgages. Are you interested in homeownership patterns, like, who's getting home loans? Or, yeah,

Matthews
Yes, absolutely that. I'm really interested. I'm sorry, go ahead.

Richardson
I was going to say, I was going to ask if you, if you're interested in focusing kind of overall in the county, or do you want to dive into individual lenders and kind of understand individual lenders who are active there?

Matthews
I would really like to see individual lenders and what demographics are actually lending to and at what rates as well.

Richardson
Yeah, we can. We can do all of that. So the mortgage data set we use is from the Home Mortgage Disclosure Act. This data is produced every year by the Consumer Financial Protection Bureau, and it includes about 90% of applications for mortgages across the US. It has. It's a very deep and complex data set, and so like I like to interview, in this case, Jacelyn, just to make sure that we are slicing the data in a way that helpful, because it's one of those data sets where you can look at it in an almost infinite number of ways, but we want to make sure it's of interest to her. So what I heard her saying is that she's interested in home purchases. Who is loaning to new home buyers? Where are they loaning? Who are they lending to, and more importantly, sometimes, who are they not lending to, and what areas are not getting investment. So I'm going to share my screen again here, and we're going to draw upon the NCRC database that goes back to 2007. So we have about 160 million rows of our mortgage application records essentially on file. We also keep databases on things such as bank branch locations, openings and closures over the last 20 years, and we have small business data going back, I think, to 2007 as well.

You know, we get a lot less detail on the business loan data, but it is, it is a resource. In addition to that, we often bring in other sources. Census data is very commonly used as a comparator. We're going to use a little bit of that today. And we often work with members that have their own data. For example, your members that help with things like pre-purchase counseling or homebuying classes, we can pull data that they have under an NDA agreement and compare that to other lender data, for example, or help them better understand their own impact in their community. So all of these are things that we do for members, and let's go ahead and focus on Jacelyn right now. So in front of you, what you see is my screen, and I'm sharing the whole screen here, so you see my little icons and everything there. But in the middle, this is a tableau worksheet, and we just use Tableau as a way to easily link to and manipulate the data that we're going to be using. What you have here are loans as a table showing home purchase, refinance and home improvement lending from 2018 to 2024. It has the number of loans, and then it has several columns that are additional variables that we are interested in. LMI B means a low to moderate income borrower. So that's somebody making 80% or less of the area median income for the city they're in. LMI CT is a low to moderate income census track, which means the income of that track is 80% or less of the area median income. MMCT is a majority minority census tract, which just means that the population is less than 50% non-Hispanic, White, and then the percent of loans to Asian, Black, Hispanic, Hawaiian or Pacific Islander and Native American borrowers.

And we have published several white papers on how we calculate these variables. So if you are a mortgage geek and you are interested, by all means, let me know, and I can go into an excruciating amount of detail about how we do that. But for today, we're going to be focusing on home purchase loans. So let's go ahead and just clip that. And it takes a second there. And then we also just take this off to simplify the view. We also know we want to look at Essex County, New Jersey. Now we can look at just about any geography. So if it's a you know, I've had a member just a few weeks ago that was interested in two zip codes in their city, not in the entire city even, and we were able to clip out just those records. So if you're not sure about the area, let me know, and we can always work together and figure out what, what's going to solve your particular need there while we're waiting on this, that was an interesting case. Jacelyn, this is what I think I did tell you about this one day they were interested in home purchases in two zip codes in Missouri or St Louis, Missouri. And in the middle of the call, they said, you know, we hear that our neighborhoods have more duplexes than anywhere else in the city, and we hear that investors are buying them up. So I was able to take that information and then go back into this data and separate out investor loans, which a lot of investors pay cash for their houses, but a lot of them will pay a large down payment, and they'll still mortgage part of it, so it does show up in our data. And sure enough, we found a huge issue in those neighborhoods of investors buying up duplexes, triplexes, etc. And, you know, it just and then just renting them out, instead of them being owner-occupied.

So okay, on the right-hand side here, you see the data has been adjusted. The numbers are a lot lower because we're just looking at Essex County. These are home purchase loans on site, built, meaning not manufactured homes, one to four unit owner-occupied properties. So this is a good way to kind of get an idea of the overall health of the neighborhood. Now we also want to dive into specific lenders and how they've done over time. But the problem with that sometimes is that there, especially over seven years, can be quite a few lenders out there. So instead, we're going to focus on, I think, the top, let's just say the top 20 in lenders in 2024, and we'll do that, and then I'll bring in the lender names so we can, I'm go there, and let's organize that.

Matthews
Jason, they're asking the chat if you can enlarge your screen a little bit, because they want to be able to see the rows and columns better.

Richardson 11:20
All right, excellent. Let me know if that is a little bit better, guys. All right, and we're going to do a couple of quick things here to make sure that we get just the largest lenders. And you'll note here, right off the bat, you can see the largest lender is guaranteed rate in 2024 with 345 loans, and then cross country mortgage with 343 and then really quickly you see how those numbers decline, until you've got some kind of small entities here that, that, that we're still going to take a look at, but that's kind of a concentrated market where it's a few lenders that really kind of dominate Essex County. And that's the first thing. If I had, like a little red flag to wave. Jacelyn, I would, I would kind of be waving that because, as an advocate, you want there to be more lenders sharing the space. And when you see a concentration, this isn't the worst I've seen, but it's not, you know, it's not great. Let me go ahead and filter.

Matthews
And I'll, I'll say, as you're, as you're speaking. I'm doing this, Jason, for those who are kind of like, this is a lot of numbers. I'm not understanding how to actually understand what the numbers mean. This is a great opportunity to work with our research team, because they can help you make sense of the numbers. And so we don't want you to kind of get overwhelmed with that. We want you to be able to utilize the services that they offer to help you make sense of the numbers for whoever you're using this report for. So if it's for your own benefit, your own organization's benefit, but if it's also for you're having a meeting with elected officials or with other banks or anything of that nature, they can help you make sense of this in the report that they do publish for you.

Richardson
Yep, exactly. Good point. All right, now, so I've, I've gone ahead and I've sorted all the data that I want to use. We've got the top lenders here. Back to here. Hang on one second. Here, and we just go ahead and call them vendors. What I'm doing now is just kind of shaping the data, and this is kind of the more important part of the process, because we can make sure that we are bringing the right thing into the report that we are going to be talking about today. I just want to double check something here, make sure it works right, and then I'm going to stop sharing my screen in a moment, because I'm going to be exporting this data to Excel, just for simplicity sake. Because the next step is, we're going to start working with the AI, and it's a kind of a key step here is to simplify your data as much as possible before you bring it into the AI and make sure you double-check your data there. I know notoriously, many of us have heard crazy stories about AI being asked for things and making things up, usually that has to do with either giving the AI too little context, confusing directions, or poorly formatted data. So we spent a lot of time just making sure that we that we properly format that data. Let me see here, and just takes a minute for these things to export, and then we'll, we'll keep moving. Jacelyn, did you have any other questions?

Matthews 15:00
I do actually. What be the benefit of reaching out to your team to collect this data, versus like me myself, going to Google, typing in some things and putting into Chat GBT on my own?

Richardson
Well, yeah, and so the data is publicly available. You can go to the CFPB website and download the Honda data yourself. Like I said, it is, it is complex data, though, and it's difficult to work with. You know, it comes in, like very large amounts, so you've got to have some facility with data already just to be able to download it, work with it, and kind of interpret it, and then we join it together with a lot of other data sets to kind of give a little more context that are make it more relevant to what you're trying to do. And then with, you know, AI, like I said, Chat GPT or Claude, which is what we're working with today, they need a lot of context that that you really need a you know. If you just give them the data, they won't inherently know what a mortgage is or how to work with it, or they won't know. You know, the AI are just large language models. They help bridge that gap between a subject matter expert and the data really well. So the report we're doing today, and I'm just downloading the other sheet now, is actually not like I've done these reports for years for members, the big shift is now, instead of it taking me 10-15, hours to do a report, by the time we get off this call, you're going to have a report.

So that's a pretty big deal, I think. All right, so I've got the data now in Excel, and I just want to pull up Claude here, and I will start sharing my screen. Now. There are several major large language models out there, ChatGPT from Open AI is probably the one that most people know. This is Claude, which is from a company called Anthropic, and that I prefer it for this kind of work, because, although it's not as good at math as chat GPT is like, I'll use chat GPT if I'm doing coding or if I'm giving it large data sets that I wanted to help manipulate, but it's I think Claude is much better at interpreting data and writing about it in an understandable way. So this is the interface with Claude, and you can see already typed in Essex County mortgage member mortgage report, because I had a little hint from Jacelyn before this that she was going to ask about Essex County, but I just wanted to kind of get it started. So this project in quad already has a lot of information, and I've given it on humda data on the format of the reports that I like, the style guide for NCRC and how we write things is already loaded here, so now it's up to me to give it instructions. AIS like this. Need have a short you want to kind of short sentence structure when you talk to it, and be very clear and explicit. So today we are writing on mortgage purchase.

Matthews 18:22
So as you're doing that, Jason just to explain to everyone else. So Jason is going to put in this information based off of the questions that I have the information I'm looking for. So here you would be able to reach out to Jason and his team to say, like, I'm looking for information on x, y, z, and give try to be as detailed as possible about what you're looking for, and then they can put this into the report. Now we understand that you might be trying to just gather data you're not really sure about exactly what you're looking for, and so once the report is kind of generated, the team can actually help you, kind of like, drill down to the actual key items that you're looking to, to be able to share out within your organization or with others that you see fit. So this is a starting point for them, and so Jason will do like an interview, almost with you or whoever from your team, to kind of talk about what is that you're looking for, what's information that you're seeking, and so that they can actually help you to drill down on the content so that they can answer on some questions in here. And obviously, this is me and Jason have had conversations, so he's able to kind of answer this already based off of, like, just the conversation we've been having. But this is something that is also a member benefit for you. You can ask, like, really pointed and direct questions so that they can actually get you the information that you need. And while Jason's doing that, I'll also mention that if you have any questions, we would ask you to put it into the Q and A box here so that we can get to them. Jason will be sure to answer the questions. And if there's something that comes up, because we know that, and when you're in the middle of a webinar, you're still processing the information. So if you need to reach out to Jason and the research team, you can do so, he will provide his email at the end of our time together. But we do want to let you know that this is a benefit, a perk for being a member of NCRC,

And Jason, actually, while we're doing that, you mentioned HMDA data initially, and can you tell us a little bit more about what that is?

Richardson
I'm sorry, yeah, that's the well, I the data set I spoke about earlier the this Consumer Finance Protection Bureau collects data on about 90% of mortgage loans every year, mortgage applications. And you can go to the CFPB website here, ffiec.cfpb.gov, because it's the easiest URL to remember, and you can go and explore the data yourself.

When we download it, we clean it up quite a bit. We combine multiple years, because generally you can only look at individual years here. But if you're looking for information on a specific lender or area, it is perfectly well within, you know, just anybody's capabilities to go in here and download the data and kind of poke around it. And if you're interested in doing that kind of work, but you're not in your again, Member Services mean a lot of different things. I'm happy to hop on a call with you and walk you through this website and get you started on doing this kind of work yourself.

All right, now we go back here. Claude has kind of looked through and it's asked me a couple of questions here, should I focus on… let me see. I'm going to tell it here - is the county wide data for all wonders - and just give it that. And you'll see this isn't like a single I don't just give it everything, and then it just goes. There's usually a back-and-forth with the AI. And constantly I have to review what it's writing. I look at the the code that it writes if it has to write any sort of code, and this is where the subject matter experience comes in, because I'll spot things here that I know instinctively just aren't right sometimes, and I'll go back and see that either the AI misinterpreted something, or I gave it a you know, instructions that weren't really clear. So right now, I'm just letting it start to write the draft or the report over here, and I still haven't given it all of the data. I still want to give it all of the data on those top lenders that we talked about. So let me copy that data on the top.

All right, you'll notice two things. First of all, my typing is atrocious, but that's okay, because it, you know, clean that up a little bit. But also, I'm asking you to do is kind of a basic analysis of the county lending. It'll talk about how that compares to national lending that it's got records of, and it will also look at the top lenders and talk about how they are doing compared to one another. So there is that data, and we'll give it to it. Here we go,

Thompson 23:45
Jason, while Claude does its thing. Would you be willing to take two questions from the audience?

Richardson 23:49
Sure, of course. Where are they in the chat?

Thompson
So I can read them off to you. So first one, can you also add loans that have gone into default or first legal filing. So is that in the data set separately?

Richardson
So Liz pendants filings would be data you would receive from the county? Yes, you can combine so that you cannot link them with specific loan records. No. But if you're looking to learn more about you know, quantify the number of Liz pendants filings in a particular area. We can help you work with that data to do so. We are seeing, I can tell you, and you can look at Urban Institute's monthly chart book that they put out, we are starting to see defaults, and 90 day defaults going up. They're still historically pretty low, you know, but, but it's Savannah, New Orleans, yeah, I Yeah. I'm sure it's a we're seeing more defaults there. But, yeah, and that's one of those great questions I get a lot. Unfortunately, there's not a national data set of defaults. It's all. Locked, kind of at the county level. But if you've got, if you're working in a county and you want help with that data, let me know. We can, we can take a look at it together and see what's available there.

All right, it looks like it's done now, with some of this, let me go through it, and we're going to now, I've given it all the data, and you see, it's got home purchase lending analysis that's all correct. Now I want to add a little bit of more detail about Essex County, so I've got a couple of reports here that I pulled down earlier. These are from Policy Map, which is an excellent resource that filters a lot of census data and puts it into a much more understandable and dynamic kind of format. So just to give the AI a little more background, I'm going to give these reports to it. Okay.

All right. And what you're going to see here is, on this right hand side, the reports going to start kind of taking shape, but and at this point, like Jacelyn, you see it kind of coming together. I'm including background on, you know, how many people live in Essex County, what their distribution is, what their demographics are. I've also included information here on rents in Essex County, rental burden as a factor. Rental burden means a family that's spending more than half of their income on rent, and it's a good measure of how likely, you know, folks are to be able to buy a home if they can't save if they're spending so much on rent. So we kind of like to include some things like that.

Matthews 27:01
And if, Jason, if my organization, for example, had some data that we wanted to include, could we send the reports to you? You could also include that in here?

Richardson
Yeah, yeah. And I think that's one of the more interesting things. Is when members come with some of their own data and they'll, they'll say, Hey, I got this, you know, we worked with a member called Covenant Community Capital, who are based in Houston. They do pre-purchase counseling, and then their clients work with partner lenders to ultimately make their purchases. And they had great data on all of those clients and what they paid in closing fees. So my team worked together with them, looked at a representative sample of other loans from Harris County, and found that CCC clients were consistently getting lower closing costs, lower interest rates, and they were able to buy more valuable homes. They're able to get bigger houses, and that's a measurable impact that CCC can then take and put in their own reporting and seek funders or other support, just to show what they really are doing.

So here we go. Essex County context. It gives a brief overview here of the race, ethnicity distribution. It shows the change in lending over time. And you can really see the impact of the pandemic here, where home purchase loans spiked by about 10% between 2019, and 2020, and then. And then they die pretty sharply after that. But you still have last year, about 4,500 home purchase loans across the county. We see here, we have a breakdown here, year over year of lending per population. Keep in mind, if we look back, for example, lending to Black borrowers is at just under 24%- 39% of the community is African American, though, so that's a pretty big gap. Meanwhile, you see Asian borrowers are overrepresented, and Hispanic borrowers are right around where we would expect them to be as a percentage. Let me see the first thing is, I didn't include the percent of White in this because we weren't really looking at that, so I'm going to tell it right now just to drop White. And then Native American and Hawaiian or Pacific Islander, you know, in a context like this, those populations are going to be relatively small, so it's a little difficult to really tell anything by that.

All right? There we go. It's a little bit clear, also easier to read. Now, these tables, right now, these reports don't come with fancy maps or graphics or anything like that.

We can do that work, but it's a much heavier lift, and our goal is to make the something that we can deliver to you quickly. but you'll notice these tables - 1t's really easy just to copy and paste this into anything you want to from Google Docs to excel, and you can make your own visuals. We have had members that have taken this as kind of a starting point for reports that they were publishing themselves. And I think that's really exciting, because that's that's something that I think beyond, is beyond the capability of a lot of small groups but it really lets them punch above their weight when they're able to publish original research specific to their community, and they're not just regurgitating stuff that, you know comes out of one of my reports, or things like that.

The geographic distribution is really interesting here. The percent of loans going into low and moderate-income tracks has steadily increased over time, majority minority tracks. It has it has remained relatively stable, 2022 you see this jump here from 44.6% to 54.3%. That's the year that they changed the boundaries for the census tracks with 2020 and it created about 30% more majority minority tracks across the country. So that's probably not a real bump there, you know, but, but the increase in lending and low to moderate income tracks is interesting, especially because we go back to let's actually do this with LMI borrower, because I saw something before when I was looking at this that I think is interesting. There we go. All right, so this table is one I want to focus on for a minute. Jacelyn. The well, still writing here.

Thompson 32:01
Hey Jason, while it's doing that, can I toss you another question that perhaps.

Richardson
Yes, absolutely.

Thompson 32:05
So there was a question on what AI applications are really good at taking raw data and converting it into charts and summaries. So you mentioned that we're not doing that here, if someone wanted to do that on their own.

Richardson 32:22
Yeah, I haven't found anything that consistently will do charts. It's a large language model, you know, that's, that's where all these are. So they, they like, it'll, it'll produce, like, I've had Claude and ChatGPT both produce really beautiful charts, but I couldn't replicate it from one report to the next, and that's not very useful for us. Now summarizing the data, I prefer Claude to chatGPT. I haven't tested Gemini, Devin uses that more. I think, I think Claude is great at taking data and then re and then writing about it in a way that's more understandable for people really quickly. That's just been my experience, though. And keep in mind, this technology is moving quickly enough that in a couple of weeks, it may be a completely different story. I would encourage you just to kind of test out different ones with similar problems and see what works best for you.

But I want to talk about this table for just a minute here, Jacelyn. So what you're seeing here is year over year, the percent of loans home purchases going to low to moderate income borrowers versus those in low to moderate income tracks and majority minority tracks. Now you can already see that there's been an increase, albeit a slight one, over time, for loans and LMI tracks, and we've seen relative stability in lending and majority minority tracks, but lending to low- to moderate-income borrowers has sharply declined from 28% in 2020 to 14% today. That's a pretty, this isn't the worst one I've seen nationally, but this is a pretty severe drop off. But there's not a drop in majority-minority track or LMI track. So do you know what you understand what that means, Jacelyn?

Matthews
No, I'm not sure. Can you explain that?

Richardson
Rich people are buying in poor neighborhoods, essentially. And this is a trend we are seeing nationally, where as low- to moderate-income borrowers get priced out of the market. We're seeing a lot more middle and upper-income borrowers active in LMI, and especially LMI majority minority neighborhoods. Yeah, so you know that's, that's a trend. We're seeing this. But I'm just telling you, Jacelyn, this is a, this is another red flag here. I think we talked earlier about the concentration of the market in just a few lenders, but this severe drop in LMI borrowers is really concerning. And you'll see this kind of goes to that last question also Devin. The text here is, it kind of summarizes the data above it pretty well, I think. And I'm just double-checking some of the figures against my Excel. One thing I will tell you, if you're working with any AI is always to spot check the some of the data points. And just because I liken working with these AI tools now as working with a very, very smart PhD student, basically, or even a very smart high school student, sometimes they can do amazing things, but you often need to double check them and make sure they actually did the work. And any parent out there knows what I'm talking about.

So we've, we've actually already kind of done this, this table here, we've added that to the information upstairs. So let's allow it to remove this table, and then we're going to dive in. I want to make sure we get through the top lender section, and then we have some time for Q and A afterwards. But let's get that through there. Now another thing is, when you talk to me and if we work together on producing one of these reports, and generally speaking, by the end of an hour session, we'll have most of this done. But it doesn't mean that we're done. So if you take the report back, you have time to absorb it, talk to other people and have more questions or you want to tweak it or change it, I keep all of these records. I keep all the files. So just email me and say, Jason, we really like this, but can we add a section that discusses this a little bit? Yeah, not a problem at all. Just shoot me an email, and we can absolutely do that.

Thompson 36:41
Jason, I got one more that I should have asked way back when, but I don't think we're gonna come back around to it. Does it matter how the data is formatted in Excel before you copy and paste it into the AI prompt?

Richardson 37:02
Yeah, so that's a great question, actually. So the data I took out of Tableau was automatically formatted like this, so I've already done the percentage calculations so I know those are done correctly. I'm not relying on the AI to do that, although, if you are going to have an AI do math like that, I would use chatGPT right now. It's just better at it, because it does the calculations using Python. And if you want to do any type of statistical testing; first, you still want to have an expert in statistics review it before you go out and claim statistical significance, but chatGPT an use R, which is a free statistical software language to run the statistical test as well. I would avoid using Claude for that. It's just it doesn't use Python. It codes everything in JavaScript, and it's just not as good at it. But I copy and pasted this, what you see in front of you, right into the chat with Claude, and understood it instinctively, but, but the whole first part of our process today was me shaping the data, cleaning it, and then and then exporting it, so I knew it was in a small enough format that the AI could understand. That's good question.

Here we've got kind of the information on the top lenders and their market share, guaranteed rate and cross country mortgage. This is not unusual looking. We're looking here at what the top eight lenders, and out of eight of them, six of them are mortgage companies. And that's important, and it's important because mortgage companies are not subject to the Community Reinvestment Act, which means that they have no obligation or duty to serve low-income communities that banks are judged on, you know, that banks get examined on. At NCRC, we have long supported expanding CRA to cover mortgage companies as well, just so there is some type of a regulatory framework that ensures that everybody is, you know, being kind of judged on how much they invest in working-class neighborhoods.

Significant contract? Yeah, I noticed here that the eight largest lenders are 40% of the market. So not only did I notice it when we first started looking at this chat, or Claude, in this case, also noticed that concentration is pretty high. And then we're looking at see. So I don't really like this. It doesn't have all of the lenders listed for one thing, and it doesn't have all of the categories. So let's see if we can go ahead and make some changes here, wanting lenders… Well.

Okay, so I'm giving a little bit of instruction here on what I want to see, and I just want to narrow it down to 2024 data when we look at the top lenders. Because if you're going into a meeting, for example, with you know United Shore, and that they're mostly a wholesale lender. So you probably never heard of United Shore, but they're top one or two lenders in the country. Almost every year, they deal exclusively with mortgage brokers. So they, you know, their patterns are often a little concerning, depending on what propelled brokers in the area are doing their loans. But, you know, I always feel like you kind of want to get the more recent data for the individual lenders, and then when I asked it to write up the report, I don't want to be overwhelming people with just repeating that same data that they see in the table. So in this case, I've asked it to kind of, you know, write the narrative, but just cite the most, the most kind of compelling figures that explain what they're seeing. And what you'll note now here. First of all, it's dropped the Native American and Hawaiians from the data set because it's just the numbers are too small. In all likelihood, a great many of these lenders probably didn't make any loans to those groups because the numbers are just, that's just how they break down. But what's interesting is to look at the great variation in lending to different groups, LMI borrowers, for example. Keep in mind, we're talking about the same loans. These are home purchase loans to buy houses, guaranteed rate 8.7% are LMI, cross country 13. See if anybody really stands out. Loan Depot, Paramount residential. They all look pretty good, and then some others that don't look that good, right? You know, and I always find it interesting which ones are great. You'll typically find the banks do a lot less lending to LMI borrowers than others. But look at Bank of America, they're actually doing pretty great here, at about 20% compared to other lenders. So, you know, this is a really kind of interesting space, I think, especially as we've seen this sharp decline to 14% of LMI borrowers in the county overall are low to moderate income. You've got quite a few players here that are doing better than that, and quite a few that are doing worse. So, you know, for example, if I were looking at this with a member, You know, North Point bank, they only made eight loans, though, but you know, we can go back over time and get more information on them. They might make a great partner organization in Essex County, you know, if they really are doing a good job, it looks like, you know, with eight loans, it's hard to tell for one year, I'd want to go back a little bit further with them, but they're, they're the out of those eight loans. They're doing loans to LMI borrowers in LMI tracks, whereas other lenders are not doing as well.

Let me scroll down here and see it does. It does do a little bit lower element borrow rates versus county averages. So it's kind of highlighting some of the lenders here, and this is stuff we can always kind of tweak and change based on what you need. Jacelyn, is this, do you think this information so far is helpful?

Oh, you're muted.

Matthews 43:40
Sorry. Thank you. Yes, I think very helpful.

Richardson 43:49
Great. Well, I tell you what I think I'm going to I'm going to export this right now to what we do is just export it to a PDF file here and save that. And I want to share this with you at the end of the call, and I'm also going to give you the raw data that I pulled here. And that way, you've got not just the written report, but you've got the raw data that I've used to create it that you can use for your own purposes. If you want to use it to double-check some of the figures that are in the report here. I'll also save all of this for us, including all the records of how we produce this report. So in a couple of weeks, after you've had time, and I'm going to stop sharing here, after you've had time to check it out and talk to your colleagues about it, if you have any other questions, or if you spot something you think is maybe off, let me know. And we keep all those records, and I can fix it for you. We can produce a whole new report. But in the last 40 minutes, you now have a comprehensive report on mortgage lending in Essex County from 2018 to 2024 that covers the top lenders in the county, how well they're doing at serving different communities and who's who's not meeting their you know, who has opportunities to improve.

Matthews 45:11
Excellent. Thank you, Jason.

Richardson 45:14
I can finally see the chat.

Matthews
Yeah. And there are some questions in the Q and A that Devin and Ralph were able to answer. But I do think that it might still be worth speaking out loud, especially because we are, you know, doing a recording of this.

Richardson 45:30
There's some good questions here. Yeah, the source of small business lending data, it's FFIEC. Flat files are what we use, which we download from, f, i, U, C, but yeah. Dennis photo was there. Devin was able to answer that one. Lucy, yes, if members of NCRC get free access to the NCRC research team, that's myself, Dr Bruce Mitchell. Myself, Jason Richardson and Dr Bruce Mitchell, as well as Jad, our GIS engineer, and Joseph Dean, who is our economist. And you know, when you can't, if you want to do a report like this, you and I'll work together, but if it's sometimes, I partner you with other people on the team, depending on what what your exact request is, and who's a better skill set match for you. Default for simplifying, I answered that one. Jamila had some good questions there. Okay, Devin, yeah, Devin uses a lot of these tools. Also, I'm a lot so, so So I think that's good answer there, yeah. And I would agree, you know, Claude and Gemini and ChatGPT all have their different use cases, and they seem to be, you know, they seem to excel in certain areas, and that can change with the next model that comes out. So I'd recommend kind of being open to you to moving back and forth between them and testing them out and seeing which one fits your needs best.

Matthews
I'm just going to reiterate her question just again for recording purposes. So the question was about AI application. AI applications are really good at preparing raw data into charts and summaries. And then, do you see using Claude?I do see using Claude. Are there any others that you would suggest to use as well to do summaries and taking raw data?

Richardson 47:34
Yeah. And then, so yeah, the three big ones are chat, GPT, Claude and in Gemini from Google. Honestly, I, you know, it depends on what you're doing. You know, some of them are better than, you know, at certain things than others. I like Claude, and I think that we're going to see the industry kind of develop in a way where different ai do different things, and eventually they'll work together on some things. But for, but for today's demonstration, I just stuck with Claude, just because it was easiest and I was taking care of all the data myself. Co pilot, yeah, that's another one. I haven't I haven't used it very much either. So

Matthews 48:21
That asked a question about looking up data. Can they request data from other counties within your state?

Richardson 48:27
Yeah, that's a great question. And obviously, you know, we can look at any state in the country, anywhere in the US or the colonies the but if you're bringing in specific data for your area, yeah, absolutely we can, we can bring that data into our database. We can relate it to our data in different ways. Yeah, we're happy to, I'd love to take a look at whatever you've got. I I'm always happy to get a, get a look at any data that isn't, you know, just public data that everybody has. You know, proprietary data is something that we're used to working with and we can. We can even have an NDA drawn up if you need, if you need us to, so we can look at the data for you.

Matthews 49:08
Another question is, what are your thoughts about balancing the need to double check the work of an AI with realizing time savings from having AI help create these types of reports?

Richardson
Yeah, that's a super question, because, I mean, that's kind of the crux, if not, why do it? So as we were working on this today, I was able to, since I since I pulled the data myself, and I actually did all the calculations myself in Tableau, and then I just gave that prepared data to Claude to work with. I was able to go through the report as it was writing and cross check a few. You know, what I'll usually do is go through and check several instances of data between my original data I gave it and what it's putting the report to make sure that it's, you know, that it's not making anything up. But when I used to, when I was doing these reports, let's just say manually, meaning that. I would download the data myself, and I would write up the whole report. This would take me a couple of days to do. You know, you're talking about hours of time spent writing and rewriting and going back and creating tables, and we created that report in 40 minutes. So the volume and the quality of the of the product that I can put in your hands very quickly is kind of crazy, to be honest with you. I think we're we're really moving into a to a space now where we're able to provide better tools that are exactly what you need for your local purpose. And you don't have to go to some dashboard somewhere, you don't have to learn how to be a data person. You know, there's none of that. So I'm excited to get started, you know, to do this kind of thing. So I want to encourage folks there, if you're members of NCRC, shoot me an email if you're interested in membership, Ralph, I'm conscious of the time here. Did you want to come on and talk a little bit about membership of NCRC, and we can put, I'll put some contact information in the chat, also for folks.

Cyrus 51:12
Yes. I'll speak real quick. I just wanted to reiterate for people are just starting out, and they're curious on how to connect with research. If there are any other, any other membership benefits that that you're that you're interested in, feel free to reach out to the membership team. We're happy to talk. We're happy to do a one-on-one conversation, walk you through both how to get customized reports, and then also, a lot of times we just like to, we do like to share the premier reports that the research team does. Throughout the year, they recently released a really interesting report on gentrification, that I think a lot of people should, you know, try to utilize first. And the other thing I would like to say is, if you are interested in having some type of data request, and you're curious on how to interpret the results, what to do with the reports is that we, we typically in a locality, also have other members who may be more experienced and reporting and using reports for advocacy. And as much as we want you to collab with us. We would love for our members to to collab among each other, so we can also connect you to your fellow members, some who or who could be research centers, fair lending, fair housing advocacy groups as well, and use these reports as a way to build larger coalitions where you are,

Richardson 53:03
Great. And I was going to put contact info in the chat, but I think we'll just send around. When we send around the recording link, Jocelyn, I'll include, obviously, our contact information for Ralph and myself. And also, there was another question in the chat I wanted to address before we wrap up here. Karen Hughes, do you also integrate qualitative data? Yes, yes. We can the it. It can, obviously, can't be linked directly to, like, in this case, mortgage lending, but yeah, we're comfortable working with all kinds of data like that. We do a lot of work around public health already. You'll know, we've published several reports on redlining and public health and a lot of Devon's work has touched on social determinants of health, in particular that you mentioned there. So yes, absolutely. I'm going to include on the email that we send out later not just the link to this session, but also links to our most recent reports on gentrification. And we are about to debut a new mortgage market series online that looks at different aspects of mortgage lending in 2024. So the first three parts of that, we hope to publish this week, and there'll be more parts coming out over the rest of the year. So make sure you're signed up for emails, and you'll hear about that.

Matthews 54:33
And Jason, I'm going to ask one last questions. I think it is an important one. What is the cost to a member to actually have your team compiled report.

Richardson
It's zero if you're a member of NCRC, this is, this is a core part of what my team is tasked to do. We produce our own research. We support other internal departments here at NCRC, such as the policy team and our leadership and we help members. So those are the three things we do. And if you come to our conference every year, you're gonna you'll see our research area that we set up with. You know, we're with big computer screens and everything like that we usually have, and we spend all day sitting there talking to members and talking about their data needs. But no if you're a member of NCRC, that is the cost. So we're very happy to engage.

Matthews
Great. Thank you and thank you. This was a very informative I know we couldn't get into all the questions, but if you have any additional questions, or, like I said before, if as you're watching this recording, you're like, Oh, wait. I you know, wanted more information about this or that, you can definitely find Jason about the actual research. Jay richardson@ncrc.org if you have questions about membership becoming a member of NCRC, you can reach out to Ralph R cyrus@ncrc.org. We can also provide that information for you as well. And again, thank you for taking the time to meet with us today. We hope this was informative for you as well.

Richardson 56:05
Thank you. Have a good day.