Insights
Market reports, forecast data, industry insights, and more from iEmergent.
Since the 2025 HMDA data came out, we've continued to analyze it from different angles to learn more about mortgage lending across the US. In part 1 of our analysis, we covered overall trends like originations and applications. In part 2, we looked at trends by race and ethnicity. Here in part 3, we’re focusing on income.
Navigation note: Under each subhead, we’ve provided an “in short” summary—a tl;dr (too long; didn’t read), if you will. You can skim these summaries and dig more into the sections you find most interesting.
This is a VERY long analysis. If you prefer to read it as a PDF, you can download that file below: Download PDF
Setting the Stage
About HMDA Income Data
Income Overview: Applications and Originations
Loan Purpose Matters
Application Denial Reasons
Average Loan Size
Originations by Loan Type
Lender Landscape
Overlap of Race and Income
What this Means for the Industry
Subscribe for in-depth data:
Income is one of the biggest single drivers in mortgage lending, and its impact shows up differently depending on where a household falls in the homeownership journey.
The first point where income becomes critical plays out before an application is ever submitted: income intersects with affordability, product access and years of accumulated (or missing) equity to determine if a household is in a financial position to apply for a mortgage in the first place.
The second major point where income has an outsized impact is after the application is submitted: how does income impact which applications make it through the process to become an originated loan and, if originated, how does income influence the characteristics of that loan?
We will dig into these points in the analysis that follows to provide meaningful insight on the past, current, and future mortgage market; how to expand homeownership; and where the next lending opportunities might be.
In short: The market conditions of the past several years didn't hit every income group the same way, which is the backdrop that shapes this analysis.
Every chart in this analysis sits inside the same broader story: as interest rates climbed and home prices stayed elevated over the past several years, the cost of buying or refinancing a home rose substantially. That cost fell hardest on low- and moderate-income (LMI) borrowers, who typically have less savings, less equity, and less room to absorb changing conditions.
The clearest illustration is refinancing. During the pandemic-era refi boom in 2021, only 16% of refinance applications came from LMI census tracts because the borrowers positioned to move fast on historically low rates were generally those who already had equity and strong credit. When refi activity collapsed 84% between 2021 and 2023, LMI's share of applications rose to 22%. That increase reflected a shift in percentages, not greater participation or a surge in opportunity. When converted back to volume, LMI refinance activity fell by roughly the same amount as everyone else's—it just fell slightly less.
2021-2025 Refinance Applications by LMI Tract Category
Purchase lending told a steadier, but still revealing, story. They declined only about 30% at their lowest point, but the LMI share of purchase applications remained fairly consistent: 20–22% over the 5-year period. This is a difference we’ll examine more in the Loan Purpose Matters section of this analysis.
2021-2025 Purchase Share By Tract Income Level
The investigation that follows uses this context to understand the dual-gap paradigm we mentioned at the start of this discussion: Who actually applies for a loan and who ultimately receives approval? Beyond that, we will also explore other questions:
In short: Higher rates and prices have narrowed the path to homeownership, but not for the borrowers you'd expect.
Between 2021 and 2025, mortgage rates and home prices moved against consumers at the same time: the average 30-year fixed rate more than doubled, from below 3% to roughly 6.6%, while prices kept climbing nationally. The result was a dramatically higher monthly payment for anyone buying the same home just a few years later.
It would be reasonable to assume lower-income households absorbed the brunt of that shift, but the data aren’t necessarily linear. Many lower-income households were most likely already well outside conventional qualification thresholds before rates and prices moved, so the changes in affordability didn't shift their odds much one way or the other.
Instead, a large impact fell on moderate- and middle-income households living near the edge of affordability - borrowers with stable employment and enough credit and savings to qualify under favorable conditions, but with little cushion.
A one-point rate increase, a bidding war, a jump in insurance or property taxes—any one of these might be manageable. Combined, they can push an otherwise qualified borrower past acceptable debt-to-income (DTI) thresholds. This is a demand story as much as a supply one: households aren't only choosing not to buy, they're being priced out, and moderate-income households sit in the narrowest band where qualification is possible but never guaranteed.
In short: Where someone lives can matter as much as what they earn, which is often concealed in national data.
National data smooths over something worth repeating: every market is unique. Housing costs vary dramatically across markets and there is a widening gap between what households earn and what homes cost. As home prices outpace wage growth, affordability challenges in high-cost markets extend well beyond low-income households. Moderate-income families who could afford a home elsewhere in the country get priced out, and either can't buy near their workplace or have to commute in from surrounding markets. This is one reason why where someone is trying to buy a home impacts how income level comes into play.
One metric that illustrates this point is HVI (Home Value-to-Income) ratios, which compare median home prices to median household income levels. (The actual calculation is median home value divided by median household income, so if a census tract’s median home value = $800,000 and its median household income = $80,000, the HVI would be 10, which means the median home value is 10 times the median income ).
Here’s an example:
Let’s first look at San Francisco, where nearly all census tracts have a home value to income ratio above 7, with many above 10
In Philadelphia, there are some tracts in the double digits, but most stay under 10.
However, in our home market of Des Moines, only one census tract (right in the heart of the city) has a ratio above 5.
Many housing agencies—from HUD to Fannie Mae and Freddie Mac—define housing affordability as spending less than 30% of your annual income on housing, which translates to a home value-to-income ratio that roughly falls between 2.5 and 4.0. The State of the Nation’s Housing Report for 2026 released by Harvard’s Joint Centers for Housing Studies indicated that the national median home price is lingering around five times the national median household income for the fifth year in a row. This metric illustrates why housing affordability is a barrier in many markets across the country.
This is why “where” matters so deeply and why localized market analysis is vital.
In short: The gaps in this analysis aren't one gap. They're two, and they don't always move together.
It's worth pausing on something we noted at the outset: income affects mortgage lending through two separate barriers, not one. The first is about entry—whether a household is in a position to apply for a mortgage at all. The second is about outcome—whether the applications that do get submitted actually become originated loans. A market can look identical on the surface and still be failing LMI households in two entirely different ways, or in just one of them.
That distinction matters for how you read everything that follows. When we show application volume by income, we're looking at the first barrier. When we show denial rates and pull-through, we're looking at the second. Where the two move together, the story is straightforward. Where they diverge (where entry improves but approval doesn't, or the reverse) that's where the more interesting, more actionable findings tend to live.
With that framework in place, we can turn to the data itself.
Now that we've set the stage, we can dig into the 2025 HMDA data.
Throughout this analysis, we look at income two ways:
For both, the income categories remain the same: Low, Moderate, Middle, Upper.
In short: The application and origination data prove that lower income levels create gaps and barriers to mortgage lending. Fewer low-income households enter the pool, and the ones who do are approved at lower rates.
Let’s first look at trends and patterns that capture the impact of income on who actually enters the homeownership household pool. The first two charts we are comparing highlight a very important point: that data on borrower income and tract income don't always tell the same story.
The chart on the left demonstrates a predictable pattern: low-income applicants generate the fewest applications (11%) and upper-income applicants generate the most (42%). However, the chart on the right, which compares applications by tract income level, indicates that the most applications are generated by middle-income tracts (5.2 million), which account for 45% of all applications. Low income tracts generate only 3% of all applications, a much lower penetration rate than the 11% generated by low income applicants.
2025 Applications by Income Level: Applicant and Tract Segments
Borrower income describes the person applying. Tract income is a neighborhood average, which naturally compresses results toward the middle. There are simply more middle-income tracts nationally than low-income ones, which pulls tract-level volume in the same direction.
A few other factors likely widen the gap further:
Where the two views diverge in the charts is not really a contradiction, but more a confirmation that we are looking through two different lenses.
As borrower income increases, both loan count and total dollars increase. Unsurprising on its own, but the gap widens sharply once you look at dollars. Upper-income borrowers originated about 6x as many loans as low-income borrowers in 2025, but accounted for nearly 18x the total loan volume.
2025 Loans and Dollar Distribution by Borrower Income
The five-year trend adds context: every borrower income group saw originations decline between the 2021 refi peak and the higher-rate environment that followed. The rates of decline were significant for all—between 47% and 58%.
2021-2025 Loan Count By Borrower Income Level
Neighborhood income and borrower income are related but not the same—borrowers of many income levels live in the same communities, so tract-level data shows where mortgage activity is concentrated rather than who is borrowing. In 2025, middle-income tracts generated the most originations (3.05 million), followed by upper-income tracts (2.41 million), moderate-income tracts (just over 1 million), and low-income tracts (170,000). Dollar volume followed a similar pattern: upper-income tracts led at $980 billion, then middle at $831 billion, moderate at $263 billion, and low at $39 billion.
2025 Loan Counts and Dollars by Tract Income Level
The five-year trend shows an interesting asymmetry: upper-income tracts saw the largest percentage decline in originations (37%) since the 2021 peak, while low-income tracts saw the smallest (24%). That's consistent with a market where households already holding equity—concentrated in middle- and upper-income areas—were better positioned to take advantage of 2021's low rates in the first place, so they had further to fall when conditions tightened.
2021-2025 Loan Count By Tract Income Level
In short: What happens after an application is another factor in analyzing how income intersects with lending, with outcomes often worse for lower-income borrowers and tracts.
Applications are only the entry point. What happens next is critical to understanding any market gaps in housing. Which applications become originated loans, and which are denied, withdrawn, or never completed?
Fallout rates rise as income falls, for both applicant income and tract income, with the gap more pronounced by applicant income.
2025 Applicant Income vs. Tract Income Applications & Fallout Comparison
Many of the tract vs. applicant income comparisons we have examined so far treat the applicant-vs-tract distinction as roughly constant—a consistent dilution effect from averaging individuals into a neighborhood figure. The five-year application action trends by income levels clarify that this comparison is not simple and constant. When looking at application pull-through, it is clear that size of the gap depends heavily on where you are in the income distribution.
Pull-Through and Denial Rate Differences: Low Income Applicant vs. Tract
The two income-level views diverge sharply at the lowest income level, and that divergence has widened over time. The origination rates for the tract-level low income category fell within a narrow band: 50–56% across 2021–2025. Applicant-level low income origination rates fell further and settled lower, 39–43% for most of the period.
2021-2025 Application Actions by Tract Income Level
2021-2025 Application Actions by Applicant Income Level
Denial rates tell a similar story with more force: the tract-level denial rates for low-income tracts ran 18–25%, while applicant-level denial rates for low-income borrowers ran 27–39%. A gap that was 9 points in 2021 and had grown to 13–16 points by 2023–2025.
Denial Rate Gaps by Income Level (Comparing Applicant to Tract Segments)
The moderate income segments show almost no gap at all—tract and applicant lines sit within a couple points of each other in every year.
Middle and upper income modestly reverse the pattern: applicant-level outcomes run slightly better than tract-level (2025 upper: 61% tract vs. 64% applicant originated; 15% vs. 12% denied).
The practical implication: tract-level low-income statistics typically understate how much worse outcomes actually look for the low-income borrowers living in those tracts—and that understatement has been getting larger, not smaller, over the past several years.
In short: Purchase lending shows the smallest income-based gaps at every stage (application, denial, and origination) because of loan programs, lending initiatives, and loan saleability.
Compared to refinance, home equity, and home improvement loans, purchase loans show the smallest differences across income groups at every stage of the process—largely because the industry has intentionally built products around expanding homeownership. Down payment assistance, community lending programs, and government-backed products like FHA, VA, and USDA loans are heavily used by LMI borrowers and concentrated almost entirely in purchase transactions. There's no equivalent infrastructure for refinance, home improvement, or home equity lending.
Those other loan purposes also require the borrower to already own a property with sufficient equity, which is exactly where the structural disadvantage compounds. Borrowers in lower-income communities are more likely to own lower-valued homes, carry higher loan-to-value ratios, and accumulate equity more slowly, which widens the income gap for these products well beyond what current income alone would predict.
Loan purpose also determines what happens on the secondary market: eligible purchase and refi loans can be sold to Fannie and Freddie, giving lenders liquidity and standardized underwriting. Home equity and home improvement loans don't have that outlet, which creates different incentives for different lender types. More on that in the Lender Type section later in this analysis.
Income influences all mortgage outcomes, but loan purpose often determines which products are available, how lenders evaluate risk, and, ultimately, who is able to access credit.
Purchase applications were the most resilient segment, moving from 7.5 million in 2021 to 5.1 million in 2024 before stabilizing in 2025. The LMI share stayed remarkably steady at about 21% throughout.
2021-2025 Purchase Applications By Census Tract LMI Status
Refinance moved the opposite way: applications collapsed from 13.6 million in 2021 to just 2.2 million in 2023, with a slight bounceback to 3.8 million by 2025.
2021-2025 Refinance Applications By Census Tract LMI Status
LMI's share of refi applications rose from 16% (2021) to 22% (2023) before slipping back to 20% (2025). Again, this is not because LMI refinancing surged, but because non-LMI activity fell further during the collapse and rebounded faster once conditions improved. Market recoveries aren't evenly distributed: when rates drop and refinancing becomes attractive again, non-LMI homeowners are the ones positioned to move on it.
Home improvement and "other" purposes rose in volume from 2021–2023 with LMI share up about 2 points, then home improvement volume dipped slightly from 2023–2025 while HELOC/home equity ticked up. LMI penetration held steady at 19% and 18% respectively.
2021-2025 Home Improvement & Other Applications By Census Tract LMI Status
In short: Borrower commitment varies by loan purpose, with purchase loans having higher completion rates and less likelihood of withdrawal or incompletion.
When it comes to the impact of income on which applications are originated, the size of the income-based gap depends enormously on what the loan is for. For all loan purposes, fallout was still mostly driven by denial rates. Differences in withdrawal, incompletion, and unbooked application rates between LMI and non-LMI tracts were narrow across the board.
2025 Fallout Rates and Type By Purpose and Tract LMI Status
Purchase has the narrowest fallout gap of any purpose (9.8 points in 2025 for differences in tract income levels, but +22 points between applicant income levels). It’s the one category with deep program infrastructure (FHA, VA, USDA, DPA) built specifically for this problem. Where the infrastructure exists, the gap is smallest.
2025 Purchase Application Fallout Rates and Type by Tract Income Level
2025 Purchase Application Fallout Rates and Type by Applicant Income Level
Refinance fallout rates sit in the middle (+11 points for tract income levels, and +20 points for applicant income) and tracks the rate cycle, but with a twist. In a high-rate environment, a lower-income borrower's reason for refinancing often shifts from "better rate" to "need cash," which is a harder transaction to underwrite and may explain why the gap widened again in 2025 just as rates began easing.
2025 Refinance Application Fallout Rates and Type by Tract Income Level
2025 Refinance Application Fallout Rates and Type by Applicant Income Level
Home Improvement and Other show the widest gaps by far. For home improvement loans, there is a 45.5% denial rate for low-income-tract applicants in 2025, nearly double the 26% rate for upper-income tracts. The gaps are even higher for home improvement between low and upper income applicants: a 50% denial rate for low income applicants, which is far more than double the 21% denial rate for upper income applicants.
Other loans (HELOC and home-equity products) show wide gaps as well. These are largely home-equity products, and equity is the constraint. It's accumulated unevenly, and there's no FHA/VA/USDA-style safety net for this category the way there is for purchase. Thus, we see wide fallout rate gaps by tract income (a +15 ppt difference in fallout between low and upper income levels) and even wider gaps by applicant income (a +21 ppt difference between low and upper income applicants).
2025 Home Improvement & Other Fallout Rates and Type by Tract Income Level
2025 Home Improvement & Other Fallout Rates & Type by Applicant Income Level
Because purchase loan size tracks home value, this gap compounds forward. A smaller home today builds less equity over time, and equity is exactly what home improvement and refinance loans draw on later. These gaps predate the recent rate/price environment; they show up in every year of the dataset, and rising home prices have turned a fairly stable percentage gap into a growing dollar gap. Closing the entry gap (getting more LMI-tract households to "yes") doesn't by itself close the value gap. The wealth-building disparity can persist even as headline approval numbers improve.
This lines up with the borrower-vs-tract distinction discussed above. DTI, collateral, and incomplete-application status are tied to an individual borrower's actual finances, so they show a real LMI gap at the applicant level that gets diluted once averaged into a neighborhood's income label. Credit history and down payment, by contrast, seem to track with income regardless of whether you're measuring the person or the area.
The trends, rates, and reasons for purchase, refi, and home improvement/other loans aren't one story—they're three, each pointing to a different lever:
Understanding where applications stall points toward specific interventions: earlier support, clearer documentation, affordability counseling, more active follow-up. Reducing fallout among applicants already in the pipeline may be one of the most immediate opportunities to expand LMI lending.
In short: Lower income borrowers face higher denial rates driven by income-related factors, with denial reasons reflecting affordability challenges.
Across purposes, the starkest difference in why applications get denied shows up not between LMI and non-LMI tracts, but between LMI and non-LMI applicants. In 2025, DTI drove 49% of denials for LMI applicants versus 28% for non-LMI applicants, a 21-point gap. By tract income, the same comparison is far smaller and runs the other way (34% LMI tract vs. 38% non-LMI tract).
Collateral and incomplete-application denials follow the same shape—a real gap by applicant income (14% vs. 21% for collateral; 10% vs. 16% for incomplete applications) that nearly disappears by tract (18% vs. 17%; 12% vs. 14%). Credit history and down payment are the exceptions. Both show similar-sized gaps regardless of which income lens you use (31% vs. 28% and 33% vs. 28% for credit history; 6% vs. 5% for down payment in both cuts)
Denial Reasons by LMI Status by Applicant Income vs. Tract Income
Changes in denial reasons over time by income level provide even more insight into how changing economic conditions impact a household’s accessibility to lending products. First, let’s look at the difference in denial reasons between LMI tracts and non-LMI tracts over time for purchase loans specifically.
Trends in Denial Reasons: Purchase Applications in LMI vs. Non-LMI Tracts
Denials due to down payment grew at a similar rate in both tract groups, with non-LMI actually edging ahead: LMI tracts +66.4%, non-LMI tracts +71.34%. Not a dramatic gap, but non-LMI grew faster, which cuts against the intuition that affordability pressure would hit LMI down-payment capacity hardest.
Trends in DTI-driven denials show a bigger difference. LMI tracts were essentially flat over the period (-0.13%, bouncing between 36–39%), while non-LMI tracts grew a real 7.14%.
Some potential explanations for why these two reasons increased most quickly for non-LMI tracts:
Trends in Denial Reasons: Purchase Applications by LMI Applicants vs. Non-LMI Applicants
Denials due to down payment grew dramatically faster for LMI vs. non-LMI applicants. LMI tracts showed down payment as a reason 8% of the time in 2021, but increased to 14% by 2025, for an increase of +82.4%, compared to the increase for non-LMI applicants by 58.87%.
Similarly, LMI applicants were denied due to credit history in 2025 at a much faster rate than in 2021, a trend not seen by non-LMI applicants.
Comparing the wider changes in LMI applicant-level down payment frequency against the changes in LMI tract-level down payment frequency suggests that many of these income-sensitive denial reasons are more volatile when looking at differences in applicant-level incomes. In fact, these swings can also be seen for other denial reasons (credit history, employment, collateral, etc).
This finding suggests that a "non-LMI tract" average is blending a much more heterogeneous population than a "non-LMI applicant" average. So if the mix of who's buying in non-LMI tracts shifted over 2021–2025 (more moderate-high earners getting squeezed into buying there as prices rose, alongside the usual high earners), the tract-level number could swing more than what a typical individual non-LMI applicant experienced. LMI tracts don't have that same compositional slack, since the income band itself is narrower, which is likely why the dilution effect runs cleanly in one direction there but not on the non-LMI side.
In short: Across the country, average loan size is higher in non-LMI census tracts compared to LMI census tracts. The gap is roughly double the size when measured by the borrower's income instead of the neighborhood's.
Unsurprisingly, across loan purposes, average loan size in LMI census tracts is lower than in non-LMI CTs. Because purchase loan size tracks with home values, this gap plausibly compounds forward: a smaller home today builds less equity over time. And equity is exactly the resource that home improvement and refinance loans draw on.
The loan-size gaps were evident before the rate/price environment of the past five years, but rising home prices are causing this gap to widen with time. Because of this, closing the application gap (getting more LMI-tract or LMI-applicant households to "yes") doesn't by itself close the value gap, which means that the wealth-building disparity may persist even if the origination penetration rate of LMI segments increases
2021-2025 Average Loan Size By Tract LMI Status and Purpose
2021-2025 Average Loan Size By LMI Applicant Status and Purpose
All loan purposes in both LMI and non-LMI tracts have increased in average size—more on that in Part 1 of our HMDA data analysis. For reasons we discussed in the Loan Purpose Matters section, the graph shows that the "Other" loan purpose category in LMI census tracts has the largest relative gap, with non-LMI census tracts seeing averages 62.8% higher than LMI CTs. That gap grows to 112.5% when compared by applicant income instead of tract.
Purchase loans have the next largest gap, with loans in non-LMI tracts averaging 46.2% higher than in LMI tracts, a premium that more than doubles to 102.8% by applicant income. Even when LMI-tract borrowers are entering the purchase market at close to the same rate as non-LMI-tract borrowers, they're buying into a meaningfully smaller-value home. And the borrower they're standing next to, income for income, is buying into one smaller still.
This compression between tract and applicant views isn't unique to Other and Purchase. Refinance loan size runs 36.3% higher in non-LMI tracts but 114.4% higher by applicant income; Home Improvement runs 36.8% higher by tract versus 90.7% by applicant. The trend adds another layer: for Purchase, Refinance, and Home Improvement, the applicant-level gap grew meaningfully between 2021 and 2025 — Refinance moved from a 70.7% premium to 114.4%, the steepest widening in the dataset — while the tract-level gap for those same purposes barely moved. Other is the exception: its applicant-level gap actually narrowed over the same period (140.6% to 112.5%), tracking its tract-level gap down as well.
Rising home prices tend to hurt LMI neighborhoods disproportionately, separating them from everyone else. As home prices have climbed, a fixed percentage gap turned into a growing dollar gap for most purposes at the tract level—and, for three of four purposes, an actively widening percentage gap at the applicant level. This shows us that LMI borrowers are financing a shrinking share of the market's total value growth even where their odds of approval haven't changed.
Affordability pressure has made borrowing harder across the board, and it has also made the gap between groups more expensive in real terms. Getting more LMI households to a "yes" is necessary but not sufficient if the home they're financing is worth a third (or, by applicant income, roughly half) less. The home value gap doesn't stay contained to the purchase transaction. The inequity resurfaces years later as a wider equity-access gap. Closing the access gap without closing the value gap risks leaving the deeper wealth-building disparity untouched even as the headline entry numbers improve.
In short: Government loan products are a saturated, steady share of LMI-segment purchase lending. They are now growing in the non-LMI segments, suggesting the safety net LMI buyers have long relied on is increasingly being reached for by buyers who didn't used to need it.
Earlier in this analysis, we referenced the trend that FHA, VA, and USDA-style products are heavily used by LMI borrowers and built specifically to support them. Let’s examine the trajectory of these government lending products over the past five years.
Trends in Government Loan Type Purchase Loan Penetration by LMICT Segments
Trends in Government Loan Type Purchase Loan Penetration by LMI Applicant Segments
In LMI tracts, total government-product penetration has held essentially flat across the entire period—31% in 2021, 33% in 2025, never dipping below 30% or climbing above 33%. In non-LMI tracts, the same figure has grown by 4 points, from 23% to 27%, approximately a 17% relative increase. FSA stays a small, steady 1–2% in both groups. VA edges up slightly more in non-LMI tracts (8% to 10%) than in LMI tracts. FHA is what drives the overall pattern: 22% to 24% in LMI tracts (essentially flat) versus 13% to 16% in non-LMI tracts (up 3 points,which is approximately 23% growth).
The flat LMI line is a sign that LMI-tract purchase buyers were already leaning on government products at high levels before 2021, which lines up with why LMI-tract down-payment denial growth stayed comparatively modest even as affordability worsened broadly. Government programs were already in place and already being used. Non-LMI-tract buyers are visibly reaching for the same products at a growing rate for the first time, which is the flip side of why the increase in down-payment denials in non-LMI tracts actually outpaced the increase in down-payment denials in LMI tracts.
There is a slightly different trend with LMI applicants, where we do see a drop in government lending from 2021 (35%) to 2025 (33%). Interestingly, the driver of this drop is not FHA or VA, but a 50% drop in the rate at which LMI applicants are utilizing the FSA rural loan products. That same drop is not seen with non-LMI applicants, because their use of FSA products has never risen above 1%. What we do see in the applicant-income level data is the same increase in non-LMI applicants using FHA and VA lending for what is most likely the same reason as the increase in non-LMI tract lending: that housing affordability barriers have pushed higher-income borrowers into loan products that were designed to overcome those barriers.
In short: IMBs seem to be growing in dominance, but it's concentrated almost entirely in the categories they've historically been weakest in and it is happening at the same pace for LMI and non-LMI applicants alike.
In this section, we’ll look at what lender types and lenders are succeeding with LMI applicants and census tracts.
In short: Each lender type’s penetration rates vary in LMI segments by applicant income vs. tract income.
Let’s start by analyzing trends in applications by lender type: is one lender type doing a better job than others bringing LMI applicants or the people living within LMI tracts into the mortgage lending pool?
IMBs dominated the purchase and refinance markets in application share in both LMI and non-LMI lending. Their highest market share across all purposes, LMI segments, and years was achieved in 2025 with an application share of 73% in both purchase and refinance lending in LMI census tracts. Much of this can be attributed to a product mix that includes more FHA, VA, and similar programs that are often used by first-time and lower-income borrowers.
2025 Application Penetration Rates by LMICT Status, Purpose, and Lender Type
2025 Application Penetration Rates by LMI Applicant Status, Purpose, and Lender Type
All three lender types demonstrated higher LMI segment penetration rates with LMI applicants than within LMI census tracts. Of the three lender types, banks had the widest divergence in applicant-vs-tract LMI penetration across all loan purposes, although the gap was the highest in home improvement/other lending.
In short: Like originations, IMBs continue to grow share in applications.
Let’s look at applications for each lender type (IMBs, banks, credit unions) over the last five years, by loan purpose, and by LMI status.
Trends in Application Share by Lender Type, Loan Purpose and LMI Tract Status
Trends in Application Share by Lender Type, Loan Purpose and LMI Applicant Status
IMBs The expansion of IMB market share is not LMI-specific—it's happening at roughly the same rate for LMI and non-LMI applicants. Within any given purpose and year, the IMB share of LMI and non-LMI segments are within a few points of each other. These data suggest that the driver of growing market share of home improvement/other applications isn't a deliberate LMI push. It looks like a broader market-share play that happens to be pulling both income groups along together.
Banks From 2021 to 2025, banks lost application share across the board: every purpose, every LMI segment, including both applicant and tract income views. This is the most consistent trend across all of the data, although the largest drops in application share are in the two loan purpose categories where they have historically been the strongest. In LMI census tracts, banks lost 7% and 16% share in home improvement and other loan purposes, respectively.
Credit Unions Credit unions show application share that is either flat or growing across most segments and purposes. The largest increases in application share were seen in refinance lending for both LMI and non-LMI applicants and census tracts. Their only drop in application share was seen for the other loan purpose in LMI census tracts, where share dropped from 30% in 2021 to 28% in 2025.
In short: Application and origination shares can differ thanks to fallout during the process.
Purchase applications show the most stability across loan purposes by every measure, reinforcing what we have seen throughout the analysis. It has the narrowest fallout across all lender types and income cuts, the most stable market share trend, and the smallest applicant-vs-tract divergence.
2025 Application Fallout by Lender Type and LMI Tract Status
2025 Application Fallout by Lender Type and LMI Applicant Status
IMBs show a consistently lower denial rate gap than banks or credit unions across loan purposes, especially between the LMI vs. non-LMI census tract segments. Tract-level 2025 data shows IMB's LMI-vs-non-LMI denial gap flat at ~3 points across all four purposes. However, the LMI applicant-income data shows gaps ranging from 5 to 11 points across different loan purposes. In purchase loans specifically, banks actually demonstrated a narrower denial gap (8 points) compared to IMBs (9 points).
IMBs’ refinance numbers complicate any simple "IMBs do better with LMI segments" narrative. Despite the narrow denial gap, IMBs had the highest total refinance fallout of the three lender types (58% LMI applicant-level; 56% in LMI tracts), driven by withdrawn (20%) and incomplete (9–10%) applications, not denials. This suggests there may be a process/follow-up problem on the operations side, as opposed to limitations to credit access.
In short: Each lender type has a differently shaped trend when it comes to LMI lending.
Applications and fallout tell us who's trying to get a loan and what happens to them along the way, but fallout doesn't hit every lender type or purpose evenly, which means the lender landscape at origination won't simply mirror the lender landscape at application. IMBs’ elevated refinance fallout, for instance, means some of the application-share gains shown above don't necessarily carry through to closed loans. This section picks up where applications leave off: which lenders are actually originating loans to LMI borrowers, and whether the patterns in market share and product mix hold once fallout has done its work.
2025 Loan Counts by Lender Type and Applicant Income Level
2025 Loan Counts by Income Level and Type
The tract and applicant level income views both show that IMBs have the highest market share across all income levels in both views, followed by banks and then credit unions. However, the distribution of market share is quite different between tract income and applicant income.
By tract, IMB's share declines cleanly and monotonically as income rises—from 66% of low-income tract share to 54% in upper-income tracts. It’s the kind of pattern that supports a simple "IMB dominates the LMI end of the market" narrative.
By applicant, however, that story doesn't hold. IMB's share is actually lowest among low-income borrowers (56%), climbs through moderate and middle income (60–61%), then drops back down for upper income (52%). It's roughly hump-shaped. The tract-level story—that IMB's market position is strongest at the bottom of the income scale—is not what the applicant-level data shows. If anything, IMB's applicant-level strength is centered in the moderate-to-middle band, not the low end
Market share for banks and credit unions show the same kind of shape mismatch, just less dramatically: Bank share rises fairly smoothly with income by tract (24%→24%→26%→30%) but dips in the middle before jumping at applicant level (29%→25%→23%→31%). Credit union market share is flatter and more consistent by applicant (15%→15%→16%→16%) than by tract (10%→13%→16%→15%).
In short: Origination share tracks application share almost exactly, with the exception of refi. Overall market share for IMBs is strongest in purchase and refi, reflecting how IMB use of government lending products matches well with the needs of lower-income borrowers.
2023-2025 Loan Share by Lender Type, Loan Purpose, and Tract Status
2023-2025 Loan Share by Lender Type, Loan Purpose, and Applicant Status
Purchase and Refi IMB dominance is strongest exactly where government products are concentrated. While IMBs have the highest market share in both purchase and refinance across all income categories, their market share in LMI purchase is the highest (72% in the LMI tract segment, and 71% with LMI applicants). Those are the purpose/income categories where lenders that lean heavily into agency products outperform those that rely more on conventional lending products.
IMBs still show strong origination market share in refinance overall, but the elevated withdrawal/incomplete rate (which we saw earlier in the Application Fallout by Lender Type charts) eats into their origination share relative to their application share. Purchase and Home Improvement show almost no gap—application share and origination share are nearly identical.
The tract-vs-applicant divergence we found in application refinance share shows up again here in originations. Tract-level IMB LMI refinance share runs ahead of applicant-level, and the gap widens over time: 3 points in 2023 and 7 points in both 2024 and 2025. Purchase, other, and home improvement loans all show only a 0-3 point tract/applicant gap by comparison. Refinance is distinctly the outlier purpose for this divergence, confirmed now at both stages of the pipeline.
Home Improvement and Other The picture changes dramatically for home improvement and other home equity-related lending. Regardless of borrower income level, banks originated nearly half of all home improvement loans, while credit unions accounted for roughly another 42% to 43%. IMBs represented only about 11% of this market. A similar pattern appears for the "Other" loan purpose category, which consists primarily of HELOCs and other home equity lending.
Although IMB market share in these categories is increasing, banks and credit unions still possess more market share. This is partly because they are able to retain these loans on their balance sheets, but also because home improvement/equity originations also frequently grow out of existing banking relationships with borrowers who already have checking accounts, savings accounts, or other consumer lending relationships with their financial institution.
In short: Concentration at the top continues and ranking lists depend on LMI segment and loan purpose.
For lending to LMI census tracts (LMICTs), concentration at the top is almost identical to the all-tract picture. United Wholesale Mortgage (UWM) and Rocket alone accounted for 13.21% of originations in 2025, up slightly from 2024, where they combined to take 12.8% of all LMICT loans. This means approximately 1 in 8 LMI-tract loans is originated by one of these two lenders.
2025 Top 25 LMICT Lender Ranking
2024-2025 Top Lender LMICT Comparison
Bank and credit union presence at the top of LMICT lending dropped year-over-year while IMB dominance grew.
One interesting takeaway: The entities doing the most LMICT lending (IMBs) sit largely outside CRA's reach, since CRA exam obligations attach only to depository institutions. The primary regulatory tool built to hold institutions accountable for LMI-community lending has limited leverage over the entities actually setting the market's terms in those communities.
2025 Top 25 LMICT Lender Ranking by Purpose
Purchase: UWM (7.49%) and Rocket (3.93%) maintain strong positions in a heavily IMB-dominated Top 10, with only Chase breaking in at #9.
Refinance: Rocket took the #1 slot in refi, with 12.89% share (46,330 loans) and UWM took #2 at 9.77%. The top 2 lenders each have roughly 3-4X the share of the #3 lender, Freedom Mortgage.
Home Improvement/Other: The top 10 in these loan purpose categories are primarily depositories, although Figure Lending (IMB) is #2 with 3% share. Of the Top 25, only 8 are IMBs and 4 are credit unions.
Now let’s pivot away from the census tract level to look at lender trends with LMI applicants.
2025 Top 25 LMI Applicant Lender Ranking
The IMB market share of the top 30 lenders for LMI applications was lower than the tract-level figure, but the difference is closing.
2024-2025 Top Lender Comparison: LMI Applicants
What happens when we add loan purpose into the applicant-level mix?
2025 Top 25 LMI Applicant Lender Ranking by Purpose
Purchase: The purchase ranking for LMI applicants looks relatively similar to the ranking for LMI tracts, although Chase is ranked much lower at #14, with only 1.2% share, and US Bank’s 0.7% share brings it to #19.
Refinance: Rocket's share jumped to 17.39% (83,486 loans), which is 35% relatively higher share and nearly 80% higher volume than its tract-level number. UWM's share actually dropped (8.37% vs. 9.77% tract) despite a higher raw count, meaning the applicant-level refinance market is simply larger overall.
Home Improvement/Other: The bank in the #1 slot with LMI applicants flips to Citizens Bank (3.17%), with PNC dropping to #2 (3.03%). Rocket and loanDepot both jumped into the top 5 at this applicant level despite ranking #6 and #10 at the tract level.
This analysis is about income, not race — but race compounds many of these same dynamics and deserves its own treatment. See Part 2 of our 2025 HMDA analysis.
What the income lens alone shows is a market where barriers are specific and locatable:
Knowing where the barriers sit, purpose by purpose, is what makes them addressable. And figuring out what to do about them is the next conversation.
Income clearly shapes mortgage lending outcomes—who applies, who's approved, which products they use, which lenders they work with, and even why applications get denied. But one of the biggest takeaways from this analysis is that income rarely acts alone. Two households with identical incomes can have dramatically different homeownership opportunities depending entirely on where they live. National HMDA data shows broad trends, but it can't explain why one metro is expanding homeownership while another struggles with affordability, or why one neighborhood has strong lending activity and while another does not.
That's why local market intelligence matters.
For lenders, the opportunity comes from understanding which borrowers are underserved, where they are located, what barriers they face, and how those barriers differ from one market to the next. In some communities, affordability may be the primary challenge. In others, it may be housing supply, down payment constraints, product availability, or high application fallout rates. The right strategy depends on the local market. And looking forward, not just backward, is what helps lenders decide where to grow next.
Mortgage opportunity forecasts, combined with localized demographic, housing, affordability, and lending data, help lenders identify emerging opportunities and build sustainable strategies to expand homeownership.
Expanding lending in LMI communities is a business opportunity. Millions of future homebuyers will come from today's underserved households and neighborhoods, and the institutions that understand those markets at the local level—and plan for where they're headed rather than where they've been—will be the ones best positioned to grow alongside them.
If you want access to analyze HMDA data, or to view what’s ahead in individual markets, you can do so in Mortgage MarketSmart. Not a user yet? Schedule a demo now.