A high-cost mortgage is a loan secured by a principal dwelling that crosses one of three price triggers set in Regulation Z. The triggers cover the interest rate, the points and fees, and the prepayment penalty. Cross any one of them and the loan becomes a high-cost mortgage, also called a HOEPA loan or a Section 32 loan, which brings additional disclosures and a list of loan terms the lender can no longer use.
The tests measure price. They say nothing about how the loan was sold.
A mortgage is high cost if it meets any one of these tests, found at 12 CFR 1026.32.
1. The rate trigger. The annual percentage rate exceeds the average prime offer rate (APOR) for a comparable transaction by more than:
2. The points and fees trigger. Total points and fees exceed:
3. The prepayment penalty trigger. The loan permits a prepayment penalty more than 36 months after consummation, or a penalty that can exceed 2 percent of the amount prepaid.
The two dollar figures are adjusted every January for inflation. The percentages and the APR spreads have held since January 2014. The $50,000 personal-property figure is nominal and does not index, which matters for manufactured housing and comes up again below.
Note that the rate test uses APR, not the note rate. APR carries the fees, so the first and second triggers are connected. A loan can clear the rate trigger because of its fee load, and the shorter the term the more pronounced that effect becomes.
Since the Dodd-Frank amendments took effect in January 2014, coverage includes purchase-money mortgages, refinances, closed-end home equity loans, and home equity lines of credit secured by a principal dwelling.
Excluded: reverse mortgages, construction-only loans, USDA Section 502 Direct loans, and loans where a housing finance agency is the creditor.
That is broader than the original 1994 statute, which reached only refinances and home improvement loans. Purchase mortgages sat outside HOEPA for its first twenty years.
Crossing a trigger carries real operational weight, which is the main reason volumes stay low.
Before closing, the borrower must receive homeownership counseling from a HUD-approved counselor, and the creditor must obtain written certification that it happened. The loan cannot carry a balloon payment, with narrow exceptions. It cannot carry a prepayment penalty. Points and fees cannot be financed into the loan. Late fees are capped at 4 percent of the past-due payment. Fees for modifications, deferrals, and payoff statements are restricted. The creditor cannot recommend default on an existing loan being refinanced.
For most lenders the practical effect is a line to stay behind rather than a disclosure to prepare.
These four terms get used interchangeably and mean four different things. Sorting them out is most of the analytical work.
A loan priced fairly for risk can land above a trigger. A loan sold to a borrower who did not understand its terms can price comfortably below every one of them. Several federal regulators made this point when they testified about predatory lending in 2000, and the Office of the Comptroller of the Currency cautioned against treating subprime and predatory as the same thing.
The measurement exists because regulators built it alongside the rule. When the Federal Reserve proposed amending Regulation Z in 2000, it proposed amending Regulation C in the same cycle, so that institutions and examiners could track the level, trend, and underwriting characteristics of high-cost lending they otherwise could not observe.
HOEPA status can be analyzed with HMDA. It is a flag the lender populates, because the lender ran the three tests at closing in order to comply. Higher-priced status has no equivalent field. It has to be constructed from rate spread, lien status, and whether the loan exceeds the conforming limit for that year, county, and unit count, then filtered to closed-end credit secured by a principal dwelling. Consuming a determination and reproducing one are different exercises, and conflating them is a common source of mismatched counts.
The HOEPA definition changed materially in January 2014, so counts before and after are not measuring the same population: the benchmark moved from Treasury yields to APOR, the fee trigger fell from 8 percent to 5, the prepayment trigger was added, and purchase loans and HELOCs came inside. Separately, rate spread was reported before the 2018 data year only when it exceeded a reporting threshold, so that field is censored at the bottom. Any higher-priced series crossing 2018 carries a break for that reason alone.
Since the 2018 data year, HMDA also reports total loan costs, points and fees, origination charges, and rate spread. Together these let an analyst determine which of the three triggers a given loan crossed rather than inferring it from an average.
HMDA does not capture all high-cost lending. Some loans come from institutions outside HMDA reporting, and some loans made by HMDA reporters are not reportable under Regulation C.
Within the consistently filtered series used here, 2025 had the highest number of high-cost mortgage originations: 9,756. The previous high was 8,400 in 2023. The broader, unfiltered HMDA data contain 10,335 originations reported with HOEPA status in 2025.
2025 HMDA data
Each tile shows 2025 originations. Hover or focus for the state name and share of the national total.
Source: Polygon Research analysis of 2025 HMDA data. Filters: originations; reverse mortgage: no; HOEPA status: HOEPA; loan terms of 12 months or less excluded; 1–4-unit properties; USDA loans excluded. D.C. and Puerto Rico are shown as inset tiles. 1 origination without a state assignment is included in the national total but is not shown on the map.
The activity was national, but unevenly distributed. Every state, D.C., and Puerto Rico reported at least one filtered HOEPA origination, while California, Florida, Ohio, Georgia, and Indiana together accounted for 39.1 percent of the total. California alone had 1,375 loans, or 14.1 percent. These counts show where activity occurred; comparing them with each state’s total mortgage originations would show where high-cost lending was most prevalent.
The 2025 count was 127.9 percent higher than 2024 and 16.1 percent above the previous high in 2023. Dollar volume rose 75.3 percent in 2025, to $1.34 billion, but remained below the $2.32 billion recorded in 2023. That combination—more originations but less total volume than in 2023—means the 2025 loans were smaller on average, about $137,000 based on the annual totals.
2018–2025 HMDA data
Annual loan count and dollar volume, with 2025 highlighted.
Source: Polygon Research analysis of HMDA data. Filters: originations; reverse mortgage: no; HOEPA status: HOEPA; loan terms of 12 months or less excluded; 1–4-unit properties; USDA loans excluded. Dollar volume is shown in billions.
Determining the binding trigger for a loan to be flagged as HOEPA takes the loan-level fields, and at that level a single lender is often more informative than a national mean.
About the analysis
This analysis uses Home Mortgage Disclosure Act (HMDA) data accessed through HMDAVision on August 20, 2026. The annual and state series include originations reported with HOEPA status for 1–4-unit properties. Reverse mortgages, loans reported as USDA, and loans with terms of 12 months or less are excluded. No loan-purpose filter is applied. These filters produce 9,756 originations in 2025; the broader, unfiltered HMDA population contains 10,335.
The resulting series has a different scope from the CFPB’s annual HMDA summaries, which present HOEPA originations for 1–4-family home purchase, home improvement, and refinance loans. For 2023, CFPB reported 7,710 such originations, compared with 8,400 in this series. Data timing can create additional differences: CFPB summaries use a static Snapshot National Loan-Level Dataset, while HMDAVision continues to incorporate late filings, lender resubmissions, and corrections. Comparisons in this article use one consistent filter set across 2018–2025.
A high-cost mortgage flag does not automatically mean a loan is predatory. Predatory lending describes harmful conduct—like equity stripping or packing unwanted insurance. The HOEPA flag simply measures price.
Congress and regulators built this threshold as a mechanical tripwire. It ensures that when a loan’s cost crosses a specific line, the borrower receives mandatory counseling and enhanced disclosures, and regulators can track exactly who is getting these loans and on what terms. A loan priced fairly for a unique risk profile can easily land above a HOEPA trigger, while a truly predatory loan can sometimes price comfortably below all of them.
To see exactly how this tripwire functions mechanically, examining a single lender is often more instructive than looking at national averages.
Consider a small independent lender that originated 696 loans in 2025. Based on their production data, exactly seven of those loans carried the HOEPA flag, while 532 were standard non-HOEPA originations and 157 fell into an exempt (NA) category.
The 532 non-HOEPA loans represented typical production: an average note rate of 6.386 percent, an average loan size of $561,297, a 71.44 percent CLTV, and a minimal rate spread of 0.03.
The seven HOEPA loans look like a different business:
The rate spread is the factor for the HOEPA flag in this case. At 9.79 points over APOR these loans clear the 6.5-point first-lien trigger with room to spare, so the rate test fired.
Given the loan term of six months, the fee load amortized over six months moves APR far more than the same load over thirty years.
The reason the rate test fired comes down to the loan term, which highlights exactly why regulators monitor this data at the loan level.
When a lender amortizes a heavy fee load over a standard thirty-year loan, the impact on the APR is diluted. But when that same $28,000+ fee load is compressed into a 6-month term—likely a bridge loan or similar short-term product—the APR spikes dramatically. In this scenario, the high cost is a mathematical reality of the specific product structure, not necessarily a predatory attack on a vulnerable homeowner.
This outcome yields two critical lessons for the industry:
National baselines are useful, but they do not explain why an individual loan was flagged. An average-sized loan in the 2025 series—about $137,000—and a $1.18 million six-month loan can reach HOEPA status through very different routes. The useful question is how much high-cost lending occurred and which trigger did the work.
Origination costs do not scale down with loan size. Underwriting, title, appraisal, and processing cost roughly the same on a $60,000 second lien as on a $600,000 first. As a percentage of the loan, they grow as the balance falls.
So the 5 percent threshold is not a constant difficulty. A $137,000 loan crosses at $6,850 in points and fees. A $50,000 loan crosses at $2,500, which fixed costs can approach on their own.
The $50,000 personal-property figure compounds this. A first lien on a dwelling titled as personal property gets the looser 8.5-point rate test only below that amount, and that figure has stayed nominal since 2013 while the two indexed figures move with CPI each January. Manufactured home prices have not stayed flat over the same period.
The qualified mortgage rule caps points and fees at 3 percent, below HOEPA's 5 percent. Mainstream production is managed to the stricter of the two, so the high-cost thresholds rarely bind. The Consumer Financial Protection Bureau anticipated this when it wrote the 2013 rule, noting that from 2004 through 2009 roughly 1,000 to 2,000 creditors reported HOEPA loans and that 80 to 90 percent of them originated fewer than ten a year.
Why the rule measures price at all is a question with a specific answer, and we covered it in Common Ground No. 5, What makes a mortgage predatory? The practices Congress heard about in 1993 had names and proved hard to legislate, so it set two numbers instead. At its 2005 peak the law reached about 36,000 loans while subprime lending ran at roughly a fifth of all originations.
National counts set a baseline. The questions that matter are local and specific: which lenders in your footprint are originating high-cost loans, in which counties, at what balances, on what terms, and which trigger is doing the work.
Start a complimentary trial to run the HOEPA status flag against your markets, your peer group, and your portfolio.
Methodology. Annual counts and dollar volume use the filtered series described in the Data and methodology note above. The single-lender example uses one HMDA reporter's 2025 originations, presented without identifying the institution; all figures are lender-level aggregates. The implied loan term is derived from the reported monthly payment, loan amount, and note rate. Threshold figures come from 12 CFR 1026.32 and 1026.35 and the CFPB's annual threshold adjustments for 2026. Historical figures on HOEPA coverage come from Federal Reserve History, "Home Ownership and Equity Protection Act of 1994," and from Federal Reserve Bank of Atlanta, Partners in Community and Economic Development, 2000 and 2002, via FRASER. The HOEPA definition changed materially in January 2014, so counts before and after that date describe different populations. HMDA does not capture all high-cost lending, because some loans come from institutions outside HMDA reporting and some loans made by HMDA reporters are not reportable under Regulation C.
A high-cost mortgage crosses one of three HOEPA price triggers. The thresholds, the exclusions, how it differs from a higher-priced loan, and how to measure high-cost lending in HMDA data.
A high-cost mortgage is a loan secured by a principal dwelling that crosses one of three price triggers set in Regulation Z. The triggers cover the interest rate, the points and fees, and the prepayment penalty. Cross any one of them and the loan becomes a high-cost mortgage, also called a HOEPA loan or a Section 32 loan, which brings additional disclosures and a list of loan terms the lender can no longer use.
The tests measure price. They say nothing about how the loan was sold.
A mortgage is high cost if it meets any one of these tests, found at 12 CFR 1026.32.
1. The rate trigger. The annual percentage rate exceeds the average prime offer rate (APOR) for a comparable transaction by more than:
2. The points and fees trigger. Total points and fees exceed:
3. The prepayment penalty trigger. The loan permits a prepayment penalty more than 36 months after consummation, or a penalty that can exceed 2 percent of the amount prepaid.
The two dollar figures are adjusted every January for inflation. The percentages and the APR spreads have held since January 2014. The $50,000 personal-property figure is nominal and does not index, which matters for manufactured housing and comes up again below.
Note that the rate test uses APR, not the note rate. APR carries the fees, so the first and second triggers are connected. A loan can clear the rate trigger because of its fee load, and the shorter the term the more pronounced that effect becomes.
Since the Dodd-Frank amendments took effect in January 2014, coverage includes purchase-money mortgages, refinances, closed-end home equity loans, and home equity lines of credit secured by a principal dwelling.
Excluded: reverse mortgages, construction-only loans, USDA Section 502 Direct loans, and loans where a housing finance agency is the creditor.
That is broader than the original 1994 statute, which reached only refinances and home improvement loans. Purchase mortgages sat outside HOEPA for its first twenty years.
Crossing a trigger carries real operational weight, which is the main reason volumes stay low.
Before closing, the borrower must receive homeownership counseling from a HUD-approved counselor, and the creditor must obtain written certification that it happened. The loan cannot carry a balloon payment, with narrow exceptions. It cannot carry a prepayment penalty. Points and fees cannot be financed into the loan. Late fees are capped at 4 percent of the past-due payment. Fees for modifications, deferrals, and payoff statements are restricted. The creditor cannot recommend default on an existing loan being refinanced.
For most lenders the practical effect is a line to stay behind rather than a disclosure to prepare.
These four terms get used interchangeably and mean four different things. Sorting them out is most of the analytical work.
A loan priced fairly for risk can land above a trigger. A loan sold to a borrower who did not understand its terms can price comfortably below every one of them. Several federal regulators made this point when they testified about predatory lending in 2000, and the Office of the Comptroller of the Currency cautioned against treating subprime and predatory as the same thing.
The measurement exists because regulators built it alongside the rule. When the Federal Reserve proposed amending Regulation Z in 2000, it proposed amending Regulation C in the same cycle, so that institutions and examiners could track the level, trend, and underwriting characteristics of high-cost lending they otherwise could not observe.
HOEPA status can be analyzed with HMDA. It is a flag the lender populates, because the lender ran the three tests at closing in order to comply. Higher-priced status has no equivalent field. It has to be constructed from rate spread, lien status, and whether the loan exceeds the conforming limit for that year, county, and unit count, then filtered to closed-end credit secured by a principal dwelling. Consuming a determination and reproducing one are different exercises, and conflating them is a common source of mismatched counts.
The HOEPA definition changed materially in January 2014, so counts before and after are not measuring the same population: the benchmark moved from Treasury yields to APOR, the fee trigger fell from 8 percent to 5, the prepayment trigger was added, and purchase loans and HELOCs came inside. Separately, rate spread was reported before the 2018 data year only when it exceeded a reporting threshold, so that field is censored at the bottom. Any higher-priced series crossing 2018 carries a break for that reason alone.
Since the 2018 data year, HMDA also reports total loan costs, points and fees, origination charges, and rate spread. Together these let an analyst determine which of the three triggers a given loan crossed rather than inferring it from an average.
HMDA does not capture all high-cost lending. Some loans come from institutions outside HMDA reporting, and some loans made by HMDA reporters are not reportable under Regulation C.
Within the consistently filtered series used here, 2025 had the highest number of high-cost mortgage originations: 9,756. The previous high was 8,400 in 2023. The broader, unfiltered HMDA data contain 10,335 originations reported with HOEPA status in 2025.
2025 HMDA data
Each tile shows 2025 originations. Hover or focus for the state name and share of the national total.
Source: Polygon Research analysis of 2025 HMDA data. Filters: originations; reverse mortgage: no; HOEPA status: HOEPA; loan terms of 12 months or less excluded; 1–4-unit properties; USDA loans excluded. D.C. and Puerto Rico are shown as inset tiles. 1 origination without a state assignment is included in the national total but is not shown on the map.
The activity was national, but unevenly distributed. Every state, D.C., and Puerto Rico reported at least one filtered HOEPA origination, while California, Florida, Ohio, Georgia, and Indiana together accounted for 39.1 percent of the total. California alone had 1,375 loans, or 14.1 percent. These counts show where activity occurred; comparing them with each state’s total mortgage originations would show where high-cost lending was most prevalent.
The 2025 count was 127.9 percent higher than 2024 and 16.1 percent above the previous high in 2023. Dollar volume rose 75.3 percent in 2025, to $1.34 billion, but remained below the $2.32 billion recorded in 2023. That combination—more originations but less total volume than in 2023—means the 2025 loans were smaller on average, about $137,000 based on the annual totals.
2018–2025 HMDA data
Annual loan count and dollar volume, with 2025 highlighted.
Source: Polygon Research analysis of HMDA data. Filters: originations; reverse mortgage: no; HOEPA status: HOEPA; loan terms of 12 months or less excluded; 1–4-unit properties; USDA loans excluded. Dollar volume is shown in billions.
Determining the binding trigger for a loan to be flagged as HOEPA takes the loan-level fields, and at that level a single lender is often more informative than a national mean.
About the analysis
This analysis uses Home Mortgage Disclosure Act (HMDA) data accessed through HMDAVision on August 20, 2026. The annual and state series include originations reported with HOEPA status for 1–4-unit properties. Reverse mortgages, loans reported as USDA, and loans with terms of 12 months or less are excluded. No loan-purpose filter is applied. These filters produce 9,756 originations in 2025; the broader, unfiltered HMDA population contains 10,335.
The resulting series has a different scope from the CFPB’s annual HMDA summaries, which present HOEPA originations for 1–4-family home purchase, home improvement, and refinance loans. For 2023, CFPB reported 7,710 such originations, compared with 8,400 in this series. Data timing can create additional differences: CFPB summaries use a static Snapshot National Loan-Level Dataset, while HMDAVision continues to incorporate late filings, lender resubmissions, and corrections. Comparisons in this article use one consistent filter set across 2018–2025.
A high-cost mortgage flag does not automatically mean a loan is predatory. Predatory lending describes harmful conduct—like equity stripping or packing unwanted insurance. The HOEPA flag simply measures price.
Congress and regulators built this threshold as a mechanical tripwire. It ensures that when a loan’s cost crosses a specific line, the borrower receives mandatory counseling and enhanced disclosures, and regulators can track exactly who is getting these loans and on what terms. A loan priced fairly for a unique risk profile can easily land above a HOEPA trigger, while a truly predatory loan can sometimes price comfortably below all of them.
To see exactly how this tripwire functions mechanically, examining a single lender is often more instructive than looking at national averages.
Consider a small independent lender that originated 696 loans in 2025. Based on their production data, exactly seven of those loans carried the HOEPA flag, while 532 were standard non-HOEPA originations and 157 fell into an exempt (NA) category.
The 532 non-HOEPA loans represented typical production: an average note rate of 6.386 percent, an average loan size of $561,297, a 71.44 percent CLTV, and a minimal rate spread of 0.03.
The seven HOEPA loans look like a different business:
The rate spread is the factor for the HOEPA flag in this case. At 9.79 points over APOR these loans clear the 6.5-point first-lien trigger with room to spare, so the rate test fired.
Given the loan term of six months, the fee load amortized over six months moves APR far more than the same load over thirty years.
The reason the rate test fired comes down to the loan term, which highlights exactly why regulators monitor this data at the loan level.
When a lender amortizes a heavy fee load over a standard thirty-year loan, the impact on the APR is diluted. But when that same $28,000+ fee load is compressed into a 6-month term—likely a bridge loan or similar short-term product—the APR spikes dramatically. In this scenario, the high cost is a mathematical reality of the specific product structure, not necessarily a predatory attack on a vulnerable homeowner.
This outcome yields two critical lessons for the industry:
National baselines are useful, but they do not explain why an individual loan was flagged. An average-sized loan in the 2025 series—about $137,000—and a $1.18 million six-month loan can reach HOEPA status through very different routes. The useful question is how much high-cost lending occurred and which trigger did the work.
Origination costs do not scale down with loan size. Underwriting, title, appraisal, and processing cost roughly the same on a $60,000 second lien as on a $600,000 first. As a percentage of the loan, they grow as the balance falls.
So the 5 percent threshold is not a constant difficulty. A $137,000 loan crosses at $6,850 in points and fees. A $50,000 loan crosses at $2,500, which fixed costs can approach on their own.
The $50,000 personal-property figure compounds this. A first lien on a dwelling titled as personal property gets the looser 8.5-point rate test only below that amount, and that figure has stayed nominal since 2013 while the two indexed figures move with CPI each January. Manufactured home prices have not stayed flat over the same period.
The qualified mortgage rule caps points and fees at 3 percent, below HOEPA's 5 percent. Mainstream production is managed to the stricter of the two, so the high-cost thresholds rarely bind. The Consumer Financial Protection Bureau anticipated this when it wrote the 2013 rule, noting that from 2004 through 2009 roughly 1,000 to 2,000 creditors reported HOEPA loans and that 80 to 90 percent of them originated fewer than ten a year.
Why the rule measures price at all is a question with a specific answer, and we covered it in Common Ground No. 5, What makes a mortgage predatory? The practices Congress heard about in 1993 had names and proved hard to legislate, so it set two numbers instead. At its 2005 peak the law reached about 36,000 loans while subprime lending ran at roughly a fifth of all originations.
National counts set a baseline. The questions that matter are local and specific: which lenders in your footprint are originating high-cost loans, in which counties, at what balances, on what terms, and which trigger is doing the work.
Start a complimentary trial to run the HOEPA status flag against your markets, your peer group, and your portfolio.
Methodology. Annual counts and dollar volume use the filtered series described in the Data and methodology note above. The single-lender example uses one HMDA reporter's 2025 originations, presented without identifying the institution; all figures are lender-level aggregates. The implied loan term is derived from the reported monthly payment, loan amount, and note rate. Threshold figures come from 12 CFR 1026.32 and 1026.35 and the CFPB's annual threshold adjustments for 2026. Historical figures on HOEPA coverage come from Federal Reserve History, "Home Ownership and Equity Protection Act of 1994," and from Federal Reserve Bank of Atlanta, Partners in Community and Economic Development, 2000 and 2002, via FRASER. The HOEPA definition changed materially in January 2014, so counts before and after that date describe different populations. HMDA does not capture all high-cost lending, because some loans come from institutions outside HMDA reporting and some loans made by HMDA reporters are not reportable under Regulation C.
A high-cost mortgage is a loan secured by a principal dwelling that crosses one of three price triggers in Regulation Z: the APR test, the points and fees test, or the prepayment penalty test. It is also called a HOEPA loan or a Section 32 loan, after 12 CFR 1026.32.
They are separate categories. A higher-priced mortgage loan is triggered at 1.5 percentage points over APOR for most first liens and carries escrow and appraisal requirements. A high-cost mortgage is triggered at 6.5 points for most first liens and carries counseling requirements and prohibited loan terms. They overlap heavily without either containing the other, since higher-priced reaches closed-end credit only and a loan can be high cost on the fee or prepayment trigger alone.
No. High cost is a price threshold in Regulation Z. Subprime describes borrower credit quality and has no regulatory definition. Predatory describes conduct such as repeated refinancing that strips equity. A loan priced fairly for genuine risk can be high cost, and a loan sold poorly can price below every trigger.
HOEPA status is a flag the lender reports in HMDA, available since the 2004 data year. Higher-priced status has no equivalent field and must be modeled from rate spread, lien status, and the conforming limit. Since the 2018 data year HMDA also reports total loan costs, points and fees, and origination charges, which allow an analyst to identify which trigger a loan crossed. Counts before and after January 2014 are not comparable, because the HOEPA definition changed materially at that point.