This article is educational and not mortgage, legal, or tax advice; unless labeled otherwise inline, cost figures reflect 2023 Home Mortgage Disclosure Act data, the most recent complete federal dataset on borrower-paid mortgage fees available at publication.
TL;DR — Quick Verdict
- Federal Reserve Bank of Philadelphia analysis of 2023 HMDA data found borrowers’ out-of-pocket upfront costs on a home purchase mortgage averaged nearly $6,500, up from about $4,900 in 2021 — a 33% increase in two years.
- Only part of Section A is genuinely negotiable. Discount points are a purchase decision, not a fee; lender administrative fees averaged over $1,000 per loan and are the real negotiation target.
- Origination charges in Section A carry a zero percent tolerance under 12 CFR 1026.19(e)(3)(i) — the lender cannot raise them at closing absent a documented changed circumstance.
- Comparison result: on an identical $300,000 loan in 2023, the average nonbank charged roughly $2,000 more than the average bank and nearly $4,000 more than the average credit union, once points and rate were combined.
- Recommendation: collect three Loan Estimates in the same week, compare Section A totals rather than line-item labels, and negotiate against a written competing offer — not against a verbal quote.
Upfront mortgage costs climbed 33% in two years while almost nobody was watching the rate. Researchers at the Federal Reserve Bank of Philadelphia, analyzing confidential Home Mortgage Disclosure Act records covering more than 1.5 million 2023 home purchase originations, found borrowers paid nearly $6,500 in out-of-pocket points and fees, compared with about $4,900 in 2021. Section A of the Loan Estimate — “Origination Charges” — is where the lender’s share of that money is disclosed, and it is the only closing-cost section where the borrower is negotiating directly with the party who set the price.
The problem is that Section A is not one fee. It bundles discount points, an origination fee, an underwriting fee, a processing fee, an application fee, and sometimes a rate-lock fee into a single dollar total, and lenders itemize that total inconsistently. Rocket Mortgage publishes an origination fee range of 0.5% to 1.2% of the loan amount; Better Mortgage markets no lender origination fee at all. Those two structures are not comparable without decomposing Section A first. This article breaks Section A into its component parts, identifies which components move under pressure and which do not, quantifies the negotiation window using federal data, and walks through the tolerance rules that determine what a lender can legally change after you receive the estimate.
What Section A Actually Contains
Section A of the Loan Estimate and Closing Disclosure holds every charge the lender collects for making the loan. The Consumer Financial Protection Bureau’s own sample Closing Disclosure itemizes it as four possible lines: discount points expressed as a percentage of the loan amount, an application fee, an underwriting fee, and any additional lender charges. In the CFPB’s illustrative form, those lines total $1,802 — $405 in points, a $300 application fee, and a $1,097 underwriting fee.
Not every lender uses that breakdown. Some collapse everything into one “origination fee” of 0.8%; others split the identical amount across a 0.3% processing fee, a 0.3% underwriting fee, and a 0.2% administrative fee. The dollar total is what matters, and comparing labels across lenders produces false conclusions. The CFPB has stated that origination charges may be more or less itemized depending on the lender, which is precisely why the agency standardized the section total.
One labeling rule is legally binding. Under 12 CFR 1026.37(f)(1)(i), only amounts paid to the creditor to reduce the interest rate may be disclosed as points. A lender charging a flat percentage of the loan amount that buys no rate reduction cannot relabel it as points, a rule the CFPB confirmed when it declined to allow the practice for tax-deductibility purposes. That matters for anyone reading a loan estimate’s key numbers and trying to distinguish a purchase decision from an administrative charge.
Component definitions per Consumer Financial Protection Bureau sample Closing Disclosure and Regulation Z §1026.37(f). CFPB Loan Estimate Explainer. Negotiability column reflects this publication’s assessment, not agency guidance.
How Much of the $6,500 Is Actually Lender Fees
Decomposing the average matters more than the average itself. The Philadelphia Fed researchers separated borrower upfront costs into three streams and tracked each from 2018 through 2023. Third-party costs — appraisal, lender’s title insurance, settlement agent, credit report, flood determination — held steady at nearly $3,000 and were largely unchanged through 2022 and 2023. Net points and fees paid to lenders more than doubled, from an average just under $1,500 during 2021 to nearly $3,500 during 2023.
Within that lender total, discount points did the work. Net points — points net of lender credits — averaged 0.67% of the loan amount across 2023 originations, against 0.04% in 2021. The residual, which the researchers describe as all other lender fees collected at or before closing, typically averaged over $1,000 across the period studied. That $1,000-plus figure is the honest size of the negotiation target on a typical purchase loan, and it barely moved while points exploded.
Run the math on a $400,000 loan using the 2023 averages. Points at 0.67% cost $2,680. Other lender fees add roughly $1,000. Third-party costs contribute about $3,000, which flow through Sections B and C rather than Section A and follow entirely different rules covered in our breakdown of which closing costs are negotiable. Total upfront: approximately $6,680. Of that, about $1,000 — 15% — is the administrative charge a borrower can argue about. The other 85% is either a rate purchase or a third-party service price.
Bhutta, N. and Lambie-Hanson, L., “The Rise in Mortgage Fees: Evidence from HMDA Data,” Discussion Paper 24-01, November 2024. Federal Reserve Bank of Philadelphia Consumer Finance Institute. Figures cover first-lien conventional and FHA 30-year fixed home purchase originations.
Why Lenders Resist Cutting the Fee — and Where the Give Is
Origination fees are not pure margin. Freddie Mac’s Single-Family analysis of retail lender financials put the average cost to produce a mortgage at approximately $11,800 in the second quarter of 2025, down from roughly $13,400 in the first quarter but still above the $11,600 recorded in the third quarter of 2023. Against that cost base, lenders reported pre-tax net income of $900 per loan. A loan officer told to cut $800 from Section A is being told to erase most of the file’s profit.
That constraint explains observed lender behavior. Personnel expenses represent roughly two-thirds of total production costs according to Freddie Mac’s 2024 Cost to Originate Study, and those costs are largely fixed per file regardless of loan size. A $200,000 loan and an $800,000 loan require similar document collection, similar verification work, and a similar amount of underwriter attention — which is why percentage-based origination fees generate far more revenue on large loans than the underlying work justifies.
The negotiation window opens exactly there. On a $750,000 loan, a 1% origination fee produces $7,500 against a production cost the lender’s own trade data pegs near $11,800 all-in — but that $11,800 includes commissions and overhead the lender recovers through secondary-market sale and servicing value, not the fee alone. Borrowers with large loan amounts, strong credit profiles, and low loan-to-value ratios represent files that close fast and rarely fall out during underwriting review. Those borrowers have leverage. A borrower with a thin file, a self-employment income structure, and a 3% down payment does not, because the lender’s expected cost on that file is genuinely higher.
Bank vs Nonbank Lender: Which Is Better for Minimizing Origination Charges?
Lender type predicts pricing more reliably than any advertised rate. The Philadelphia Fed study regressed 2023 pricing outcomes on lender type while controlling for application date, county, loan type, loan size, credit score, and combined loan-to-value ratio — meaning the comparison holds borrower risk constant. Nonbank mortgage lenders charged interest rates about 11 basis points higher than banks and about 0.211 points more. Credit unions charged nearly 0.238 points less than banks and rates about 9 basis points lower.
Translate that into cash. Assuming a 25 basis point rate buydown per point, the researchers calculated that a borrower taking a $300,000 mortgage from the average nonbank paid roughly $2,000 more than the same borrower at the average bank, and almost $4,000 more than at the average credit union. Their itemized derivation: $630 in additional points, plus $1,320 to equalize the 11 basis point rate gap.
Nonbanks are not uniformly expensive. Breaking the category apart, the study found FinTech lenders priced similarly to or slightly better than banks, with rates 9 to 10 basis points lower offset by 0.211 to 0.256 more points. Builder-affiliated lenders showed rates 70 basis points below bank rates in 2023, though the authors caution that builder financing is bundled with the home sale and offsetting costs may not be visible in the data. Roughly 53% of the 2023 sample came from nonbanks, 27% from banks, 11% from FinTechs, 5% from credit unions, and 4% from builder lenders.
Verdict
For a borrower whose sole objective is minimizing origination charges and rate, credit unions were the cheapest lender category in 2023 by a wide margin — nearly $4,000 on a $300,000 loan versus the average nonbank. Nonbanks earn their place through breadth of loan programs, speed, and willingness to underwrite non-standard files, not price. The practical rule: if you qualify for credit union membership and your file is straightforward, get an estimate there before anywhere else. If your income documentation is complex or you need a program a depository won’t write, expect to pay for that access and negotiate the administrative fees rather than the points.
The Tolerance Rules That Govern What Can Change
Regulation Z assigns every closing cost to a tolerance category, and Section A sits in the strictest one. Under 12 CFR 1026.19(e)(3)(i), origination charges carry a zero percent tolerance: the amount the borrower pays cannot exceed the amount disclosed on the Loan Estimate. The CFPB states the rule plainly — origination fees generally cannot increase at closing except under certain circumstances.
Third-party services the borrower is permitted to shop for fall under §1026.19(e)(3)(ii), where the aggregate may rise by up to 10% over the disclosed total. That distinction is the entire reason Section A is worth negotiating hard: a number you extract there is locked, while a number you extract in Section C can drift upward by a tenth before any cure is owed. The same logic governs the treatment of owner’s versus lender’s title insurance, which sits outside the zero-tolerance bucket entirely.
Changed circumstances reset the baseline. A lender that documents a valid changed circumstance — a loan amount increase, a program change, information that was unavailable and not reasonably knowable at disclosure — may issue a revised Loan Estimate reflecting higher charges. Absent that documentation, an increase in Section A at closing requires a tolerance cure: a corrected Closing Disclosure, a refund to the borrower, and a letter of explanation. Loan file audits routinely flag these, and borrowers who compare their Loan Estimate against their closing disclosure line items line by line are the ones who catch them.
One asymmetry favors lenders. The CFPB clarified that if a creditor decreases an estimated charge on a revised disclosure, it is not required to use the decreased estimate when determining good faith — it may compare against the amount originally disclosed. A mid-process reduction, in other words, does not permanently lower the ceiling.
What Most People Get Wrong About Negotiating Section A
Four errors show up repeatedly, and each has a measurable cost.
Mistake 1: Treating “no origination fee” as free
Lenders advertising zero lender fees typically recover the cost through a higher interest rate. Navy Federal Credit Union, per Forbes Advisor’s 2026 lender review, waives its typical 1% origination fee in exchange for adding 0.25% to the mortgage rate — verify current terms at navyfederal.org. On a $400,000 loan, that trades $4,000 upfront for a rate premium lasting the life of the loan. Correct action: compute the breakeven month before accepting the waiver, and compare annual percentage rate rather than the headline fee.
Mistake 2: Negotiating against a verbal quote
The CFPB recommends obtaining Loan Estimates from at least three lenders. A verbal quote carries no tolerance protection and cannot be used as leverage, because the competing lender has no obligation behind it. Correct action: request three written Loan Estimates within the same week so rate movement doesn’t contaminate the comparison, then present the lowest Section A total as a document.
Mistake 3: Confusing lender credits with a fee reduction
Negative points appear as lender credits and reduce cash at closing, but the Philadelphia Fed data show something borrowers rarely see: for loans with net points between zero and negative one, the relationship between credits and total borrower-paid points and fees flattens, suggesting credits are being offset by relatively high other lender fees. Correct action: check whether the Section A subtotal actually fell, not just whether a credit line appeared in Section J.
Mistake 4: Assuming points bought a rate discount
Comparing across lenders, higher points correlate with higher rates, not lower — lenders charging more points also tend to charge more interest. Within a single lender, controlling for borrower risk and application date, each point purchased bought an average rate decline of nearly 23 basis points in 2023. Correct action: evaluate the points-rate trade-off only within a single lender’s quote sheet, never by comparing one lender’s pointed offer against another’s par offer.
Mistake 5: Waiting until the Closing Disclosure to object
Section A is negotiable during the shopping window and effectively frozen afterward. Once the Closing Disclosure is issued, the zero tolerance rule protects you from increases but creates no mechanism to demand a decrease, and reopening the file risks the closing timeline entirely. Correct action: conclude all Section A negotiation before rate lock.
Who Should Push, and Who Should Not Bother
Negotiating leverage tracks file quality and loan size, not persistence. A borrower with a 760-plus credit score, 20% down, W-2 income, and a loan above $400,000 represents a file the lender expects to close on schedule with minimal rework. That borrower can reasonably ask a lender to match a competitor’s Section A total, waive an application fee, or absorb a rate-lock extension. Freddie Mac’s finding that lenders using automated underwriting capabilities extensively experience 40% fewer loan defects underlines why clean files are worth discounting.
Conversely, a borrower with a 640 score, 3.5% down, self-employment income requiring two years of returns and a profit-and-loss statement, and a $180,000 loan amount is a file where the lender’s fixed production cost consumes most of the fee. Pushing hard there tends to produce a rate increase rather than a fee reduction, and in tight files it can surface the underwriting fragilities that drive mortgage denial and reapplication costs. Shopping across lender categories delivers more savings than negotiating within one.
Two situations change the calculus entirely. When a seller is contributing toward costs, the negotiation shifts from the lender to the purchase contract, subject to seller concession limits toward closing costs — and the Philadelphia Fed data confirm this happens at scale, showing that borrowers on average did not directly pay discount points above two points, with those marginal points covered by sellers or other parties. When a borrower expects to sell or refinance within five years, paying points to lower the rate rarely recovers, and the correct move is to accept lender credits and a higher rate. Refinancing borrowers should also account for how a new escrow setup interacts with prepaid insurance, tax, and interest at closing, since those amounts dwarf a $500 fee concession.
Frequently Asked Questions
Can a lender increase my origination fee between the Loan Estimate and closing?
Generally no. Origination charges in Section A carry a zero percent tolerance under 12 CFR 1026.19(e)(3)(i), meaning the amount you pay cannot exceed the amount originally disclosed. A lender may only increase it after documenting a valid changed circumstance and issuing a revised Loan Estimate. Without that documentation, the lender owes a tolerance cure — a corrected Closing Disclosure, a refund, and a written explanation.
How much can I realistically expect to save by negotiating?
The Philadelphia Fed’s 2023 HMDA analysis indicates lender fees other than points typically averaged over $1,000 per home purchase loan. That is the negotiable pool on an average file. Larger savings come from switching lender categories rather than haggling: the same borrower on a $300,000 loan paid roughly $2,000 more at the average nonbank than the average bank in 2023.
Are discount points worth paying in a high-rate environment?
Only with a long enough holding period. Within a single lender, each point bought an average rate decline of nearly 23 basis points in 2023 according to Federal Reserve Bank of Philadelphia estimates. On a $400,000 loan, one point costs $4,000 for roughly a 0.23% rate reduction. Compute the monthly payment difference and divide into the upfront cost to find your breakeven month before committing.
Do lender credits actually reduce what I pay?
Sometimes less than they appear to. Philadelphia Fed researchers found that for loans with net points between zero and negative one, lender credits tended to be offset by relatively high other lender fees, meaning those borrowers were not receiving a real net credit. Verify by comparing the Section A subtotal across competing Loan Estimates rather than reading the credit line in isolation.
How We Researched This Article
Cost figures in this article come from three primary sources, each named at the point of use. Our principal dataset is the Federal Reserve Bank of Philadelphia Consumer Finance Institute Discussion Paper 24-01, “The Rise in Mortgage Fees: Evidence from HMDA Data,” published November 2024 by Neil Bhutta and Lauren Lambie-Hanson. That paper analyzes confidential Home Mortgage Disclosure Act records for first-lien conventional and FHA 30-year fixed-rate home purchase originations on single-family, owner-occupied, site-built properties, restricted to loans receiving an automated approval recommendation and excluding loans below $25,000 or above $2 million. The 2023 regression sample contains 1,524,186 loans; the 2021 sample contains 2,329,793. Regression estimates control for application date, county, and fully interacted bins of loan type, loan size, credit score, and combined loan-to-value ratio, with standard errors clustered at the lender level. Data are available through the Philadelphia Fed Consumer Finance Institute.
Regulatory treatment of Section A charges was verified against Regulation Z as codified at 12 CFR 1026.19(e)(3) and 1026.37(f), and against published Consumer Financial Protection Bureau guidance on origination services and the agency’s Loan Estimate explainer. Lender production cost figures come from Freddie Mac Single-Family’s 2025 Updates to the Cost to Originate Study and its 2024 Cost to Originate Study, both published at Freddie Mac Single-Family Insights.
Two limitations deserve explicit statement. First, HMDA origination charge and total loan cost fields capture only borrower-paid amounts, so they understate true closing costs where sellers, builders, or other parties pay a portion — the paper’s authors flag this directly. Second, HMDA lender credit reporting captures only general credits listed in Box J of the Closing Disclosure, not item-specific credits, which means net points may be overstated in some periods. Where this article presents 2023 figures, those are measured values from the HMDA sample, not projections. Where this article presents a $400,000 or $750,000 loan scenario, those are modeled calculations applying measured 2023 averages to a hypothetical loan amount and are labeled as such in the text. Vendor-specific fee terms are described as ranges with a direction to verify current pricing at the lender’s own site, because vendor pricing is not verifiable through a primary regulatory source and changes without notice. Research was last conducted July 2026. All figures were verified against named primary sources before publication.