Pricing Power in the Mortgage Market (Bhutta, Keys & Raajkumar, 2026)

Summary

Neil Bhutta (Federal Reserve Bank of Philadelphia), Benjamin J. Keys and Adithya Raajkumar (both University of Pennsylvania, The Wharton School; Keys also NBER), "Pricing Power in the Mortgage Market: Evidence from GSE Fee Waivers," NBER Working Paper 35907 (October 2026), ask who receives a subsidy delivered through intermediaries: a waiver of upfront fees that Fannie Mae and Freddie Mac (the GSEs) charge on loans they buy, directed by the Federal Housing Finance Agency (FHFA) for lower-income first-time homebuyers.

The authors estimate that lenders pass through less than 40 percent of the waiver on average (baseline 39.3 percent; 31.5 to 39.3 percent across specifications), and that intermediaries capture over 60 percent of $1.5 billion in waivers over 2023-2024. The most expensive quintile of loan officers passes through less than 20 percent. The authors frame the result as market power that "may arise not only from firms, but also from individual salespeople who can exploit informational advantages over consumers" (abstract).

The Fee Waiver and the Data

The program. The GSEs charge lenders loan-level pricing adjustments (LLPAs), upfront fees that vary with credit score and loan-to-value ratio and are generally unseen by borrowers. On 24 October 2022 the FHFA announced that the GSEs would eliminate LLPAs for first-time buyers with income below the area median income (AMI), or below 120 percent of AMI in designated high-cost areas, effective for loans purchased on or after 1 December 2022. For a borrower with a 650 credit score and a 90 percent loan-to-value ratio the waiver is worth 2.75 percent of the loan, or $6,875 on $250,000 (section 2.1); the average present value is about $2,500.

The data. The authors merge confidential Home Mortgage Disclosure Act (HMDA) records, covering nearly all U.S. mortgage applications with lender and loan officer identifiers, with public Fannie Mae and Freddie Mac loan-level files that supply a first-time-buyer flag. About two-thirds of GSE records match (section 3.2).

Identification. The main design is difference-in-differences: eligible borrowers against those just over the income cutoff, before and after December 2022. The outcome is the note rate adjusted for discount points, using a monthly rate-point tradeoff (36 basis points per point on average). Checks include a placebo test on non-first-time buyers, which shows no divergence, and triple-difference and regression-discontinuity estimates, which the authors report are close to the baseline.

Findings

Pass-through. Under full competition, eligible borrowers' rates would fall by the whole value of the waiver. The authors call the outcome "a striking deviation from this competitive benchmark." Pass-through is 31.5, 36.6, 39.3 and 36.5 percent across the four columns of their main table as controls are added; the third is the baseline.

Heterogeneity by borrower and lender. Pass-through is lowest for Black borrowers, at just over 20 percent. Banks and credit unions pass through more than nonbanks; the ten largest nonbanks pass through less than 30 percent, about half the bank rate. The authors do not estimate race differences within lender, so this is not evidence of disparate treatment.

Heterogeneity by loan officer. The authors rank lenders and loan officers by "expensiveness," from rates on their 2018-2022 loans holding borrower and loan characteristics constant; the top-to-bottom quintile gap is 26 basis points for lenders and 22 for loan officers within lender. The most expensive lender quintile passes through about 25 percent, the most expensive loan officer quintile less than 20 percent, and each effect holds with the other included.

What explains it. The authors' preferred reading is a search-friction model in which some borrowers do not shop. In that model every lender earns a positive margin and pass-through of a cost change equals the fraction of borrowers who shop (equation 1). Survey data they cite show nearly half of low-income first-time buyers considered only one lender, and 61 percent called an agent referral important. The authors say that separating search, referrals and compensation incentives would need data on borrowers' beliefs and search behavior, which they lack.

On policy, they write: "Subsidies delivered through intermediaries may fail to reach intended beneficiaries when pricing discretion and borrower inattention are pervasive" (section 6).

Relation to the Georgist Case

The paper is not about land. The word does not appear in it, and it measures no house or land prices. It is no evidence for or against land value taxation, and should not be cited as either.

The paper's "rent" is also not the Ricardian rent of economic rent. It is the margin lenders and loan officers earn because some borrowers do not compare offers, a profit that vanishes in the authors' model as the share of shoppers approaches one. The authors assume that mortgage supply is highly elastic. That is the opposite of land, whose supply is fixed.

The relevance is narrow. The paper is a worked case of who captures a demand-side subsidy, a question that arises for land through tax capitalization and the Common Agricultural Policy, where a subsidy attached to land is predicted to end up with owners. Here the leakage runs to intermediaries with pricing discretion under elastic supply: a different mechanism and recipient. It also offers a micro-level example of intermediary extraction of the kind discussed on rentier, at the finance frontier of the rent gradient, where whether such margins count as rents in the Georgist sense is contested and this study does not settle it.

Nuances and Limits

• Working paper. The cover says NBER working papers "have not been peer-reviewed." The figures are the authors' estimates.
• The $1.5 billion total. The text states the figure but does not show how it is derived. The "over 60 percent" share is consistent with one minus the baseline pass-through, but the paper gives no dollar amount retained.
• Unit of the estimate. The abstract says "mortgage intermediaries" while the body says lenders; brokers and real estate agents are not measured separately.
• Posted rates and assumptions. Pass-through depends on the rate-point tradeoff: a rule-of-thumb 25 basis points gives about 50 percent (footnote 25), and ignoring points altogether gives 28 percent (section 5.1).
• Alternative explanations. Price dispersion can reflect unobserved lender costs, such as "high-touch" service. The authors read the lower pass-through of more expensive lenders as evidence of pricing power rather than cost, but this is an inference. Strategic income reporting near the cutoff is possible, and the authors say they cannot cleanly estimate the demand response.
• External validity. This is one program, in a high-rate period of muted activity, for a targeted population. The authors note that other studies of g-fee changes found full or nearly full pass-through, and speculate the difference reflects the less experienced borrowers targeted here. The paper describes the sample as 2022-2023 in section 3.1 and 2022-2024 in section 4.2.

See Also

• Economic Rent — the Ricardian concept; the paper's search-friction "rent" is a different, imperfect-competition use of the word
• Rentier — income from ownership or position rather than production, including contested boundary cases
• CEA 2016: Market Power — the macro-level market-power brief; this paper is point-of-sale evidence in one consumer credit market
• Financialization of Land — mortgage credit and land; this paper concerns credit pricing, not land
• Tax Capitalization — the mechanism by which land absorbs taxes and subsidies, in contrast to the intermediary capture found here
• The Rentier Economy — the narrative that rents are a growing share of income; this paper is a frontier case, not support for it

Sources

1. Neil Bhutta (Federal Reserve Bank of Philadelphia), Benjamin J. Keys (University of Pennsylvania, The Wharton School, and NBER) and Adithya Raajkumar (University of Pennsylvania, The Wharton School), "Pricing Power in the Mortgage Market: Evidence from GSE Fee Waivers," NBER Working Paper 35907, October 2026, DOI 10.3386/w35907. nber.org — used for the program, data, identification, pass-through estimates, heterogeneity, mechanism, quotations and limits above (B-claim; Tier 1 authors by rubric; working paper, not peer-reviewed; read in full from a text rendering of the PDF, and page and section references are to the working paper's own numbering).