The Tri-Merge Standard: Credit Reporting and Mortgage Market Safeguards

The Issue

Mortgage affordability should be measured by what a household pays for sustainable access to credit. A smaller fee at closing is a benefit, but it is not a complete measure of reform. We recommend retaining tri-merge as the default for loans delivered to Fannie Mae and Freddie Mac until public evidence establishes that a reduced-report alternative can preserve underwriting quality and produce net consumer savings.

Tri-merge combines information from Equifax, Experian, and TransUnion. A credit report supplies account-level information; a credit score summarizes risk using a model. Changing the model and changing the number of reports are distinct decisions. Treating them as one reform obscures what is being tested and which change produces a particular outcome.

FHFA announced an optional two-bureau approach in 2022, but in January 2025 moved its implementation to a date to be determined. On September 9, 2026, the Enterprises expanded VantageScore 4.0 availability to all approved lenders for eligible loans. Classic FICO remains eligible; FICO 10T is not yet eligible for delivery. FHFA says the introduction of VantageScore 4.0 does not initially change existing credit-reporting requirements.

The Mortgage Bankers Association’s December 2025 proposal would relax the tri-merge requirement for borrowers with credit scores of at least 700. Its strongest argument merits consideration: requiring three purchases can weaken price competition, and additional information should justify its cost. In April, MBA President Bob Broeksmit cited American Enterprise Institute research finding similar predictive power from a single score when selection is not manipulated. He supported advance bureau selection and submission of all reports when multiple reports are pulled.

The information problem is separate from the fee problem. SPPI’s February commentary identified uneven reporting across bureaus as a reason to preserve a broader view of a borrower’s obligations and payment history. Combining reports cannot guarantee that every debt or payment is captured, and it does not make an inaccurate item correct. It does, however, reduce dependence on the contents of one bureau’s file. A replacement standard should demonstrate how it preserves useful information while addressing errors and duplication.

Consider two possible omissions. If a report leaves out a debt obligation, an underwriter could overestimate repayment capacity unless another verification process identifies it. If it leaves out a record of timely payments, the borrower could appear less creditworthy than a fuller record would suggest. These are mechanisms to test, not estimates of how often either outcome occurs. Counting additional approvals alone cannot distinguish better identification of qualified borrowers from weaker screening.

Andrew Davidson & Co.’s February 2026 study used VantageScore 4.0 data representing 245 million scored consumers in October 2023. It found that 35% had at least one bureau score 10 or more points from the three-score median; 18% had a gap of at least 20 points, and 7% at least 40 points. Figure 1 presents these overlapping thresholds. The population includes people who were not mortgage applicants; these are score differences, not observed rates of denial or default.

For a $350,000 loan at 90% loan-to-value, the study illustrated a $3,000–$5,000 present-value change in borrowing and mortgage-insurance costs from moving between adjacent pricing tiers. That is a scenario, not an average loss or a forecast for every borrower. Its significance is the size of the potential pricing consequence relative to an origination fee. Whether that consequence materializes depends on the applicable pricing rules, the selected score, and the borrower’s circumstances.

We should also distinguish prediction from decision consistency. A model can rank risk well across a large population while individual applicants receive different treatment depending on which report supplies the inputs. Conversely, disagreement with the tri-merge median does not itself prove that the alternative score is less accurate. Public testing should connect score differences to realized repayment outcomes and actual lending decisions. Neither a correlation statistic nor a count of score changes, standing alone, answers the affordability question.

Selection incentives make this comparison harder. If a participant can inspect several results and deliver only the most favorable one, the delivered score reflects both credit history and the selection process. MBA expressly recognizes this danger and calls for enforceable restrictions against cherry-picking. The policy task is to specify how compliance would be verified, including repeat pulls and resubmissions. A written prohibition is insufficient if the purchaser cannot determine which reports the originator obtained.

An evaluation should therefore compare two separate situations: a bureau selected before the score is known, and a process in which participants can respond to information already observed. The first isolates the effect of reducing the information set; the second tests behavior under the proposed rules. Combining them could conceal whether any measured difference comes from missing data or selective submission. Audit trails should record report orders and bureau choices without making personal credit histories public.

SPPI’s April commentary warned that apparent origination savings could be offset by downstream repricing and risk transfer. That is a policy concern to investigate rather than a measured forecast of a particular rate increase. If borrowers receive lower fees but higher financing costs, the reform has shifted the bill. If investors or guarantors absorb additional losses, a lower borrower charge may reflect a subsidy rather than an efficiency gain. A credible cost comparison should identify both the amount and the recipient of each saving or additional cost.

The same discipline should govern claims about inclusion. We recommend separate reporting for first-time buyers, applicants with limited credit histories, and borrowers near eligibility or pricing thresholds. An average result may hide offsetting gains and losses. Policymakers should ask whether a reduced-report process rejects applicants whom fuller information would support, as well as whether it approves applicants whose repayment capacity was overstated. Any use of positive alternative data should be tested for availability across the permitted reporting channels.

What We Recommend

1. Retain tri-merge as the Enterprise default pending public validation. FHFA, Fannie Mae, and Freddie Mac should publish matched-applicant comparisons of one-, two-, and three-bureau underwriting, with results for approvals, actual loan prices, mortgage insurance, repurchases, and subsequent performance.
2. Separate price competition from changes to the information standard. FHFA should continue score-model competition while requiring a comparable accounting of reporting, licensing, reseller, and implementation costs sufficient to test whether savings reach borrowers.
3. Limit any reduced-report pilot and make its safeguards auditable. The Enterprises should require advance bureau-selection rules, disclosure of every report obtained, independent compliance reviews, subgroup results, and published expansion or suspension criteria tied to net consumer benefits and risk.