We ran a Fairspire Review on the 2024 home-purchase decisions of three mid-size New Jersey lenders, using the public mortgage data every lender has to report. All three passed the four-fifths rule. One of them still showed a gap that would worry any compliance team.
The data
Under the Home Mortgage Disclosure Act, lenders report every mortgage application and its outcome, and the federal government publishes the data. In 2024, 564 lenders reported more than 73,000 home-purchase applications in New Jersey that were approved or denied. We picked three mid-size lenders, each with between 1,100 and 1,500 applications, and kept them anonymous: Lender A, B and C.
What the four-fifths rule said
The four-fifths rule compares each group's approval rate with the most-approved group's. Below 0.80 is the usual warning line. By that test, all three lenders looked fine:
| Lender | Lowest impact ratio | Group |
|---|---|---|
| Lender A | 0.89 | Hispanic or Latino applicants |
| Lender B | 0.93 | Black or African American applicants |
| Lender C | 0.81 | Black or African American applicants |
No ratio below 0.80. A review that stopped here would report no concerns.
What the full review found
At Lender C, we compared applicants with similar financial profiles, the applications a lender would normally approve. Among them, Hispanic applicants were denied 3.3 times as often as everyone else. The difference is statistically significant (p = 0.005), so it is unlikely to be chance.
Lender C's ratio for Black applicants, 0.81, also sits right on the warning line, from a small group. One more year of similar decisions could take it below.
Public mortgage data has no credit scores, so these are signals for a closer look, not findings of discrimination. That is exactly what a review is for: knowing where to look before a regulator, an auditor or a plaintiff does.
Why it matters
The four-fifths rule is a rule of thumb, not a legal safe harbor. Regulators and courts look at whether similar applicants are treated differently, and New Jersey's disparate impact rules apply to lending decisions whether or not a model makes them. Passing a single ratio is not the same as being fair.