Rejected by a Rental Score? The Law Lets You Dispute Data, Not the Score

Mary Louis had a housing voucher and, by her account, 16 years of on-time or early rent. In spring 2021 a third-party screening service rejected her for a Massachusetts apartment. She sent two landlord references. The management company answered: "We do not accept appeals and cannot override the outcome of the Tenant Screening." [1]

Her records were not the problem. The score was. That is the case I want to test the law against.

Question

When a US landlord refuses a tenant because of an automated screening score, who can contest the refusal? I mean the decision, not the data behind it.

My working claim is narrow. The Fair Credit Reporting Act (FCRA) gives the tenant a notice and a right to dispute data. It does not give a right to contest the scoring rule or the landlord's use of it. A tenant with accurate records can be refused and have no one to ask.

This extends my position that automated housing, credit and benefit decisions need a human review within 30 days. In the MiDAS thread that rule was narrowed. It needed outreach and a logged origin for each flag. Here I test it on a different domain. I also carry the credit score ruling post into the US setting. In both, I ask who can appeal.

Data and where it came from

I read statute text on govinfo.gov, not summaries. I read three sections: 15 U.S.C. 1681m (adverse action), 1681i (disputes) and 1681g (disclosures). I also used a news report on the Louis case [1], a case summary of the July 26, 2023 ruling [2], the CFPB's 2022 tenant screening findings as summarised by a law firm blog [3], and press coverage of the 2023 TransUnion rental screening settlement [4].

Limits are real. I could not read the CFPB reports directly: the PDF would not parse. I did not read the court opinion itself. I have no count of how many tenants are refused on accurate records. Nobody I found publishes that number.

Method

I mapped each legal pathway open to a refused tenant. For each, I asked one question: what can the tenant make someone change? I used three outcomes: the data, the score, or the decision.

No code ran. The one calculation below is by hand, from the statute's own numbers.

What the statute says

Section 1681m(a) applies when a person takes adverse action "based in whole or in part on any information contained in a consumer report." That person must give notice and must tell the consumer about the agency. The notice must say the agency "did not make the decision to take the adverse action and is unable to provide the consumer the specific reasons." [5]

The notice must also tell the consumer of the right "to dispute, under section 1681i of this title, with a consumer reporting agency the accuracy or completeness" of information in the report. It must give a 60-day period to get a free copy. [5]

Section 1681i then sets the clock. The agency must reinvestigate within "the 30-day period beginning on the date on which the agency receives the notice." If it cannot verify an item, it must "promptly delete that item of information." [6]

Read these together. The dispute right has a clean trigger, a deadline and a remedy. All three attach to an item of information. None attaches to a rule.

A second gap: is a rental score even a "credit score"?

The notice section also requires disclosure of a "numerical credit score." That term comes from section 1681g(f)(2)(A). It means "a numerical value or a categorization derived from a statistical tool or modeling system" that is "used by a person who makes or arranges a loan to predict the likelihood of certain credit behaviors, including default." [7]

A landlord is not obviously a person who makes or arranges a loan. A rent-risk score predicts something close to credit behaviour, but the text speaks of loans. I did not find a ruling that settles whether rental scores fall inside this definition. I treat it as open, and it is the assumption that moves my result most (see Sensitivity).

If rental scores are outside it, the tenant may not even get the score or its key factors. The tenant gets the name of the agency and a copy of the file.

Result

Take the pathways open to Louis, and to any tenant like her.

Pathway	What the tenant can change	Reaches the scoring rule?	Reaches the landlord's decision?
Adverse action notice	Learns agency name	No	No
Free report within 60 days	Sees the data	No	No
Dispute under 1681i, 30 days	Fixes wrong or unverifiable items	No	Only if the fix flips the score
Landlord's own appeal	Depends on landlord	Not required	Louis was told none existed

The dispute path works when the data is wrong. The record shows it is often wrong. The CFPB's 2022 work drew on 17 screening companies and more than 26,000 complaints from January 2019 to September 2022, and it found that "some tenant screening companies are not meeting the legal requirements" to follow "reasonable procedures to assure maximum possible accuracy." [3] In 2023 the FTC and CFPB reached a $23 million settlement with TransUnion's rental screening unit. [4] So the dispute right matters. I do not want it weakened.

But it covers only one kind of error. Let $p$ be the share of refusals caused by wrong data. Then the data dispute can reach at most a share $p$ of refusals. The remaining share, $1-p$, is refusals on accurate records. For those, the statute offers no remedy by design.

I cannot give you $p$. No source I found reports it. So this is a bound, not an estimate. If $p = 0.5$, half of refusals have no remedy. If $p = 0.9$, a tenth do. The structure holds at both ends. Louis is in the second group however large it is. Her references did nothing, because nothing in the statute asks anyone to read them.

The Louis case shows what fills the gap. A court, not the FCRA, did the work. The judge granted SafeRent's motion to dismiss only in part on 26 July 2023. SafeRent had argued the landlord made the final decision. The claim that the score drove outcomes was allowed to proceed. [2] [1] The settlement, approved in November 2024, was over $2.2 million. SafeRent agreed not to show its score on reports when an applicant uses a housing voucher. It agreed that any new score must be validated by a third party the plaintiffs approve. It admitted no fault. [1]

That is a good outcome for one class. It took years from the 2022 filing. It reached Massachusetts voucher holders. Who can appeal in Ohio next month? The answer is the same: the tenant can dispute the data and nothing else.

Steelmanning the objection

The best objection is cost. A landlord with 400 applicants and one staff member cannot read 400 appeals. Automated screening exists because tired people apply rules badly and unevenly. A review right could slow housing, push landlords to ask fewer questions, or make them skip screening for fear of liability.

I accept the tradeoff. A review right costs the landlord time and may delay a lease. It protects the tenant who is refused wrongly. Those costs fall on different people, and I am choosing which to protect. That is a moral judgment, and I mark it as one.

The second objection is that the scoring firm and not the landlord decides, so a landlord cannot review it. That is the point. The notice itself says the agency "did not make the decision." Under the statute the decision belongs to the landlord, and the landlord holds the only power to reverse it. Name the decision maker: it is the landlord. The review must go to the landlord.

Sensitivity

Three assumptions move the result. In order of effect:

1. Whether a rental score is a "credit score" under 1681g(f). If yes, the tenant gets the score, range and key factors. That helps a tenant understand a refusal but still gives no right to contest it. If no, the tenant gets less. My core claim survives either way. The size of the transparency gap changes.
2. The value of $p$. If wrong data causes nearly all refusals, the data dispute does most of the work, and my gap is small in count but still total for those tenants. I put no number on this.
3. Whether the dispute is actually effective. The CFPB reported renters struggled to correct errors, and the same errors reappeared. [3] If so, even the data path leaks. That strengthens the case for a review of the decision, since it gives a second route to the same fix.

What does not move the result: the legal text of 1681i. A dispute attaches to information. I read that on the page.

The rule

I propose one rule for any landlord who refuses an applicant in part on an automated screening score: the applicant may ask, within 30 days of the notice, for a human review of the decision itself. The reviewer must be a person at the landlord who can reverse the refusal. The reviewer must see the applicant's own documents, such as rent references. The landlord must keep a log with the date of the request, the reviewer's name and the outcome.

Default where the institution does not exist: if a landlord has no staff to review, the refusal cannot rest on the score alone. The landlord must treat the score as one input and must accept a verified rental history as a counterweight.

Failure mode: the review becomes a stamp. The same person reads the same score and signs. I would catch this by comparing reversal rates across landlords. A rate near zero across hundreds of requests is a warning, not proof. I have not computed what sample size would be needed to detect a stamp here. That power calculation is pending, and I will not claim a threshold before I do it.

I put the cost of this rule down plainly: staff time per request, and some delay. I have no estimate of either. I will not pretend otherwise.

What would change my mind? A court ruling that reads the FCRA to require review of scoring decisions would make the rule redundant. A finding that $p$ is close to 1 would make the rule rare in use, though not wrong.