And Then You're Dead

The Permanent Record / Dossier · Tenant Screening

The Vendor in the Middle

A landlord who turns you down has usually not read anything about you. They have read a summary produced by a company you have never heard of, which bought its raw material from a second company you have never heard of, which scraped it from a courthouse database that warns on its own website that it is not an official record.

In November 2022 the Consumer Financial Protection Bureau published a market study of that industry. It found no independent, publicly available evidence that tenant screening reports are reliably predictive of anything, no federal registration system that would even reveal how many such companies exist, and a cost structure in which pulling a court record automatically runs under a dollar while having a person check it runs about seven. Of the seventeen firms it examined, five advertised any manual verification at all.

And Then You're Dead · October 2026 · 10 min read

Join the discussion →
A hand-lettered sign in a New York City window in 1938 advertising apartments for rent, above a row of tenement storefronts
61st Street between First and Third Avenues, New York, August 1938, photographed by Walker Evans for the Farm Security Administration. In 1938 the landlord decided. The decision has since been outsourced to a vendor the applicant cannot name in advance. Photo: Library of Congress, public domain, via Wikimedia Commons.

Start with the thing the market study could not establish. The CFPB looked for evidence that tenant screening reports work — that they predict who will pay rent, who will damage a unit, who will stay. It reported that there is no independent or publicly available evidence that tenant screening reports are either reliably predictive of future rental behavior or otherwise reduce risks and costs to landlords. Most residential landlords use them anyway. Roughly 68 percent of renters pay an application fee, and those fees are frequently what pays for the report.

That is the arrangement in one sentence: the applicant funds the production of a document about themselves, which they usually cannot see before it is used against them, produced by a company chosen by the landlord, using data whose accuracy nobody in the chain is paid to guarantee.

Thirteen Thousand Courts

The raw material is public records, and public records in the United States are not a database. Screening companies obtain eviction and criminal record information from more than thirteen thousand federal, state, and local courts. There is no single government repository that spans them. Jurisdictions differ in what identifying information they publish, whether they release bulk files or only individual search results, and what they call a given offense.

Because no company can maintain a live connection to thirteen thousand courts, most of them buy. The market study names LexisNexis Risk Solutions, First Advantage and RealPage as examples of private collectors who aggregate court records into centralized databases and sell access downstream. So the file that decides your application has usually passed through at least two companies before the landlord sees it, and the errors of the first are inherited by the second.

The National Center for State Courts has told background check companies as much in writing, warning that online court databases are not official records and may be subject to errors or omissions requiring further verification for accuracy, currency and completeness. That warning is the industry's central economic problem, because verification is exactly the step the business model is built to skip.

A Dollar Against Seven

The CFPB's market monitoring put a price on it. Fees a screening company pays for a credit report or a court record are rarely more than a dollar. A manual verification — a person contacting a county courthouse, pulling the physical record, or simply checking whether the first name, last name and address on the result match the applicant — costs around seven dollars. Seven times as much, for the one step that catches the wrong-person report.

Out of the seventeen tenant screening companies the CFPB reviewed, only five advertised any form of manual verification process, and some of those did it only when a consumer disputed something or a landlord specifically asked. The Bureau's summary of the resulting incentives is unusually blunt for a market study: market incentives generally value comprehensiveness of derogatory information at the expense of accurate information, and data quality checks can be costly, providing further incentives against ensuring accuracy. A screening product competes on how much dirt it surfaces, not on how much of that dirt belongs to the applicant.

0
federal, state and local courts that tenant screening companies draw eviction and criminal records from, with no single government database spanning them
0
of the 17 tenant screening companies the CFPB examined that advertised any form of manual verification — the step that costs about seven dollars against under one for the automated pull

Wildcards

What replaces verification is matching, and matching is where the wrong person enters the file. The practice lawsuits and researchers keep surfacing is wildcard matching: searching court databases on a partial name, such as a first initial with a last name, on the theory that a broader net catches more true positives. It also catches strangers.

The Federal Trade Commission described the mechanism concretely when it sued RealPage in October 2018 in the Northern District of Texas. According to the FTC, RealPage's automated system used broad criteria to match applicants to criminal records and applied limited filters to the results, requiring an exact match on the applicant's last name together with a non-exact match on first name, middle name, or date of birth. A search for an Anthony Jones born on a given date could therefore return records belonging to an Antony Jones, an Antonio Jones, or an Antoinette Jones sharing that birth date. RealPage paid three million dollars.

In November 2021 the CFPB issued an advisory opinion stating the legal consequence directly: a consumer reporting company that uses insufficient identifiers to match information to consumers — name-only matching among them — is not following reasonable procedures to assure maximum possible accuracy under the Fair Credit Reporting Act.

Two years earlier the Bureau had described a still stranger failure mode. After a consumer successfully disputes an erroneous record, the same matching logic can simply collect it again on the next run. The correction and the error are produced by different processes, and only one of them is automated.

Verification costs seven dollars and the record costs one. Everything else about this industry follows from that ratio.

The Score That Hides the File

On top of the records sits a second product. Larger vendors sell proprietary tenant risk scores — a number, a grade, or a bare accept-or-decline recommendation — generated from the underlying data by models they do not publish. TransUnion markets ResidentScore; SafeRent markets the SafeRent Score; Entrata, RealPage, National Tenant Network and others sell comparable products, some described as artificial intelligence models.

The CFPB's finding about these models is a statement about absence, and it is worth quoting precisely: in contrast to the documented model risk management in the financial services space, the Bureau wrote, it was unaware of objective validation of tenant screening company models or detailed descriptions of the specific variables or weights used in a given model. It also noted that two larger companies with ostensibly sophisticated models define a bad tenant outcome differently from each other — TransUnion in terms of evictions, three or more late payments and insufficient funds; RealPage in terms of payment defaults, balance owed after departure and share of move-outs — so a landlord cannot compare the two even if they wanted to.

The more consequential design choice is what the score conceals. Some reports deliver the recommendation without the underlying records, so the landlord who declines you cannot see the court entry that drove it and therefore cannot tell you. Private litigation over CoreLogic's CrimSAFE product, in the District of Connecticut, alleged exactly this: a product returning a yes-or-no indicator against the client's pre-set criminal criteria, with the underlying record suppressed. A scoring layer that sits on top of unverified data does not just pass the errors through. It makes them unreviewable.

Meanwhile the one data source that would obviously predict whether someone pays rent barely exists. Industry estimates put coverage of rental payment history in the consumer reporting system at between 1.7 and 2.3 percent of American renters. The screening report knows about your judgments. It does not know that you paid rent on time for nine years.

An Industry Nobody Counts

The reason so much of this has to be described through lawsuits is that there is no other window. There is no required federal registration system for the consumer reporting industry, and, as the CFPB put it, the total number of market participants is unknown. The market is bifurcated: hundreds of small regional companies serving independent landlords, and a handful of large firms. A few states — Maine, Maryland, New York — register credit reporting agencies; California and Vermont register data brokers. Nationally, an applicant cannot look up which company will produce the report about them, which means they cannot check it in advance, which means the first time they encounter the file is when it has already been used.

What the CFPB does have is complaints. Between January 2019 and September 2022 it received roughly 26,700 complaints related to tenant screening, rising from about 300 a month at the start of that window to nearly 700 a month by the end. More than 17,200 concerned incorrect information appearing on a report. Another 5,000 concerned a problem with the company's investigation into an existing issue. More than 3,200 concerned improper use of the report.

What Enforcement Has Established

Three federal actions have put specifics on the record. The RealPage settlement in 2018 established the matching criteria. In December 2020 the FTC settled with AppFolio for $4.25 million, alleging that until at least April 2019 the company had no reasonable procedures to ensure that criminal and eviction records it received from a third-party vendor were accurate before including them — producing reports with records belonging to people of a different name or birthdate, records with a missing or inaccurate offense name, type or date, records with a missing or inaccurate disposition, and multiple entries for the same underlying case. The same order addressed the inclusion of non-conviction criminal records and eviction records more than seven years old.

In October 2023 the FTC and CFPB together settled with TransUnion Rental Screening Solutions and its parent Trans Union LLC in the District of Colorado for fifteen million dollars — eleven million in consumer redress and four million in civil penalties, the largest sum the FTC has recovered in a tenant screening matter. Among the allegations: that until April 2021 the subsidiary often reported developments within a single proceeding as separate events; that it labelled claimed amounts as a "Judgment Amount," implying a court had awarded them; that it lacked reasonable procedures to keep court-sealed eviction records out of its reports; and that it failed in many instances to tell consumers the name of the third-party vendor its information came from — which is to say, the applicants could not find out who had assembled the record, because the company that sold it would not say. The settlement requires it to disclose those sources, to give consumers free access to their complete file, and to supply landlords with a template adverse action letter.

And Then You're Dead

The screening report is a manufactured good, and it is manufactured cheaply because nobody in the transaction is the customer for accuracy. The applicant pays for it and cannot see it. The landlord reads it and cannot audit it. The vendor sells comprehensiveness, buys its raw material from an aggregator, and skips the seven-dollar step. The score on the front page is produced by a model no outside party has validated, using a definition of a bad tenant that differs between the two largest firms selling one. And the entire industry sits outside any registry that would let you learn its name before it has already answered a question about you.

That's it. That's the whole thing.

Comments

Loading…

Comments are anonymous and moderated.

Sources