As of November 2023 none of the federal agencies GAO interviewed reported an effort or plan to measure evictions nationally, and in February 2024 GAO recorded that, according to HUD, the department has no legal authority to require states and localities to collect eviction data

What is happening?

The United States has no single national source of data on evictions. Eviction is governed mainly by state and local law and handled by local courts, and the Government Accountability Office reported in February 2024 that researchers do not share one definition of what an eviction is.

The Department of Housing and Urban Development told Congress in October 2021 that no national survey or data source captures the full magnitude of court-ordered evictions, and that nothing at national scale covers how common court-ordered, extra-legal and administrative evictions are, where they occur, or who is involved.

Four years later the position had not moved. As of November 2023, none of the federal agencies GAO interviewed reported an ongoing effort or plan to collect data measuring how common evictions are nationally or who the parties to them are. GAO recorded in February 2024 that, according to HUD, the department has no legal authority to require states and localities to collect eviction data or report it to a national database.

What exists instead is an outside estimate. The most recent year for which a national filing rate is available is 2018, at 7.8 filings per 100 renter households, produced by a university research team from the court records it could obtain and from statistical modeling where it could not. GAO described that in February 2024 as the most recent year with a national rate.

Whose problem is this?

The body that would have to build the count is not the body that holds the records, and the records sit in thousands of separate local courts that answer to neither.

GAO also recorded who would use better data, and what currently limits that use. Existing federal programs in this area, emergency rental assistance and HUD grants for eviction prevention legal aid, are temporary or small, which GAO said limits the opportunity to use improved data. The benefits the interviewed court jurisdictions themselves named were local: placing services where filings are concentrated, and comparing one jurisdiction against another.

Where does this problem end?

This document is about whether anyone measures the thing, not about whether the thing should happen more or less often.

What is the state now, and what should it be?

Now

What is missing here is not a design but an instrument, because a design exists and a price has been put on it.

Needs a new measurementthe target state: no source opened here names a deadline, a statutory target, or an adopted plan for measuring evictions nationally. HUD wrote in October 2021 that there are sound reasons for the government to track how common evictions are nationwide, and set out three strands to do it covering court-ordered evictions, evictions outside court, and evictions inside HUD-assisted housing. That is a direction rather than a target. Nothing opened here states by when a national count should exist, how complete it would have to be, which agency would own it, or what margin of error would be acceptable.

How big is it?

There is no national count of the households this affects. The population file for this document records not-derivable rather than a range, and the reason is visible in the table below: the available figures are in three different units, from three different years, and no source opened here combines them.

The same figure carries two meanings inside one report. HUD uses 3.7 million in its summary as households affected each year, and in its first chapter as eviction cases filed in 2016. A household can face more than one case across a year and a case can involve more than one person, so the two are not interchangeable. This document keeps them apart and does not reconcile them.

Multiplying any two of these numbers together would produce a figure that no source states and that nobody could check.

Under what conditions does it arise?

1. Two different events are being measured and only one of them leaves a document. A court filing creates a record by definition. An eviction that happens without a court creates nothing, because the event is a move and nothing labels a move as an eviction unless somebody asks the household. In the only national estimate available, the informal share was the larger of the two.

2. The records that do exist are held by thousands of separate courts. Definitions, case management systems and disclosure rules differ by state and by county. Of the 8 jurisdictions GAO interviewed, none collected demographic characteristics of tenants or owners, 5 cited staffing constraints and 4 cited a lack of technology. An official in one jurisdiction told GAO that collecting additional data would bring little benefit, and some said the purpose of a court is to resolve cases. Those are statements GAO relayed, not findings of this document.

3. Access to what exists is uneven, and the report is not internally consistent about it. A survey cited by GAO found 35 states and territories publishing annual eviction summary statistics, 25 states and the District of Columbia releasing civil records at no charge, and 10 releasing individual records for a fee. On sealing, the same GAO report gives two numbers for the same survey as of 2021: 9 states in the body and 7 in a footnote. This document reports both and chooses neither.

4. No federal instrument reaches the record holders. GAO recorded in February 2024 that, according to HUD, the department has no legal authority to require states and localities to collect eviction data or report it to a national database. HUD wrote the narrower version of this itself in October 2021: without additional legislation from Congress it cannot compel states to undertake the work of changing their court systems and information technology so that national data becomes possible.

5. The most complete court-ordered eviction data is not a public product. HUD wrote in October 2021 that the most comprehensive data of that kind relies heavily on records assembled by a commercial court-records vendor, described that reliance as practical but not ideal, and named as a potential ethical concern that the vendor gathers those records in order to sell them to landlords for tenant screening. Those are HUD statements carried here as HUD made them.

A record that exists in a courthouse is not the same as a record that anyone can add up.

What has been tried?

Two directions were pursued at once and neither produced a recurring national measure. Federal work did not stop, and the record should not be read as though it had. HUD produced its own recommendations in the 2021 report. A university research team worked with the Census Bureau to link 38 million court records from 2007 through 2016 to administrative data and reported a match rate of 65 percent, which is how the characteristics of the parties have been studied at all. HUD was updating its assisted-housing tenant system while GAO was reporting. GAO also recorded that neither of the two main approaches, a national court-records database or stronger survey questions, serves every purpose.

What was found?

Why is it still unsolved?

Measurement absent — the records that would answer the question are held by thousands of local courts, and no federal instrument reaches them.

The measurement problem is two problems wearing one name. Court filings can in principle be counted, because a filing creates a document with a date on it. Evictions that happen without a court cannot be counted that way, because the event is a move, and a move records itself as nothing in particular unless somebody asks the household what happened. The only national attempt to ask was a survey module fielded once in 2017, and GAO reported in February 2024 that its data had not been released publicly as of that date and that the questions were not repeated. In the estimate that module produced, the uncounted half was the larger half.

The second part is that being asked for data and being able to require it are different powers. Congress directed a feasibility study, HUD produced one with three designs and a price attached to each, and the chain stops there. HUD wrote in October 2021 that without additional legislation from Congress it cannot compel states to undertake the work of changing their court systems and information technology, and suggested that Congress consider attaching, alongside additional funding, a requirement that states take part so far as that is feasible. As of November 2023 GAO found no federal agency among those it interviewed with an effort or plan under way.

The third part, which is what keeps the pattern stable, is that the gap produces no event. A missing count does not arrive on a date, does not expire, and does not fail an audit. Filings keep being recorded in courthouses one case at a time, the 2021 feasibility report remains the most recent federal design, and every few years a partial estimate arrives from outside the federal government and gets read as though it were the national figure. Nothing in that sequence creates a moment at which anyone has to decide anything.

What observation would mean it is solved?

Candidates — (a) a federal agency publishes a recurring national count of evictions with a stated definition, a stated coverage and a release schedule (b) the share of county-years with reliable court-issued filing counts rises toward full coverage (c) a national survey publishes a repeated measure of physical eviction that includes moves which never reach a court.

(a) alone is weaker than it looks. A national count built on court records measures only the half of the event that generates a document. In the only national estimate available, the informal share was the larger one, so a court-based count could be complete on its own terms and still miss most of what its name claims to describe.

(b) alone counts courts rather than households. A filing count records that a case was opened. GAO reported in February 2024 that court records frequently do not record whether the tenant actually moved, so rising coverage of filings would raise confidence about filings and say nothing about outcomes. It would also leave the parties unknown, since none of the 8 jurisdictions GAO interviewed collected characteristics of tenants or owners.

(c) alone has already happened once, and once is what the problem is. A module was fielded in 2017, produced the only national estimate of physical eviction there is, and by GAO record in February 2024 had not been released publicly and had not been repeated. A single unreleased release is not a measure. The test for this candidate is repetition and publication, not existence.

The three have to be read together, and the definition question sits underneath all three. A count of filings, a count of judgments and a count of completed moves are three different numbers, and no source opened here says which of them a national measure is supposed to report.

What is it connected to?

Fills with researchrental housing affordability and the supply of rental assistance, tenant screening and the secondary market in court records, homelessness measurement, right to counsel programs, and the measurement of forced residential moves generally. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• No source opened here names a deadline, a statutory target or an adopted plan. Neither the HUD report of October 2021 nor the GAO report of February 2024 says by when a national count should exist, how complete it would have to be, or which agency would own it.
• GAO does not reconcile its own two counts of states that seal eviction records. The body of the February 2024 report gives 9 states with sealing or expungement laws as of 2021, and footnote 51 gives 7 for the same underlying survey. Neither number is corrected, and neither is explained.
• GAO states that the national geographic range of physical eviction rates is unknown. Which places are high and which are low is not established by anything opened here.
• GAO states that there is no other national estimate against which the 2017 informal eviction figure could be compared.
• HUD wrote in October 2021 that its own administrative systems do not collect evictions from its rental assistance programs. The system update described above was still under way when GAO reported, and nothing opened here reports its result.
• GAO records that HUD had not released the 2017 module data publicly and that the questions were not repeated. GAO reports a HUD staff statement that the sample might be too small, and stops. No source opened here goes further than that.
• The February 2024 report read here contains no recommendations section. Its contents run from background, to information on prevalence, to court records and surveys, to key considerations, to agency comments; and the agency comments section records only that HUD, the Census Bureau and the Consumer Financial Protection Bureau provided technical comments which were incorporated. That is not the same thing as GAO deciding to recommend nothing, and this document does not read it that way.
• No source opened here carries a HUD position later than October 2021. The position of the department in this document is taken from its own report of that date and from the statements GAO recorded in February 2024. A more recent public HUD response was looked for and not found.
• The 2025 tracking figures were read only in republication. The report of the university research team and the original news article both refused automated requests, so those filing counts rest here on a regional outlet that republished them.
• The successor household survey is not established as counting evictions that happened. The April 2026 release lists risk of losing current housing to foreclosure or eviction among its topics and calls itself an experimental data product. No source opened here says whether the forced-move question carried over, or whether any question in it counts completed evictions rather than risk.
• The bill status record read here ends at committee referral. What happened to H.R. 8185 after 2026-04-02 is absent from every source opened here, and no source says that it failed.
• No result document was found for the research grant linking eviction records to voucher data. GAO reported that initial results were expected in autumn 2024; nothing opened here reports them.

See the evidence

Two federal primary documents were read directly and they carry most of this dossier. The February 2024 report was read as a 38-page PDF and the October 2021 report to Congress was read in its government deposit copy; between them they supply the authority statement, the absence of any federal effort as of November 2023, the coverage and quality limits of the outside dataset, the 2017 survey module and its non-release, the three costed designs, and the survey cost benchmarks. Two further primary records, the text and the status of the 2026 House bill, were read directly. Where the two reports differ from each other the difference is left visible rather than resolved: the wording HUD used about its own limits in 2021 is narrower than the wording GAO used to summarize it in 2024, and both wordings appear above in the form each document used. Where a single report differs from itself, as with the 9 states against 7 states for sealing laws, both numbers appear and neither is chosen. The weakest material here is the set of 2025 tracking figures, which could be read only in republication because the report and the original article both refused automated requests; those figures are graded low and nothing load-bearing here rests on them. This is a Path A output, a definition assembled from public documents rather than from a registry source, so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range Not derivable

The reason and what is missing are listed under “What is missing” below