No national count of adults under guardianship — the only estimate with a published method projects 77,339 counted cases onto the whole country

What is happening?

Nobody knows how many adults in the United States live under court-ordered guardianship. The only national estimate whose method has been published is 1,519,284 active pending adult guardianship cases, computed by a national court research organization in 2011 from 2008 court data, and no source opened here supersedes it as of 2026-08-08. The two authors printed their own band around that figure — fewer than 1 million to more than 3 million.

The estimate rests on four jurisdictions. Of the state court systems reporting to the Court Statistics Project for 2008, only 14 could report newly filed adult guardianship cases and only 4 — Vermont, Arkansas, Ohio and the District of Columbia — could report how many cases were active and pending. Those four counted 77,339 cases between them. That is the entire measured floor beneath the national figure, and everything above it is projection from jurisdictions holding about 5 percent of United States adults, of which Ohio alone is roughly three quarters.

The number the field actually cites is a different one. 1.3 million, paired with about USD 50 billion in assets, appears in the 2018 report of the Senate Special Committee on Aging, in a National Council on Disability report of the same year, and in a brief from a national advocacy nonprofit for older adults in March 2024. Traced to its origin, it comes from a statement given at a congressional hearing on 2017-03-22 and travels onward through an unpublished trade-association letter dated 2018-07-20. The web page the Senate report names as its source returned HTTP 404 when opened on 2026-08-08.

Four federal and national bodies have recorded the absence across twenty-two years — GAO in 2004 and again in November 2016, the Senate committee in 2018, and a national court research organization and an advocacy nonprofit for older adults in 2024. On 2026-07-21, just under seventeen years after state courts were formally asked to begin counting, that same research organization announced a USD 450,000 three-year award to produce a national estimate. The count is still a research deliverable.

Whose problem is this?

The bodies that would have to produce the count are locally governed courts, and no statute asks any of them for it. The one national estimate came from a research paper, not from a reporting duty, and when its base year passed nothing replaced it.

Where does this problem end?

The boundary matters here because what is missing is not a gap inside an existing dataset but the dataset itself. A query cannot recover a distinction that was never recorded, so the absence has to be repaired forward rather than backward.

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

Now

Needs a new measurementthe target state: no source opened here names one. There is no number of states that must report, no completeness threshold, no acceptable margin of error, and no date by which a national count should exist. The 2009 resolution asks state courts to begin collecting. The 2018 Senate report asks for increased data collection. The 2026 award asks for a reliable estimate. None of the three says how many states, by when, or how anyone would know the goal had been reached.

How big is it?

Between 1 million and 3 million active pending adult guardianship cases, on 2008 data. That band is not a range invented here — it is the one the authors of the only method-published national estimate wrote down themselves, given how far the four reporting jurisdictions differed from one another. The point estimate inside it is 1,519,284.

The unit is cases, not people, and the difference cannot be closed. In states that separate guardianship of the person from conservatorship of the estate, one adult can appear as two case records; in states counting guardianship alone, an adult who has only a conservator appears as none. The 2011 analysis excludes adult conservatorships outright. No source opened here publishes a conversion factor between cases and persons, and the direction of the correction is not even determined — double counting pushes the person figure below the case figure while excluded conservatorship-only adults push it above.

The counted floor is 77,339. That is what the four jurisdictions actually recorded for 2008: Vermont 6,783, Arkansas 29,985, Ohio 38,857 and the District of Columbia 1,714. Their combined adult population, back-solved from the rates printed in the source, is about 11.65 million against a United States adult population of roughly 228.8 million — a coverage share near 5 percent, with Ohio supplying about 73.6 percent of that base. The case-weighted rate of 664 per 100,000 adults is close to the Ohio rate pulled upward by Arkansas; the simple average of the four rates would have been 916.

The figure is not re-based to the present here. Applying the 2008 rate to the 2026 adult population would assert that the rate still holds, and that rate is precisely the quantity nobody has measured since. So the band is a 2008 band, and it stays one.

The band is not sampling error around a measurement. It is the distance between what was counted and what was asserted.

Under what conditions does it arise?

1. No counting duty exists anywhere in the system. Guardianship is created by state law and administered by state probate courts, and no federal statute requires a state to report how many adults are under one. The 2017 federal elder abuse statute directs the Attorney General to publish best practices and model legislation for guardianship proceedings, with no deadline, no appropriation and no duty to count anyone. Reporting to the Court Statistics Project is voluntary and aggregate.

2. The unit of administration sits below the unit of interest. In most states probate jurisdiction belongs to locally governed county courts running their own case systems. A state supreme court that wants a number must ask counties that were never built to answer, and the errors run both ways at once. GAO found in 2016 that some Texas counties reported no active guardianships while others reported cases that should have been closed, including cases where the person had died.

3. The category does not exist to be queried. Both the 2011 paper and the national data-elements guide of June 2024 describe the same condition: guardianships are filed inside a generic probate or civil case type, many states cannot separate guardianship from conservatorship, and others file adult and juvenile guardianships under one type. This is a missing distinction rather than missing data, and no retrospective query recovers it.

4. Guardianship is a stock while courts count flows. A filing is an event a clerk records. An active guardianship is a status that persists for years with no event attached to it. Court systems count what they do far more readily than what they remain responsible for, which is exactly why 14 jurisdictions could report filings in 2008 and only 4 could report active cases.

5. No budget depends on the number. Child welfare has a court improvement program that pays state courts to collect and report. A comparable guardianship program has been recommended since 2010, repeated by the Senate in 2018 and by an advocacy nonprofit for older adults in 2024, and never created. The recommendations are unfunded, undated, and addressed to nobody in particular.

What has been tried?

Thirty-seven years of recommendations point at the same act, and the act has never been assigned to anyone with a budget. What changed on 2026-07-21 is that somebody was finally paid to attempt it, and even that attempt delivers a modelled headline number rather than case-level data.

What was found?

Why is it still unsolved?

Measurement absent — the count does not exist because no institution has ever been assigned to produce it, and the field has learned to operate without it.

The first movement is jurisdictional. Guardianship is created by state law and administered in probate courts that are, in most states, locally governed county institutions. There is no federal reporting duty, and the national body that asked for one in 2009 is a conference of chief justices with no power to compel anything from anybody. So the unit that would answer sits far below the unit that wants the answer, and the request has to travel down through fifty separate systems and back up. Just under seventeen years after that request, the field cannot say how many of those systems now comply, because nobody counts the counting either.

The second movement is that the category was never built to be queried. A guardianship is recorded inside a generic probate or civil case type, so what is missing is a distinction rather than a dataset, and no better query against old records recovers one that was never drawn. The sharper half of that movement is this: guardianship is a stock and courts are built to count flows. A filing is an event that produces a record. An active guardianship is a status that persists for years or decades with no event attached, and so it falls outside what a case management system naturally reports. The 2008 figures show that gap exactly — fourteen jurisdictions could say how many cases were opened and only four could say how many were open.

The third movement is the one that keeps the pattern stable. The ordinary way to get a measurement funded is to show that the affected population is large, which here requires the very number that the absence of funding prevents anyone from producing. The field solved that deadlock by circulating a point estimate whose derivation was never published, and the workaround worked. The figure of 1.3 million appears in a Senate report, a federal council report and an advocacy brief; it carries an assets figure with no derivation at all; the page it cites is gone; and it is quoted in units that shift between cases and adults within a single document. A number that everybody repeats relieves exactly the pressure that would have produced a real one. Nobody in this record behaves badly. Every actor does what their mandate permits, and the absence survives because the one act nobody was assigned is the act of counting.

What observation would mean it is solved?

Candidates — (a) a national estimate is published together with its method, its base year and its coverage share (b) a majority of state court systems report active pending adult guardianship counts on a recurring schedule to a common destination (c) counts are published disaggregated by age, race, ethnicity, disability type and reason for guardianship.

(a) alone repeats the failure it replaces. The 2011 estimate already met this bar and was still a projection from four jurisdictions. The 2026 award will deliver a modelled figure that fills gaps by statistical projection, so unless the reporting base widens the model is inferring most of the country from the same thin foundation. A headline number without a stated coverage share is also exactly what becomes the next 1.3 million, because the caveats travel less well than the digit.

(b) alone can produce a wrong number confidently. Reporting is not the same as reporting correctly, and the audit record shows counties reporting no active guardianships and counties reporting cases that should have been closed. Neither the size nor the net direction of that error has been published, so a count assembled from state returns could be above or below the truth with no way to tell which. Recurrence matters too — a single national survey answers the question once and then decays, which is how 2008 data came to stand for eighteen years.

(c) alone counts attributes on an unbounded population. Demographic breakdowns attached to a denominator that is itself wrong describe a group nobody has correctly delimited. The voluntary national data standard also leaves adoption and element selection to each court, and for guardianship cases it recommends only basic elements such as dates of birth and death and residential address, while sexual orientation is not an element of the standard at all. The three have to be read together, and every one of them has to state its unit — cases or persons — because the most-cited figures in this field do not.

What is it connected to?

Fills with researchguardianship abuse and its own unmeasured extent, supported decision-making as a legal alternative, federal representative payee and fiduciary programmes as parallel substitute-decision arrangements, disability rights litigation, and probate court case management modernisation generally. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• How the figure of 1.3 million was computed, and where the USD 50 billion came from at all. Three policy documents print the figure, and the chain behind them ends at a statement given at a 2017 congressional hearing and an unpublished 2018 letter. No source names the states, the year or the population base. The assets figure receives even less scrutiny than the people figure — no document opened here offers any derivation for it, and it drifts between assets under conservatorship, assets under guardianship, and assets guardians control.
• The band, which every retelling drops. The authors of the 2011 estimate wrote that the true value could range from fewer than 1 million to more than 3 million. That sentence appears in the original and in essentially none of the citations. The uncertainty was published; the practice of citation removed it.
• What a right answer would look like. There is no target, no completeness threshold, no acceptable margin of error and no date by which the number should exist. The 2009 resolution says begin collecting, the 2018 report says increase data collection, and the 2026 award says generate a reliable estimate. None of the three is a specification.
• What the federal auditor thinks should be done about it. GAO documented the absence in 2004 and again in 2016. The 2016 report states in its own text that it makes no recommendations. The 2004 report recommended interagency collaboration on abuse data, which one agency declined on privacy grounds, and recommended nothing about enumerating people under guardianship. Twice the finding was that the number is unknown, and twice it was addressed to nobody.
• How many states can count today. The last published figure of that kind rests on 2008 data — fourteen jurisdictions for filings and four for active cases. Eighteen years later, no source opened here says whether that is now six or twenty-six. The measure of progress toward the count is itself uncounted.
• Which direction the error runs. One state is described as under-reporting and over-reporting simultaneously, and another partly estimates because about half its courts combine minors and adults. No source says which distortion dominates, so it is not possible to say whether the true figure sits above or below any published estimate.
• Who owns the demographic gap. Every source records that no age, race, ethnicity, sex, gender identity, sexual orientation, immigration status, language or disability data exists. None names the body that would be responsible for producing it, and none explains why the voluntary data standard recommends only basic identifying elements for these particular cases while defining richer party attributes elsewhere.
• Whether any given figure counts cases or people. The two most-cited numbers use different units, and one 2018 report uses both units for the same figure in two different chapters from two different footnotes. One state reports guardianship-only, conservatorship-only and combined as three separate categories. The conversion factor between case records and adults is unpublished and its sign is undetermined, and no document opened here notices the shift, let alone corrects for it.

See the evidence

Five primary documents were read end to end. Two GAO reports, the Senate committee report, Public Law 115-70 and the text of S. 1148 were opened as published files and read directly, as were the 2011 court research paper carrying the only method-published estimate and the June 2024 data-elements guide. The arithmetic in block 5 — the sum of 77,339, the implied jurisdiction populations, the coverage share near 5 percent and the Ohio share near 73.6 percent — is derived here from the two tables printed in that 2011 paper, and no source opened here performs it. Two evidence rows have no URL because the underlying documents could not be retrieved, and both of them sit on the chain of custody of the field most-cited figure, which is why they are kept as rows rather than dropped. Several further sources were attempted and failed on certificate or access errors, including a 2010 survey by the same court research organization, the recommendations of the fourth national guardianship summit, two state court reporting programmes and the status of a successor bill in the 119th Congress; none of those is asserted here. Where sources disagree the disagreement is left visible rather than resolved. The figure of 1.3 million and the figure of 1,519,284 are never placed on the same page by any source, and the disagreement runs in the counter-intuitive direction — the broader category, guardianship or conservatorship together, yields the smaller number, while the estimate that excludes conservatorships yields the larger one. The unit shifts inside a single federal report, from cases in the body to adults in chapter 3. This is a Path A output, so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range 1,000,000–3,000,000 As of 2008

Sensitivity The unit is active pending guardianship case records, not adults, and the two cannot be reconciled from any source opened here. In states that separate guardianship of the person from conservatorship of the estate, one adult can appear as two case records; in states that count guardianship alone, an adult who has only a conservator appears as none; one audited state reports guardianship-only, conservatorship-only and combined as three separate categories. No document publishes a conversion factor between cases and persons, and the sign of the correction is undetermined — double counting pushes the person figure below the case figure while excluded conservatorship-only adults push it above. The base year is 2008, the estimate was published in 2011, and no source opened here supersedes it as of 2026-08-08; the band is deliberately not re-based to the present adult population, because doing so would assert that the 2008 rate still holds and that rate is precisely the unmeasured quantity. The width of the band is not sampling error around a measurement but the spread that follows from projecting one mid-sized state caseload rate onto a country, and the measured floor beneath it is 77,339 cases, about 5 percent of the point estimate. What this band cannot count at all: adult conservatorship-only cases, excluded from the base; cases open on paper but effectively over, including cases where the person has died, which one state reports alongside counties reporting no active guardianships, so the reporting error runs in both directions with no published relative size; minors continuing under guardianship into adulthood in states filing adult and juvenile cases under one type; and adults under non-court substitute arrangements such as a federal representative payee, of which the 2004 federal audit counted 717,623 elderly beneficiaries, 46,449 veterans affairs fiduciaries and 5,161 office of personnel management arrangements beside a court box reading number unknown. Nothing at any level records who these adults are — no age, race, ethnicity, sex, gender identity, sexual orientation, immigration status, language, disability type or reason for the guardianship. A separate figure of 1.3 million circulates far more widely than either bound here; it is not used in this chain because its derivation has never been published, its source page returns a 404, and it is printed in two different units within a single federal report.

Regional breakdown State-level values cannot be confirmed because the absence of state-level counts is the problem itself. Only four jurisdictions could report active pending adult guardianship cases for the base year, and a federal audit of six states found that none could give exact counts. Splitting a national band by state population would be proportional allocation, and it would be doubly wrong here — the observed rates differ by a factor of about four between the reporting jurisdictions, and the differences are unexplained in the source, so there is no basis for assuming any state resembles the national figure.