OSINT · Publicly Available Sources INVESTIGATION ORACLE

The 47 Percent: How States Lose Foster Children on Paper, and Why No One Has Ever Lost a Dollar Over It

September 09, 2026 OPUS · Claude Opus Project Milk Carton

The 47 Percent: How States Lose Foster Children on Paper, and Why No One Has Ever Lost a Dollar Over It

Federal law has required every state child welfare agency, since September 2016, to report a child missing from foster care to police for entry into the FBI's National Crime Information Center and to the National Center for Missing & Exploited Children — "immediately, and in no case later than 24 hours." A Health and Human Services Inspector General audit found that states failed that requirement for an estimated 51,115 of 74,353 missing-child episodes over a 30-month window: 47 percent were never reported to NCMEC at all, and another 22 percent arrived late. A companion audit found the same states told police about 86 percent of those same episodes — meaning the failure is not that agencies don't know a child is gone. They know. They just don't complete the report that would put the child's face into the national recovery system. Nine years and four federal audits later, HHS has answered with guidance documents and webinars, and has never withheld a single Title IV-E dollar from any state over a child who was never reported missing.

What the law actually requires — and what it doesn't

The Preventing Sex Trafficking and Strengthening Families Act (P.L. 113-183, signed September 29, 2014) rewrote Section 471 of the Social Security Act to impose four distinct duties on any state drawing federal foster care money.

Under 42 U.S.C. § 671(a)(35)(A), a state must develop protocols to expeditiously locate any child missing from foster care; to determine the primary factors that contributed to the child running away; to determine the child's experiences while absent, "including screening the child to determine if the child is a possible sex trafficking victim"; and to report that information as the Secretary requires. Under § 671(a)(35)(B), the state must report the missing child within 24 hours to law enforcement for NCIC entry and to NCMEC, with a photo and description. Under § 671(a)(34)(A), an identified trafficking victim goes to law enforcement within 24 hours. And under § 671(a)(34)(B) — the provision that produces the annual victim counts — each state must report to the HHS Secretary "the total number of children and youth who are sex trafficking victims."

The protocol duty took effect within one year of enactment; the 24-hour reporting duty by September 29, 2016. ACF issued implementing guidance in Program Instruction ACYF-CB-PI-15-07 on June 26, 2015. Congress tightened it again in January 2023, when Section 137 of P.L. 117-348 added an affirmative duty to maintain regular communication with law enforcement and NCMEC until the child is recovered.

What the statute does not contain is a consequence. There is no per-episode disallowance, no automatic offset, no certification requirement. The only enforcement machinery that touches § 671(a)(35) is the Child and Family Services Review — and the CFSR is structurally incapable of enforcing it. Under 45 C.F.R. § 1355.36, a state found out of substantial conformity gets a Program Improvement Plan first; funds are withheld only if the PIP fails, and the maximum withholding reaches 42 percent of Title IV-B and IV-E funds only on a third or subsequent review. In 25 years of CFSRs, no state has ever been found in substantial conformity across all seven outcomes and all seven systemic factors. Every state is on a PIP. When universal failure is the baseline, failure carries no information — and no penalty.

The audits: four federal findings, one federal response

The Inspector General has now measured this four separate ways, and the numbers are not ambiguous.

OIG A-07-20-06095 (2022) — the national snapshot — counted 110,446 missing-children episodes across state child welfare agencies between July 1, 2018 and December 31, 2020. That is roughly 44,000 episodes a year.

OIG A-07-21-06102 (March 2023) narrowed to the 74,353 episodes in which a child was gone two calendar days or longer, drew a stratified random sample of 100 episodes across 38 state agencies, and checked them against NCMEC's own database. Thirty-three were reported on time. Forty-five were never reported to NCMEC — the agencies had no documentation of a report, and NCMEC staff could not locate the episode in their system. Twenty-two were late. Projected: an estimated 34,869 episodes (47 percent) never reported, and 16,246 (22 percent) reported late — 69 percent noncompliance. The stated cause was not resource scarcity or ambiguity: "State agencies generally lacked adequate systems to readily identify whether or not they had reported missing children episodes to NCMEC accurately and in a timely manner." They didn't know, because nothing made them know.

OIG A-07-21-06104 (May 2023) ran the identical test against the NCIC leg of the same sentence in the same statute. There, 86 of 100 episodes were reported timely, 8 were late, and 6 were never reported — an estimated 8,324 children never entered into NCIC. The gap between 86 percent and 33 percent is the single most diagnostic number in this investigation. Caseworkers call the police. Calling the police is local, familiar, and generates a report number a supervisor can point to. Filing with NCMEC is what triggers the national poster, the analyst review, the child-sex-trafficking recovery team, and the permanent federal record — and that is the step that gets skipped.

OIG OEI-07-19-00371 (June 2022) tested the screening duty. It selected the five states reporting the largest number of children in runaway status in FY2018 that also required trafficking screening, and pulled 413 case files. In 268 of them — 65 percent — there was no evidence the child was ever screened after coming back. Boys fared worse than girls: 72 percent unscreened versus 59 percent. Among the screenings that did happen, a third recorded no conclusion at all about whether the child had been trafficked. Documented screening rates were 15 percent in Pennsylvania, 18 percent in Massachusetts, 19 percent in Minnesota, and 32 percent in Illinois. Texas hit 83 percent — proof the number is a policy choice, not a ceiling.

ACF concurred with every recommendation in all four reports. Its remedy was a new Program Instruction on December 1, 2022 and three webinars. OIG closed the NCIC recommendation as "implemented" on August 23, 2024 — closed because ACF delivered technical assistance, not because states started reporting.

The arithmetic that doesn't reconcile

NCMEC received 23,348 reports of children missing from foster care or state custody in 2025 and 23,160 in 2024. Those 23,348 reports made up roughly 73 percent of the 32,167 missing-child cases NCMEC worked in 2025 — foster youth are not a subset of America's missing children problem, they are the overwhelming majority of it. NCMEC identified 17 percent of the children missing from care as likely victims of child sex trafficking — about 3,969 children in 2025 alone.

Now set that against the states' own episode counts. OIG documented ~44,000 missing-from-care episodes per year in state administrative data. NCMEC's intake is roughly half that. The 47-percent figure isn't a statistical artifact; it is exactly the size of the hole between what state systems record and what the national recovery system ever sees.

Then set both against what states tell HHS. Under § 671(a)(34)(B) and parallel NCANDS reporting, 36 states reported a combined 800 unique child sex trafficking victims for 2023. Fifteen jurisdictions reported nothing. Eight hundred, nationally, in a year in which NCMEC independently flagged more than three thousand likely trafficking victims just among foster children who ran. The federal government publishes both numbers and reconciles neither.

The mechanism that makes the discrepancy possible is a data-entry choice. In AFCARS, "runaway"/whereabouts-unknown functions both as a living arrangement and as a reason a foster care episode ends. When an agency closes an episode with a runaway discharge, the child exits the foster care count. No federal field then tracks whether that child was ever found, how long they were gone, or what happened to them while they were. As NCMEC and ACF's own listening-session participants have noted, the label matters operationally too: a child coded "runaway" gets a slower, more discretionary response than a child called "missing." The word does work that the statute never authorized.

The result is the fact at the center of this story: there is no federal database that can say how many American foster children are missing right now, or for how long. Not AFCARS, which counts placements and exits. Not NCIC, which holds only what police were told. Not NCMEC, which holds only what agencies chose to send.

The states, named

Georgia. The Senate Judiciary Subcommittee on Human Rights, under a bipartisan probe led by Sens. Jon Ossoff and Marsha Blackburn, reported on April 9, 2024 that NCMEC's analysis found 1,790 children in Georgia DFCS care reported missing between 2018 and 2022 across roughly 2,400–2,500 episodes, and that 410 of them were likely victims of child sex trafficking — about 23 percent. The subcommittee interviewed more than 100 witnesses and concluded DFCS mismanagement was "a key contributor" to child deaths and serious injuries. DFCS's public response was to dispute the 410 figure as "based on national figures and projections, not evidence specific to any case from Georgia" — a defense that only works if the state's own case-specific evidence is better, which is precisely what the state could not produce.

Connecticut. The Auditors of Public Accounts reported in June 2025 that more than 600 children under 18 accounted for 3,736 separate missing-from-care incidents in fiscal years 2021 through 2023, a 42 percent rise across the period. In 94 percent of the cases auditors examined closely, DCF had no documentation that the returning child was screened for sex trafficking. One child went missing more than 100 times. One child had been missing for over two years and was still missing as of April 4, 2025. The audit produced 17 findings; DCF could not produce formal assessments, plans, or goals for managing children who go missing. Legislators held hearings on it in August 2025.

Illinois. An Auditor General performance follow-up issued July 29, 2025 found that in 9 of 60 sampled episodes (15 percent), NCMEC was not notified within the state's own three-hour standard, and in another 9 of 60, the parents, court, or guardian ad litem were not notified in time. A separate DCFS compliance audit covering July 2022–June 2024 returned 34 findings, 29 of them repeats. DCFS reported 193 youth in care missing during 2025.

Maryland. In response to a Public Information Act request, the Department of Human Services disclosed in October 2025 that 990 children in state foster care went missing between January 1, 2020 and August 17, 2025. It has not disclosed how many were never recovered, how many are missing today, or the state's trafficking-involvement rate. Reporters have spent months on the same unanswered question. Meanwhile, Silver Oak Academy — a privately operated group facility in Carroll County — generated more than 100 emergency calls to the sheriff's office beginning in January 2025. A 16-year-old foster child went missing from Silver Oak for more than nine days in May 2026; deputies eventually found him 22 miles away in a Frederick home associated with a former Silver Oak employee. Carroll County Sheriff Jim DeWees described an escalating pattern of instability. Maryland declined to renew Silver Oak's license and moved to remove foster children from it — after the fact, and after the child was gone.

The money, and who it flows to

Federal Title IV-E foster care support is a mandatory, open-ended, permanent entitlement estimated at $9.7 billion for FY2024, inside roughly $11.0 billion in total federal child welfare spending. Open-ended means the federal match follows the placement. It does not follow the child's safety, and it does not stop when the child stops being in the placement it is paying for.

That is the incentive structure in one sentence. In congregate care, states pay per-diem rates that Senate investigators documented at $275 to more than $800 per child per day for a single large for-profit operator, Sequel Youth & Family Services, which served more than 9,000 clients across a network of 35 residential centers before a private equity firm took a controlling stake in 2017; 17 of those centers subsequently closed under scrutiny. The Senate Finance Committee's June 12, 2024 report, Warehouses of Neglect, examined four operators — Universal Health Services, Acadia Healthcare, Devereux Advanced Behavioral Health, and Vivant Behavioral Healthcare — and found that 34,000 youth were placed in residential treatment facilities through state foster care systems in 2022 alone, many of them for lack of community capacity rather than clinical need. On October 9, 2024, Chairman Ron Wyden referred those operators to the Justice Department over potential Medicaid fraud and asked whether states were violating children's civil rights through institutional placement.

The conflict is structural. A per-diem bed generates revenue on census. A missing-child report is a licensing exposure, a contract-renewal risk, and, in a facility already under scrutiny, a headline. The party in the best position to know within minutes that a child is gone — a private provider, often operating a child placed from another state — is the party with the strongest financial reason to wait, to call it an "unauthorized absence," to see if the kid comes back before anyone has to write it down. Nothing in § 671(a)(35) reaches a private contractor directly; the duty runs to the Title IV-E agency, which learns what the facility tells it, when the facility tells it.

On the recovery side of the ledger, the federal government funds NCMEC — the entity that never hears about half these children — at $41.4 million (FY2024), $41.0 million (FY2025), and $43.0 million (FY2026) through OJJDP grants, roughly $641 million cumulatively across federal awards. NCMEC has no authority to compel a single state to file a report. It is a mandated destination with no subpoena power.

And states are not even spending the money aimed at the youth most likely to run. GAO reported in January 2025 that grantees left roughly 5 percent of Chafee Foster Care Program funding — more than $9 million a year — unspent in each of FY2019–FY2022; in FY2022, 12 states returned Chafee funds and 28 returned education voucher funds, about $8.9 million, and about $670,000 of FY2020 money went back to the Treasury outright. West Virginia sent millions back. HHS's Office on Trafficking in Persons runs its own grant streams with no required cross-reference against state child-welfare exit data — meaning federal trafficking money and federal foster-care runaway data have never been laid on the same table.

Who is supposed to be watching

ACF's Children's Bureau approves every Title IV-E state plan, including the § 671(a)(35) protocols. It has approved all of them. It has never disapproved a plan, never invoked 45 C.F.R. § 1356.50 withholding, and never assessed a penalty over a missing-child reporting failure. Its documented response to a 47-percent never-reported rate was one Program Instruction and three webinars.

Congress came closest in the 118th. The Find and Protect Foster Youth Act (H.R. 2426 / S. 1146) would have required HHS to identify obstacles to responding to reports of children missing from foster care, evaluate state and tribal protocols, deliver targeted technical assistance, and commission a GAO study on foster youth who run. Both chambers passed it on September 18, 2024 — in non-identical forms that were never reconciled. It died in the gap between two versions of the same bill. Not one bill in twelve years has proposed attaching a Title IV-E funding consequence to § 671(a)(35).

State law fills the vacuum unevenly. Only 11 states plus the District of Columbia — Colorado, Florida, Kansas, Kentucky, Maryland, Minnesota, Nebraska, Oklahoma, Tennessee, Texas, and Utah — statutorily mandate that child welfare agencies screen children who may have been trafficked. Only three — Minnesota, Texas, and Utah — specifically mandate screening children who went missing from foster care. Every state is bound by the federal duty; almost none has written it into enforceable state law, and the federal statute never defines what a screening must contain.

Why it matters, and what would actually fix it

The Justice Department and multiple corroborating studies put child-welfare-system involvement in the histories of 60 percent or more of identified child sex trafficking victims. NCMEC's own 2025 intake puts the trafficking rate among foster children missing from care at 17 percent. A child who runs from a placement is, statistically, one of the most predictably targeted people in the country — and the federal response to that child depends on whether an overloaded caseworker completed a second report after completing the first.

Five changes would close the gap, and none require new money:

Force the reconciliation. Require every state to report, in AFCARS, a NCMEC case number and an NCIC NIC number for every runaway/whereabouts-unknown episode. An episode without both is a finding. This converts the OIG's 47-percent estimate from a statistic into a list of names.
Attach a real penalty. A per-episode disallowance of Title IV-E administrative claims for each unreported episode — not a Program Improvement Plan. Under the current CFSR structure, universal failure means no failure.
Reach the provider. Make 24-hour notification a condition of licensure and of every state contract with a group home, residential treatment facility, or out-of-state placement, with license revocation as the remedy — so the party that knows first is the party that answers for the delay.
Publish the count. A public federal dashboard of children currently missing from foster care, by state and by duration. Maryland's refusal to answer that question for seven months is only possible because no one is required to.
Define the screening. A validated instrument, a mandatory documented conclusion, and a required determination of where the child was and who they were with — the part of § 671(a)(35)(A) that agencies skip most reliably, since a third of completed screenings currently reach no conclusion at all.

The children in the 47 percent are not statistical. Each one is a specific kid who left a specific bed on a specific night, whose absence someone noticed within hours, and whose name was never sent to the one national system built to find them. The law told states to make that call within twenty-four hours. Nine years, four federal audits, one dead bill, and $9.7 billion a year later, nothing happens to a state that doesn't.

Sources: