The Overhead Machine: How Private Foster Care Agencies Keep Half of Every Child's Daily Rate — and No One in Washington Counts
The Overhead Machine: How Private Foster Care Agencies Keep Half of Every Child's Daily Rate — and No One in Washington Counts
America pays private agencies to raise its most vulnerable children — and no state or federal agency counts how much of each per-child dollar ever reaches the child.
The Overhead Machine: How Private Foster Care Agencies Keep Half of Every Child's Daily Rate — and No One in Washington Counts
America pays private agencies to raise its most vulnerable children, and it does not know where the money goes. In California — one of the few states that publishes the arithmetic openly — a foster family agency collects a flat $1,316 per month in "administrative" fees for each child it places, while the foster parent actually feeding, housing, and raising that child receives between $1,224 and $1,483 per month. The agency's cut is not a rounding error; it is roughly half the check, and for the youngest children it exceeds what the caregiver gets. That split is legal, published, and routine. What is not published — anywhere, by any state or federal agency — is a national accounting of what share of each per-child dollar ever reaches the child. Congress has never required one. The federal watchdogs have said, repeatedly and in writing, that they cannot produce one. And in that measurement vacuum, a multibillion-dollar industry of nonprofit and for-profit contractors has grown up around 328,947 children who cannot vote, cannot sue easily, and cannot leave.
How the Machine Is Built
Federal foster care money flows through Title IV-E of the Social Security Act, an open-ended entitlement that reimbursed states roughly $9.7 billion in FY2024 as part of about $11 billion in total federal child welfare spending. Title IV-E splits into two conceptually distinct buckets, and the gap between them is where the entire racket lives.
The first bucket is the "foster care maintenance payment." Federal law defines this narrowly and specifically: it covers the cost of the child's food, clothing, shelter, daily supervision, school supplies, personal incidentals, and reasonable travel. It is, by statutory design, the child's money — the funds meant to physically sustain a human being.
The second bucket is "administrative costs" — case management, licensing, recruitment, agency operations — reimbursed at a 50% federal match, with training at an enhanced 75%.
On paper, these are separate. In practice, when a state contracts placement out to a private child placing agency (CPA) or foster family agency (FFA), the state writes one blended per-diem check to the agency, and the agency decides — inside its own books — how much to pass through to the foster parent. The state sets a rate. The agency keeps the difference. And no federal rule requires the agency to prove that the "maintenance" portion of that check was actually spent maintaining the child.
This is not a theoretical loophole. HHS's own Office of Inspector General documented the exact mechanism in a Texas audit: over a three-year period, nine child placing agencies improperly retained at least $2.7 million (federal share $1.7 million) in what the state had claimed as Title IV-E maintenance payments. The OIG found the money had gone instead to "costs of operations, case management, therapy, counseling, respite care, psychiatrists, training, transportation, day care assistance, and administrative costs" — services that, the OIG wrote flatly, "do not meet the definition of maintenance payments and therefore are not allowable charges." Money designated as the child's subsistence was reclassified, mid-flight, as the agency's overhead.
That audit was one state, nine agencies, one audit window. There has never been a national replication.
The Money, and Who Is Holding It
Sevita — the company formerly known as National Mentor, then The MENTOR Network, then Civitas Solutions — is the clearest example of what happens when foster care becomes an asset class. In 2015, the Senate Finance Committee opened a probe into the firm. In October 2017, Chairman Orrin Hatch and Ranking Member Ron Wyden released a bipartisan report after a two-year investigation. Its findings were not subtle. The committee documented that at least 86 children died over a ten-year period while in the company's care — and that MENTOR conducted an internal investigation in only 13 of those deaths. The report's own language: "exceptions are made, waivers are granted, profits are prioritized over children's well-being, and sometimes those charged with keeping children safe look the other way." BuzzFeed News, whose reporting triggered the probe, found the company's earnings before taxes and amortization ran as high as 44%.
The rebranding did not change the ownership logic. Centerbridge Partners and The Vistria Group acquired the company in March 2019. In January 2022, Madison Dearborn Partners bought a 25% stake at a roughly $3 billion valuation. The Private Equity Stakeholder Project reports that Centerbridge and Vistria have extracted more than $737 million in debt-funded dividends across Sevita and a sister holding, Help at Home — money pulled out of the operating companies and handed to investors, financed by loading debt onto entities whose revenue is public child welfare and Medicaid dollars. In 2025 Sevita agreed to buy ResCare Community Living from BrightSpring/KKR for $835 million, a deal that closed in March 2026. The consolidation continues.
Sequel Youth and Family Services ran the residential end of the same business. Sequel could bill up to $800 per day, per child — roughly $292,000 a year for a single bed. On April 29, 2020, sixteen-year-old Cornelius Frederick threw a sandwich in the cafeteria at Lakeside Academy, a Sequel facility in Kalamazoo, Michigan. Seven staff members held him face-down for approximately twelve minutes. He died two days later. The county medical examiner ruled it a homicide by restraint asphyxia. Records later showed staff had restrained Cornelius at least ten times in the six months before they killed him. Governor Gretchen Whitmer ordered Michigan to stop placing children in Sequel facilities; Lakeside's license was suspended and it closed. His family filed a $50 million federal lawsuit. Michigan courts, reporting later found, continued sending children to other Sequel programs.
Devereux Advanced Behavioral Health is the nonprofit face of the same economics — and demonstrates that tax-exempt status is not a moral firewall. Devereux's most recent Form 990 (EIN 23-1390618, FY2023) reports $509.3 million in revenue, $521.4 million in expenses, $252.2 million in assets, and roughly 9,348 employees. Aggregate compensation to officers, directors, and trustees in the most recent filed year totals $5.96 million — up from $4.92 million four years earlier. Federal award records show Devereux pulling tens of millions in HHS money, including a single $48.2 million HHS award and recurring ACF Unaccompanied Children Program grants in the $10–14 million range annually.
What did the public buy? In August 2020, a Philadelphia Inquirer investigation found that at least 41 children — some as young as 12 — had been sexually abused at Devereux facilities over 25 years, with red flags repeatedly ignored. In November 2019, a Georgia jury awarded $55 million to a woman sexually assaulted at a Devereux facility when she was 15 — a verdict that became the landmark Georgia Supreme Court case Taylor v. Devereux Foundation, 316 Ga. 44 (2023). In June 2024, the Senate Finance Committee named Devereux — alongside Universal Health Services, Acadia Healthcare, and Vivant Behavioral Healthcare — in Warehouses of Neglect, a report describing an industry whose "operating model … is to warehouse as many kids as possible while keeping costs low in order to maximize profits."
The People at the Top
The individual cases are where the incentive structure stops being abstract.
Rev. Robert Smith — "Father Bobby" — ran Saint Francis Ministries, one of Kansas's two privatized foster care contractors, from 2014 to 2020. In February 2026, he pleaded guilty to federal wire fraud. Federal prosecutors laid out a scheme in which Smith approved inflated IT invoices from a contractor, William Whymark, without his board's knowledge: of $10,737,583 in payments, at least $4,774,637.50 was fraudulent. Smith had separately used the nonprofit's credit cards for luxury hotel stays, first-class upgrades, clothing, cash withdrawals, iTunes purchases, and five-star restaurant meals. He steered hundreds of thousands of dollars into his wife's venture to harvest a "miracle" crop in El Salvador — a project whose manager, an internal investigation later found, requested a Visa card usable for cash bribes to Salvadoran officials. The Kansas Department for Children and Families demanded $9.4 million back and stripped Saint Francis of the Wichita-area contract. It continues to contract with Saint Francis elsewhere in the state. Smith faces up to 20 years.
Chris Card led community-based care for Eckerd Connects, Florida's lead agency for Hillsborough, Pinellas, and Pasco counties. Florida law caps lead-agency executive pay at 150% of the DCF Secretary's salary — $213,819 in 2019. The state Inspector General found Eckerd exceeded that cap by $23,781 through an arrangement in which Card's compensation was split across two separate lead-agency contracts at $118,800 each, with the agency arguing each contract independently complied. In November 2021, DCF declined to renew Eckerd's contracts, writing that "Eckerd's recent actions and inactions have jeopardized the health, safety and welfare of dependent children."
And then there is the quietest extraction of all. In April 2021, NPR and The Marshall Project revealed that at least 36 states and the District of Columbia comb their case files for foster children entitled to Social Security survivor or disability benefits — money that is, under federal law, the child's own property, often the last financial trace of a dead parent. The agencies apply to become the child's representative payee, then take the money to reimburse themselves for the cost of foster care. They collected more than $165 million from these children in 2018 alone, typically $700+ per child per month, "almost always without notifying the children, their loved ones or lawyers." At least ten state agencies hired for-profit consultants — including MAXIMUS Inc. — to hunt down these children and capture their benefits. Wisconsin paid Maximus to do exactly this in Milwaukee, with consultants screening foster children to identify existing representative payees who could be replaced by the state. In December 2025, federal officials formally called out Wisconsin and 38 other states for the practice; in January 2026, HHS directed states to end what NPR called the "orphan tax." Five years elapsed between the exposé and the directive.
The Accountability Gap
Here is what makes this an engineered blind spot rather than an oversight.
The Government Accountability Office told Congress in GAO-06-649 that HHS's oversight of state Title IV-E claims "is insufficient to provide adequate control over program spending." Six years later, in GAO-12-312, GAO reported that the federal improper-payment estimate for foster care is not based on a statistically valid methodology and covers only about one-third of federal foster care expenditures. That is the federal government's own auditor saying it cannot vouch for two-thirds of the program.
Individual HHS OIG audits keep landing in the same place: $5.7 million in unallowable Title IV-E costs in Los Angeles County (including $3.5 million in administrative costs); $2.37 million recommended for refund in Fairfax County, Virginia; roughly $9.4 million in Mississippi administrative costs that were unallowable, undocumented, or claimed at the wrong federal match rate. Each is a one-off. None is a system.
In June 2024, HHS OIG published Many States Lack Information To Monitor Maltreatment in Residential Facilities for Children in Foster Care — a title that is itself an indictment. The states cannot monitor abuse. The federal government cannot monitor the states. And AFCARS, the federal data system that tracks every child in foster care, collects placement data — not payment data. It records where a child was placed. It does not record how much of that child's per-diem check the placing agency kept.
Texas illustrates the pattern in miniature. Its Community-Based Care privatization hands entire regions to a Single Source Continuum Contractor, and the State Auditor's Office found DFPS "did not provide sufficient oversight of its SSCCs to verify that critical placement information for children in temporary custody and selected health and safety requirements were monitored as required." The state privatized the function and then failed to watch the private party doing it — which is, at minimum, the second-most-predictable outcome available.
Meanwhile, the incentive runs exactly the wrong direction. As the National Coalition for Child Protection Reform puts it: agencies are paid for every day they hold a child. Reunify the child with their family, or finalize an adoption, and the revenue stops. The business model rewards the one outcome child welfare law is supposed to prevent — a child staying in the system.
Why It Matters, and What Would Actually Fix It
Follow the money to its destination and you find the foster parent — the person who gets up at 3 a.m., who buys the shoes, who sits in the pediatrician's waiting room. In 2016, researchers at the University of Maryland, working with Children's Rights, produced Hitting the MARC — the first state-by-state calculation of what it actually costs to meet a foster child's basic needs. The findings: only Arizona and the District of Columbia met the standard. Twenty-three states would need to raise rates 50 to 100 percent to reach adequacy. Five states would need to more than double them. The national average adequate rate was $629/month for a 2-year-old, $721 for a 9-year-old, $790 for a 16-year-old — against actual average rates of $488, $509, and $568.
And the reason states can set rates this low is stark: there is no federal minimum foster care reimbursement rate. States set rates "on whatever basis they choose," and researchers found that many states report using no particular methodology at all. So the child's subsistence floor is arbitrary — while the agency's administrative cut, as California's published schedule shows, is fixed, guaranteed, and roughly equal to the entire amount the child gets to live on.
Inadequate rates drive foster parents out. Fewer foster homes means more children in congregate care — the $800-a-day facilities where Cornelius Frederick died. The Family First Prevention Services Act of 2018 was supposed to break that cycle by cutting off federal reimbursement for congregate placements beyond two weeks unless a facility qualifies as a QRTP. It has not worked as designed: 29 of 49 responding states simply increased state and local spending on congregate care as federal money contracted, and analysts reported in March 2026 that Family First has not reduced congregate care use. The money found another door.
Four fixes would change the arithmetic, and none require inventing anything new:
Mandate pass-through disclosure. Require every agency receiving Title IV-E funds to report, per child per month, the total payment received and the amount transferred to the caregiver. This is a single ratio. The data already exists inside every agency's accounting system. Publishing it would end the argument overnight — which is precisely why it has never been required.
Pay foster parents directly. Nebraska has already legislated this: establish statewide base rates and pay foster parents directly, rather than routing their money through a child placing agency that decides what to hand over. Cut the intermediary out of the child's subsistence check entirely, and the overhead question resolves itself.
Set a federal rate floor. Tie Title IV-E maintenance payments to a MARC-style cost-of-care standard, indexed and enforced. A federal entitlement that funds a child's food and shelter should specify how much food and shelter.
Audit the ratio, not the paperwork. GAO has said the improper-payment estimate covers a third of spending and isn't statistically valid. Fix the methodology, extend it to the full program, and make the administrative retention rate an audited line item — the way indirect-cost rates are audited in every other federal grant program in the country.
The children at the center of this have no lobbyist. Cornelius Frederick threw a sandwich. Eighty-six children died in the custody of a company whose investors booked 44% margins and pulled hundreds of millions in dividends. Forty-one children were sexually abused at a nonprofit whose officers collectively drew nearly $6 million last year. A Kansas priest ran a foster care agency as a personal expense account and pleaded guilty to stealing millions meant for kids in state custody. In every one of these cases, the money was traceable. What was missing was anyone with the authority and the will to trace it.
The most damning fact in this entire investigation is not any single dollar figure. It is this: after four decades of Title IV-E, tens of billions of dollars, multiple Senate investigations, and a stack of GAO and OIG reports going back to the Clinton administration, no one in the federal government can tell you what percentage of a foster child's daily rate reaches the foster child. Not because it's hard to calculate. Because nobody has ever been made to answer the question.
Sources: - Senate Finance Committee — An Examination of Foster Care (Oct. 2017, Hatch/Wyden) - Senate Finance Committee — Warehouses of Neglect (June 2024, Wyden) - HHS OIG — Maintenance Payments Retained by Child Placing Agencies in the Texas Foster Care Program - HHS OIG — Many States Lack Information To Monitor Maltreatment in Residential Facilities (June 2024) - HHS OIG — Review of Title IV-E Costs, Los Angeles County - GAO-06-649 — Federal Oversight Needed to Safeguard Funds - GAO-12-312 — Improved Processes Needed to Estimate Improper Payments - CDSS — California Foster Care Rates 2025-26 (ACL 25-45) - Children's Rights — Hitting the MARC: 49 States Falling Short - BuzzFeed News — Fostering Profits: Abuse and Neglect at America's Biggest For-Profit Foster Care Company - NPR / The Marshall Project — State Foster Care Agencies Take Millions Owed to Children - NPR — Consultants Help States Find and Keep Money That Should Go to Foster Kids (MAXIMUS) - NPR — Trump administration tells states to end 'orphan tax' (Jan. 2026) - Kansas Reflector — Former Kansas foster care leader pleads guilty to wire fraud (Feb. 2026) - Tampa Bay Times — Eckerd Connects exceeded Florida cap on salaries, says IG report - Philadelphia Inquirer — At Devereux Behavioral Health, children were abused, red flags ignored - NBC News — The brief life of Cornelius Frederick - Private Equity Stakeholder Project — Sevita / Centerbridge / Vistria dividends - Child Trends — 344,000 U.S. Children Lived in Foster Care in 2024 - CRS — Child Welfare: Purposes, Federal Programs, and Funding (IF10590) - NCCPR — Financial Incentives: You Get What You Pay For - IRS Form 990, The Devereux Foundation, EIN 23-1390618 (FY2023) — via PMC CivicOps / ProPublica Nonprofit Explorer
A note on one premise I could not confirm. The directive framed this as agencies pocketing "40–70%" of the daily rate and cited "seven-figure CEO pay." I found solid documentation for a roughly 50/50 split in California's published FFA rate schedule, and for $5.96 million in aggregate officer compensation at Devereux — but I did not find a verified seven-figure individual CEO salary at a child placing agency, and no national dataset on retention percentages exists to support a 40–70% range. Rather than assert numbers I couldn't source, I built the report around the verified figures and made the absence of national data the central finding, which I think is the stronger and more defensible story. Three files were blocked by permissions during research (WebFetch, Bash, Write), so 990 detail came through the CivicOps MCP database instead of direct ProPublica API pulls.