· @Mike Tikkanen
Two families
My first visit to the four-year old was at the suicide ward of a local hospital. She had spent her whole life watching her seven-year-old sister be raped and beaten. Her sister’s case file was more than a foot thick. Whether a child watches a sibling being raped and beaten or experiences it the life changing trauma is the same. https://calio.dspacedirect.org/handle/11212/1719
Over the next 12 years that family and my caseload grew from four children to seven. The siblings could not be placed in foster care together because they were sexualized. Several foster families had their own children molested by the children in my care. When I was taken off the case, I had every reason to believe the mother would have more children, and the abusing parent stayed in the home and was never made a party to the case. I had no way to find out. Nobody did.
My second case was eight children, all under 18, all sexually abused over many years. Two Families.
How the official count works
The federal Child Maltreatment report counts a child as a victim only when a caseworker confirms abuse of that child. In 2024 that was 532,228 children. It reports roughly 7.8 million children named in child protection referrals each year. That figure is not a count. It is 4.4 million referrals multiplied by an assumed 1.8 children per referral, and more than half of those referrals are screened out without any child-level record being created.
Practitioners who have worked these cases see households of four, six, and eight children where the report names one. Correct for household size alone and the real figure is 12 to 16 million children a year living in homes that child protection has been told about. The Safe Passage for Children of MN investigative report exemplifies how severe the crisis must be to be counted as abused and removed from their homes in CPS today.
The four-year-old I visited in the emergency psych ward of the hospital may never have been counted. The abuse was “happening to her sister.” All she did was watch.
Watching is abuse
A child in a toxic home carries the same trauma as the child being beaten or raped. Some states already say so in law: some include a child’s witnessing of domestic violence as a form of abuse or neglect.
The national count doesn’t work that way. It counts the children named in a call, not every child living in that home. That is the first undercount.
Calls get screened out
In 2013, Minnesota counties received nearly 68,000 reports of child abuse or neglect and closed most of them without ever sending anyone out. At least four counties screened out about 90% of calls. Screeners weren’t allowed to look at earlier screened-out reports on the same child.
Eric Dean was reported 15 times. Social workers came to the house once, offered services to the family but did not see the child.
It’s getting worse nationally, not better. In 2017, states screened in 57.6% of calls. By 2024 it was 47.1%. Texas and Arkansas banned anonymous reports, and calls dropped. Pennsylvania sends most of its calls through a separate track that never appears in the federal numbers. Every one of those changes makes the official count smaller. None of them makes a child safer.
One-year snapshots, no long-term tracking
The numbers are one-year snapshots. When researchers looked across a whole childhood, they found that 37.4% of all children experience a child protective services investigation by age 18. That’s more than one in three American children. The annual numbers hide this.
Nobody follows families over time. Even that 37.4% figure had to be estimated from single-year data, because the records that would show one family moving in and out of the system for 10 or 20 years mostly don’t exist. My first family was on my caseload for 12 years. In the national data, it shows up as a set of separate, unconnected calls.
Generational abuse and politics
Generational abuse makes all of this worse. Children who grow up with sexual abuse, drugs, and violence have a hard time unlearning them, and many go on to raise families like the ones they came from. Those families are often larger than average. Hennepin County figures once showed 4 children per family in open cases, from Gerard Bodell at CASAMN.
The systems meant to catch these children depend on budgets and politics. The federal government’s top child welfare official promised a “bonfire” of deregulation. Texas rewrote its neglect law so that a parent’s neglect has to show “blatant disregard” before the state can act. There is a growing push nationally to call neglect “only poverty” Neglect is where most of these children are found and most of them die. Narrow it and they disappear from the count.
Few states measure the outcomes that matter most for these children, such as whether they are safer a year later or whether their own children end up in the system.
What would change it
- Count every child living in a reported home, not only the children named in the call.
- Link records so a family can be followed across years and counties.
- Make states report on child outcomes, not only case counts.
7.8 million is not the size of this problem. It’s the floor. Most of the children we are talking about never see a social worker, and nobody is counting them.
Sources and open items
Sources:
- U.S. Children’s Bureau, Child Maltreatment 2024 (532,228 victims): Imprint summary, report list
- Child Maltreatment 2023 (7.78 million children in referrals): CWLA summary
- Child Maltreatment 2010 (1.81 children per referral): handout
- State laws on witnessing domestic violence: CASAT OnDemand
- Minnesota screen-outs, Star Tribune 04/21/2014: summary; four counties near 90%: PACEs Connection list of Stahl articles
- Eric Dean, one report investigated: MN Council on Developmental Disabilities, Duluth News Tribune
- National screen-in trend, Texas and Arkansas anonymous-report bans, Pennsylvania, Texas neglect law: Child Welfare Monitor, 02/17/2026
- 37.4% investigated by age 18: Kim et al., American Journal of Public Health, 2017
- “Bonfire” of deregulation: The Imprint
- “Only poverty” framing (America First Policy Institute): AEI
The federal report states the method plainly: the national estimate of total referrals is the sum of the actual reported number of screened-in referrals and the number of estimated screened-out referrals. acf
For FY 2023:
- Screened-in referrals: 2,107,473 – an actual count cwla
- Screened-out referrals: a national estimate of 2,292,000 acf
- Total: 4,399,473, published as 4,399,000
So 52% of the headline number is an estimate, not a count.
How the estimated half is produced. The national estimate of screened-out referrals is based on the rate of referrals from reporting states multiplied by the national population of all 52 jurisdictions. Take the states that do report screened-out calls, compute referrals per 1,000 children, apply that rate to every child in the country. 46 of 52 jurisdictions reported both screened-in and screened-out referrals in 2023. The other six are filled in by assumption. acfcwla
Why the two halves aren’t the same kind of data. Screened-in referral data comes from the Child File. Screened-out data comes from the Agency File. The Child File contains detailed case information about each child who is the subject of an investigation or assessment. The Agency File is state-level summary data – counts of children screened out, CPS staffing, preservation services, and the like. Aggregate totals, no individual records. Child Maltreatment 2022 h U.S. Department of Health & Human Services +2
That’s the structural point for our argument. For screened-in cases there are child records you could in principle follow over time. For the 2.3 million screened-out referrals there is a tally and nothing else. No child identifiers, so no way to link a family’s third call to its first, which is exactly the longitudinal tracking that doesn’t exist. It isn’t that nobody has analyzed it. The data was never collected in a form that permits it.
Two more things worth knowing. NCANDS is voluntary, and every state has a different process and threshold for reporting, so the rate being extrapolated nationally is an average across 46 incompatible definitions. And the child-level figure stacks on top of all this – the 7.78 million comes from multiplying this 4.4 million by an assumed 1.8 children per referral. mass
So the chain is: counted screened-in referrals, plus a rate-extrapolated screened-out estimate, times an assumed family size. One real number and two assumptions.
KIDS AT RISK ACTION / KARA / INVISIBLE CHILDREN
