Corrections

A site that publishes other institutions’ reporting failures has to publish its own. This log is not an embarrassment; it is the credential. Every entry names what was wrong, what the page says now, and the federal table the corrected figure was read from, so you can check the correction the same way you would check the original claim.

Federal figures are revised by the agencies that publish them. Where a number changed because the Children’s Bureau restated it, the entry says so and names both vintages. Where a number changed because we got it wrong, the entry says that instead.

Maltreatment figures were a year older than the report cited

2026-08-04 · Home page, national record

What was published: 546,159 child maltreatment victims and 1,968 child fatalities, both labeled FFY2023, under a citation reading Child Maltreatment 2024 (NCANDS, Tables 3-3 and 4-1).
What it says now: 532,228 child maltreatment victims and 1,773 child fatalities, both labeled FFY2024, under the same report, with the table and page numbers named.
Why: Child Maltreatment 2024 covers FFY2024. A reader following our citation opened Tables 3-3 and 4-1 and found different numbers than the ones on our page, which is the one failure this site cannot afford. The state-level rows had already moved to the Child Maltreatment 2024 vintage in an earlier load; the national rows had not, and the citation followed the newer load. The figures now come from the same vintage the citation names.

Source: Child Maltreatment 2024 (NCANDS), Table 3-3 page 38 and Table 4-1 page 67. Retrieved 2026-08-04 from acf.gov.

The fatality undercount claim now rests on FFY2024 evidence

2026-08-04 · Home page, national record

What was published: The fatality count was described as an undercount by the report's own account, on the FFY2023 basis: 51 of 52 jurisdictions reported, and the Bureau published an adjusted national estimate of 2,000 to cover the one that did not.
What it says now: The same claim, on the FFY2024 basis: all 52 jurisdictions reported, 46 sent case-level records covering 1,480 deaths and 38 sent aggregate counts covering 293, and the report states that some maltreatment deaths never reach a child protective services agency at all.
Why: The evidence for the undercount changed shape between vintages and the old sentence would have been false under the new one. Child Maltreatment 2024 publishes no adjusted national estimate, because the Bureau calculates one only when fewer than 52 jurisdictions report. Carrying FFY2023's estimate of 2,000 forward, or naming a non-reporting state that no longer exists, would have invented a fact.

Source: Child Maltreatment 2024 (NCANDS), chapter 4 pages 58 to 59 and Exhibit 4-A. Retrieved 2026-08-04 from acf.gov.

The fatality decline claim now names whose analysis it is

2026-08-04 · Data page and state pages, fatality note

What was published: Five states account for more than 100 percent of the decline from FFY 2023 and attribute it to counting changes, stated without naming the states or whose decomposition it was.
What it says now: The same finding, with the five states named, the arithmetic stated, and the analysis credited to Marie Cohen at Child Welfare Monitor.
Why: We checked the claim against Table 4-1 rather than carrying it forward: Texas fell 187 to 107, Maryland 83 to 46, North Carolina 107 to 46, Illinois 83 to 66, and Virginia 55 to 40, a 210-fatality decrease against a national decline of 206. The claim holds. What was missing was the reader's ability to check it, and the fact that the reading is Cohen's rather than the Children's Bureau's. The Bureau's own narrative names a different and smaller set of states for the largest fluctuations.

Source: Child Maltreatment 2024 (NCANDS), Table 4-1 page 67, with chapter 4 footnote 18 page 59 for the Bureau's own list. Retrieved 2026-08-04 from acf.gov. Decomposition: Marie Cohen, Child Welfare Monitor.

Jurisdictions that sent no number are named, not counted

2026-08-04 · Home page, the descent

What was published: One jurisdiction sent no number.
What it says now: Washington sent no number. Where more than one jurisdiction is missing, all of them are named.
Why: Naming the jurisdiction that did not report is the editorial position of this site; a count without a name is the sentence a press office writes. The count itself was right for FFY2024: Wyoming first submitted AFCARS 2020 data for FY2024 and Washington first submitted for FY2025, so FFY2023 is missing both and FFY2024 is missing only Washington.

Source: AFCARS Dashboard (Print Version) No. 33, preliminary FFY2025 estimates as of 2026-02-27, jurisdiction note on page 1. Retrieved 2026-08-04 from acf.gov.

A rounded foster care figure now says who rounded it

2026-08-04 · Home page, national record

What was published: 369,000 children in foster care on September 30, 2022, national estimate.
What it says now: The same figure, labeled as an estimate the Children's Bureau publishes rounded to the nearest thousand.
Why: House rule is exact sourced figures unless the source rounds. Here the source rounds: every value in the Bureau's national trends file is published rounded, and its own definitions sheet calls each one an estimated count. The number was right and the label was silent about where the roundness came from. For the same snapshot the AFCARS Dashboard publishes an exact 368,530 under a different vintage and reporting population; the two are not interchangeable and are not mixed.

Source: AFCARS, Trends in Foster Care and Adoption FY2013 to FY2022, national file, In Care On Sept 30th row, submissions as of 2023-05-09. Retrieved 2026-08-04 from acf.gov.

The pace panel's stated averages did not match its own intervals

2026-08-04 · Home page, pace panel

What was published: About 8,400 CPS responses, about 1,500 confirmed victims, and about 500 children entering foster care per day, beside intervals of one every 10 seconds, one every 59 seconds, and one every 2m 49s. The daily averages were written by hand and the intervals were computed from different annual figures, so 500 per day sat next to an interval that implies 511.
What it says now: About 8,200 CPS responses, about 1,500 confirmed victims, and about 480 entering foster care per day, beside one every 11 seconds, one every 59 seconds, and one every 3 minutes. Every figure on the panel, including the screen reader summary, is computed from one set of annual denominators, so a rate and its stated average can no longer drift apart.
Why: The annual figures also moved to current vintages: CPS responses and confirmed victims to FFY2024 from Child Maltreatment 2024, and foster care entries to the preliminary FFY2025 AFCARS count. FFY2025 rather than FFY2024 for entries because Washington did not submit AFCARS 2020 data until FY2025, so the FFY2024 figure is missing a state outright and understates the pace.

Source: Child Maltreatment 2024 (NCANDS), chapter 3 page 24 and Table 3-3 page 38; AFCARS Dashboard (Print Version) No. 33, Numbers at a Glance. Both retrieved 2026-08-04 from acf.gov.

Found something wrong?

If a figure here cannot be traced to the table cited beside it, that is a correction we owe you. Write to info@invisiblechildren.org and say which page and which number. See the methodology for how vintages are chosen and how gaps are labeled.