Skip to content
ParamergeParamerge
Stephen Lieberman
Through Paramerge
Contact

PAN Lab example

US Birth Match

Birth Match: a two-registry join that is itself the referral

Five US states join two registries and call the result a referral. New birth registrations meet a roster of parents with prior terminations of parental rights, serious-harm findings, or certain convictions, and a hit is the whole decision: the documented mechanism has no risk score, no threshold and no ranking, and its only tunable parameters are which findings arm the roster and how far back it reaches — two years in Texas, ten in Maryland and Missouri, back to 1978 in Michigan. Modelled on the five-state birth-match programmes: their shape, not the real systems. Where an accurate match meets no open case, policy requires the response; screening judgment is limited to whether the identity is right. What discretion survives sits downstream, inside a mandated response on a clock. And the roster is fed by the system's own output — a response that ends in a termination places the parent back on the list, permanently in two states, so the next birth meets the same trigger with more history behind it. Watch what the record then shows about the machine nobody was reading: match volumes moved 1,186 to 515 in one state and 243 to 124 in another without the agencies noticing, the independent measurement a legislature ordered in 2018 was never performed, and fewer than one in ten matched families in any state with data received a service. Almost every figure here comes from one proponent researcher's public-records requests, which she flags herself as of limited and dubious accuracy. The question is how you govern a mechanism whose whole model surface is a lookback window. Before you pick a target level: this board cannot be won under Service and Safety Targets or All Governance Targets. With every tool the Lab currently offers, no affordable combination brings this system inside the win condition at those settings. That is a measurement of the deployment this network is derived from, not a puzzle waiting to be cracked. Explore and Service Targets Only can be won.

Stylized model of a documented deploymentChild welfare & family services

Open this example in PAN Lab v0.1 to apply pressures and levers and watch what the system does.

What this models

This example runs on the Birth-Match-class two-registry referral trigger network: 12 components and 25 pathways between them. Every context in the Lab is a stylized model, never a reconstruction of any actual deployment, and each assumption behind it carries a provenance label.

Evidence base: 4 assumed · 7 published baseline. In the Lab, the shaded evidence band behind each headline readout draws its width from the least-established class below.

  • assumed

    This models the two-registry referral-trigger pattern documented across the five US birth-match states, not a reconstruction of any one state's system. Michigan is the anchor for the mechanics that differ between states (linkage at birth-record creation, mandatory screen-in); where another state's record is the source of a value, the element's copy names that state. The dossier's own framing is used throughout: 'mandatory response with no screening discretion', never 'mandatory investigation', because Maryland's mandated response is formally a voluntary prevention assessment outside Child Protective Services (CPS) that a family may refuse entry to, with tracing and escalation duties attached to the refusal.

  • baseline

    Both record-to-model ingests sit at full strength because the documented coverage is total on both sides: every in-state birth is screened with no opt-out, and registry membership is binary with no time bound in two of the five states (Michigan's records reach 1978; Minnesota's lookback is open-ended). The two ingests are marked privacy-sensitive because the payload is a whole birth cohort matched against named prior findings about their parents.

  • baseline

    The mandatory referral channel sits at full strength and the assignment hand-off at full strength because the documented discretion at both steps is zero: Michigan's manual permits screen-out only for an inaccurate identity match or an already-open case and otherwise requires the referral to be screened in and assigned, and Maryland's 2025 policy permits it only where the conviction does not qualify or the child was adopted. There is deliberately no staff-to-model pathway anywhere on this map: the record is explicit that no operator may judge that a decades-old finding is stale, that circumstances changed, or that the referral is pointless.

  • baseline

    The identity-and-open-case screen is drawn as a bounded output check on the match's own product, at a substantial level, rather than as an operator input, because that is what the record describes: central-intake screeners applying a closed, enumerated accuracy-and-dedup test, clerical and not judgmental. The kind is the model-output check rather than a peer check because its grounds are fixed by policy and exhaust at two items, so it screens the referral rather than reviewing anyone's work; the screeners are people, and nothing here asserts an automated component the record does not describe. Its strength comes from documented coverage — about half of Michigan matches already carried a pending investigation or open case, about two thirds of Texas matches had an existing case at receipt, and 45.3 percent of Maryland's 2019 matches were already in services — and its ceiling comes from the same source, which records that the screen reaches identity and case status and nothing else about the family.

  • baseline

    The store-side cycle is the shape's defining structure and every hop is documented: the response record reaches court for a small share of matches (Michigan FY2019 1,188 matched to 24 removals; Texas FY2019 1,138 to 28), a termination places the parent on the roster permanently in two states, agency findings arm the roster with no court step at all, and the roster is one of the two inputs to the match. The reconciliation that would close the loop honestly is drawn empty because no state's record describes a staleness or retirement mechanism, and Michigan stakeholders documented families receiving repeated complaints that each ended in a no-threatened-harm finding.

  • baseline

    The conviction feed sits faint, neither empty nor substantial: the 2018 Maryland statute created it and it has delivered, but the documented volume is six names from the Administrative Office of the Courts between October 2018 and December 2021. A legislated pathway that carries almost nothing is a real strength in this vocabulary, not an absence.

  • baseline

    Two check pathways are drawn empty and both are documented absences rather than authoring convenience. The throughput pathway from the match to a body that reads it: Michigan's volume moved 1,186 to 515 across FY2019-FY2021 and Maryland's 243 to 124 across 2019-2020, and the operating agencies were unaware of the change or declined to explain it. The independent measurement of the match: the 2018 Maryland statute ordered an entity to measure its sensitivity, specificity and predictive value, and a public-records request found no responsive documents. The one check that has fired, the fatality-review channel, drawn faint, is drawn with its documented record in both directions — a statutory expansion won in Maryland, and a recommendation withdrawn in Texas after the agency refused it and stopped supplying the promised quarterly reports.

  • baseline

    baselineDemand 3 comes from the mandated response's documented clock and volume: a three-hour contact standard in Missouri, 48-hour verification, 24-hour intake, five-day home visit and 30-day assessment battery in Maryland, arriving on top of an existing statewide child-protection caseload of 846 to 1,213 birth-match complaints a year in Michigan and 852 to 1,138 families a year in Texas. manualCapacity 3 comes from the documented overlap with human reporting, which is unusually high for this catalogue: 12 of Missouri's first 18 matched families were already the subject of a birth-time report from a medical professional, about half of Michigan's matches already had a pending investigation or open case, and about two thirds of Texas's had an existing case at match receipt. The human channel already reaches most of what the join finds, and the marginal detection is the residual third.

  • assumed

    The whole quantitative record for this deployment flows through one researcher's ad hoc public-records requests (an American Enterprise Institute report and the same author's blog analysis). The author is a birth-match proponent, the agencies generated the tables to order, and she flags the data herself as limited and of dubious accuracy. No state publishes birth-match statistics in any regular report and no independent outcome evaluation exists anywhere, so every baseline on this map is a reading of that single chain, not of an audited series. Maryland's 5-of-124-served figure pairs a federal-fiscal-year numerator with a calendar-year denominator, per the source's own labelling.

  • assumed

    Served newborns, parents and families are not in these dynamics. A match, a referral or an assessment on this map is an institutional signal and never a person, and the documented family-side harms — a newborn removed at the hospital, a family under repeated mandated response, a parent moved toward relinquishment to avoid arming the trigger — live in the case file. No birth-match-specific racial-disparity figure has been published in any state and none is asserted or modelled here; the poverty and race findings in the 2026 legislative testimony concern child-welfare investigation predictors generally. The comparison of this store-to-store shape to entity-resolution deployments in other domains is the atlas's own structural reading, not a claim any cited source makes.

  • assumed

    There is no vendor node and no vendor-mediated pathway on this map: all five states are in-house government builds, and the record documents no commercial vendor and no scoring product anywhere in the programme.

What this example does not show

  • Served newborns, parents and families are not modeled here; the Lab models institutional propagation only, and those outcomes are documented in the case file and measured outside any diagram like this one. The documented family-side harms — a newborn removed at the hospital, a family under repeated mandated response, a parent moved toward relinquishment to avoid arming the trigger — land on people who are outside these dynamics by construction. No birth-match-specific racial-disparity figure has been published in any state and none is asserted or modeled here; the poverty and race findings in the 2026 legislative testimony concern child-welfare investigation predictors generally, not measured properties of the match.
  • The response is described throughout as a mandatory response with no screening discretion, never as a mandatory investigation. Maryland's mandated response is formally a voluntary non-CPS prevention assessment a family may refuse entry to — though the 2025 policy attaches benefits-record and hospital tracing, certified letters, and escalation on suspicion to that refusal, which is why the wording at the trigger stage matters. Central intake's screen-out is clerical: identity accuracy and whether a case is already open, and the map draws it that way.
  • Nearly every quantitative figure on this map flows through one researcher's ad hoc public-records requests — a birth-match proponent working from tables the agencies generated to order, which she herself flags as limited and of dubious accuracy. No independently audited or regularly published statistics exist in any state, no independent outcome evaluation exists anywhere, and the Maryland 5-of-124-served pairing mixes a federal-fiscal-year numerator with a calendar-year denominator, per the source's own labelling.
  • The live 2025-26 Maryland legislative fight targets the reasonable-efforts waiver that makes a match consequential, not repeal of birth match itself: HB 944 of 2025 died without a committee vote, and HB 48 of 2026 was heard on 29 January 2026 with no vote recorded as of the bill page's 30 June 2026 update, so its fate needs re-verification before publication. Two of the five programmes rest on an interagency agreement and a memorandum of understanding rather than statute and can change without legislative process; one of those manuals was reissued unchanged in April 2026, which confirms survival and not permanence. The 'functionally state sterilization' characterization is advocacy testimony in the legislative record, attributed as such.
  • The five-state boundary is about the automated matching mechanism. One state's 2021 law requires investigating births to parents with a child in state custody but uses no data matching, and informal hospital-notification practices exist elsewhere; neither is modeled here. Comparing this store-to-store shape to entity-resolution deployments in other domains is the atlas's own structural reading, not a claim any cited source makes, and the founding-case narratives reach this record through one report's citations of press coverage whose original links are largely dead.

Sources and evidence

What this example rests on, claim by claim. Every entry resolves to the same ledger the Evidence Registry publishes.

  • Five US states (Michigan 2001, Minnesota 2001/2006, Maryland 2009, Texas 2013, Missouri 2021) operate 'Birth Match' programs in which an identity record-join between new birth registrations and a registry of parents with prior terminations of parental rights, serious-harm findings, or specified convictions fires a mandatory child-protective response with no risk score, no threshold, and no screening discretion: Michigan's current policy (PSM 712-2, reissued 2026-04-01) permits screen-out only for an inaccurate identity match or an already-open case and otherwise requires the referral to be screened in and assigned for investigation, and Maryland's 2025 response policy limits screen-out to a non-qualifying conviction or an adopted child while requiring a 24-hour intake coded 'Risk of Harm: birth match' — a mandatory response with no screening discretion, formally structured in Maryland as a voluntary non-CPS assessment families may refuse, with tracing and escalation duties attached. The only tunable parameters anywhere are registry scope and lookback: 2 years in Texas, 10 in Maryland and Missouri, unlimited in Michigan (records to 1978) and Minnesota.

    empirical
    • Investigative Cohen, Learning from the Past: Using Child Welfare Data to Protect Infants Through Birth Match Policies (American Enterprise Institute, 2022) https://aei.org/wp-content/uploads/2022/05/Learning-from-the-Past-Using-Child-Welfare-Data-to-Protect-Infants-Through-Birth-Match-Policies.pdf
    • Government Michigan Department of Health and Human Services, Children's Protective Services Manual PSM 712-2: CPS Intake, Special Cases, Birth Match Section (PSB 2026-002) (2026) https://mdhhs-pres-prod.michigan.gov/olmweb/ex/PS/Public/PSM/712-2.pdf
    • Government Maryland Department of Human Services, Social Services Administration, SSA/CW 25-03: Birth Match Response Policy (2025) https://dhs.maryland.gov/documents/SSA%20Policy%20Directives/Child%20Welfare/SSA%2025-03%20CW%20Birth%20Match%20Response%20Policy%208.8.25.docx.pdf
    • Government Minnesota Revisor of Statutes, Minn. Stat. 260E.14 subd. 4 with 260E.03 subd. 23 (Birth Match Screening and Threatened-Injury Trigger Classes) (2021) https://www.revisor.mn.gov/statutes/cite/260E.14#stat.260E.14.4

  • The Birth Match trigger is self-arming: the registry that fires it is populated by the system's own outputs, so a birth-match response that ends in a new termination of parental rights writes the parent back onto the match list — permanently in Michigan and Minnesota, which have no lookback limit or retirement mechanism — and every subsequent birth re-fires with strictly increasing history; Michigan additionally adds manual list entries for severe cases without any termination. Maryland law accelerates the loop by letting the agency skip reasonable reunification efforts on account of a prior involuntary termination, and the 2025-26 Maryland legislative fight (HB 944 of 2025, died without a committee vote; HB 48 of 2026, heard 2026-01-29 and dead in committee without a vote when the session adjourned sine die on April 13, 2026) targeted that waiver, not birth-match repeal — with Civil Rights Corps testifying that the combination is 'functionally state sterilization' and coerces parents into 'voluntary' relinquishment, an advocacy characterization in the legislative record.

    empirical
    • Investigative Cohen, Learning from the Past: Using Child Welfare Data to Protect Infants Through Birth Match Policies (American Enterprise Institute, 2022) https://aei.org/wp-content/uploads/2022/05/Learning-from-the-Past-Using-Child-Welfare-Data-to-Protect-Infants-Through-Birth-Match-Policies.pdf
    • Advocacy Richardson and Steckel (Civil Rights Corps), Testimony in Support of HB 48, Judiciary Committee Testimony Compilation (2026) https://mgaleg.maryland.gov/cmte_testimony/2026/jud/32900_02042026_155551-335.pdf
    • Government Maryland General Assembly, HB 48 (2026): Right to Fight Act, Bill Status (2026) https://mgaleg.maryland.gov/mgawebsite/Legislation/Details/HB0048?ys=2026RS
    • Government Michigan Department of Health and Human Services, Children's Protective Services Manual PSM 713-08: Special Investigative Situations, Birth Match Investigation (2025) https://mdhhs-pres-prod.michigan.gov/olmweb/EX/PS/Public/PSM/713-08.pdf

  • The Birth Match automation runs without ownership or evaluation: match volumes fell by more than half in Michigan (1,186 in FY2019 to 515 in FY2021) and Maryland (243 in CY2019 to 124 in CY2020) without the operating agencies noticing or being able to explain the change; fewer than 10 percent of matched families in any state with data received services (Maryland: 5 of 124 matched families in the source's mixed FFY2020/CY2020 pairing); Maryland's unanimous 2018 expansion statute ordered an independent evaluation of the match's sensitivity, specificity, and predictive value that a public-records request found was never implemented; and the only oversight that demonstrably changed the system — child fatality review teams auditing the matcher's misses — widened the trigger both documented times. Nearly all of these figures flow through one proponent researcher's ad hoc public-records requests (AEI 2022), which the author herself flags as limited and of dubious accuracy; no state publishes birth-match data in any regular report.

    empirical
    • Investigative Cohen, Learning from the Past: Using Child Welfare Data to Protect Infants Through Birth Match Policies (American Enterprise Institute, 2022) https://aei.org/wp-content/uploads/2022/05/Learning-from-the-Past-Using-Child-Welfare-Data-to-Protect-Infants-Through-Birth-Match-Policies.pdf
    • Reference Cohen, Using Child Welfare Data to Learn from the Past: Why Is It So Unpopular? (Child Welfare Monitor, 2022) https://childwelfaremonitor.org/2022/09/07/using-child-welfare-data-to-learn-from-the-past-why-is-it-so-unpopular/
    • Government Maryland General Assembly, 2018 Md. Laws Ch. 497 (SB 490): Child Abuse and Neglect, Disclosure of Identifying Information and Investigations (2018) https://mgaleg.maryland.gov/2018RS/chapters_noln/Ch_497_sb0490E.pdf
    • Government Baltimore City Child Fatality Review Team, Eliminating Child Abuse and Neglect Fatalities in Baltimore City (Subcommittee on Child Abuse and Neglect) (2017) https://www.healthybabiesbaltimore.com/_files/ugd/6a76e9_e2244a6672dd4862b9e1b3d43e6cb175.pdf

Where this connects

Institutional pressures in this domain

  • Workload surge — Demand outruns staffing; per-case attention shrinks and review becomes triage.
  • Deadline pressure — Statutory or managerial timeliness rules reward fast approval of machine output over slow disagreement.
  • Staff turnover — Experienced skepticism leaves; new staff calibrate their trust on the tool itself.
  • Data & policy drift — The world, the intake process, and the rules change under a system trained on how things used to be — two mechanisms with different remedies: the statistical properties of what the system processes move (concept drift), or the mixture of inputs arriving in deployment differs from the mixture it was trained on (covariate shift).
  • Compliance over substance — Paper controls (sign-offs, checklists) satisfy audits while the behavior they describe erodes.

All of them in context on the Child welfare & family services domain page.

Levers available here and the patterns behind them

Documented case histories