Patriot University — Legislative Impact Scoring (LIA)
Two-track impact assessment of a single piece of federal legislation or executive action. Produces a Legislative Impact Assessment (LIA) with:
- a Provenance gate (Layer 0) that identifies the action to a citable primary text before anything is scored;
- a Q-track (quantitative: fiscal / distributional / sectoral) that aggregates and attributes others’ estimates — CBO, JCT, TPC, Tax Foundation, PWBM, Yale Budget Lab — never presenting them as PU’s own, each carrying the assumptions (“knobs”) that produced it;
- an R-track (rights & liberties) that runs the doctrinal-legal workflow — rights-exposure, burden/chilling, and institutional severity — via PU’s legal-expert and monitoring skills;
- a shared back half: implementation, ex-post/realized, and a bipartisan divergence layer that names the single knob behind each disagreement.
The two tracks are never combined into a single composite score (methodology §0, knob 22). A bill can be fiscally favorable and rights-corrosive, or the reverse; presenting that tension honestly — rather than netting it into one number — is the product.
LIA scores legislation/executive action, not a person. It is a sibling of, but categorically distinct from, the person-scoring instruments — malice-evaluator (a person’s actions), spokesperson-misinformation-assessment (a person’s statement veracity), and dangerous-rhetoric-assessment (a person’s rhetoric).
Canonical rules live in docs/legislative-impact-methodology.md — this skill operationalizes them. If the skill and the charter disagree, the charter governs. The 22 knobs and the orientation-paired source reference class live in docs/legislative-impact-knob-baseline.md.
When to Use This Skill
Run the LIA when analyzing the impact of a specific:
- Statute — an enacted or pending federal bill (with a bill number + Congress).
- Executive order, presidential memorandum, or proclamation.
- Major agency rule — one carrying or statutorily requiring a Regulatory Impact Analysis.
- Enforcement-posture change — where a documented shift in enforcement priorities has a scoreable cohort or rights impact (R-track primary).
Do NOT run this skill for:
- Roll-call scorecards / vote ratings — those measure position-taking, not consequences; they live in a separate namespace and never appear inside an LIA block.
- A request for a single “how bad is this bill” number — categorically refused; produce the two-track scorecard instead.
- A candidate’s or private citizen’s speech — route to the person-scoring instruments / inclusion gate.
- Statement veracity →
spokesperson-misinformation-assessment; rhetoric →dangerous-rhetoric-assessment; fact-checking a claim →citation-checker.
First step, always: run the router to resolve the current SKILL.md paths of the legal-expert and monitoring skills you will compose — several are product-local and must not be hardcoded:
make route Q="<the specific right / sector / doctrine this action touches>"
The Non-Collapse Rule (read first)
There is no neutral score. Every framework LIA cites embeds a dozen contestable normative choices inside one public number. PU’s position is not “we produce the true number” — it is “we make every knob visible, we show how the number moves when you turn it, and we tell you who turned it which way.“
And there are two families of impact that do not reduce to one number. LIA runs them as parallel tracks that are never summed, averaged, netted, or rendered as a single meter or grade. Any output that produces one composite scalar is a defect. This is the load-bearing rule of the whole instrument.
Section 1: The Provenance Gate (Layer 0 — runs FIRST)
No LIA exists without a fully-identified subject. Populate before any measurement:
| Field | Requirement |
|---|---|
action_type |
statute / executive_order / presidential_memo / proclamation / agency_rule / enforcement_change |
identifier |
Bill+Congress (HR-1-119), EO number, Federal Register cite, or docket ID |
official_title |
Verbatim |
sponsor_or_issuer |
Member(s)/committee, or President/agency |
status |
introduced / reported / passed-one-chamber / enacted / in-effect / enjoined / stayed / revoked |
text_source |
Congress.gov / GovInfo / Federal Register / eCFR permalink + retrieval date |
official_scores |
Any CBO/JCT/OCACT/CMS-OACT/RIA score attached, cited (may be empty) |
as_of_date |
The date the assessment reflects |
Use policy-analyst-legislative-specialist / federal-register-regulatory-analysis for identification. If the action cannot be identified to a citable primary text, STOP — no LIA is produced.
Section 2: The Q-track (quantitative — attribute, don’t assert)
Not every layer fires; mark each applicable / not-applicable / not-yet-assessed. Sparsity is visible, never imputed.
2.1 Layer 1 — Fiscal
- Locate the official score (CBO/JCT) if one exists; declare its baseline (knob 1) and window (knob 2).
- Note what it omits: deficit financing not distributed (knob 11); unquantified effects (knob 13).
- Prefer ranges to point estimates (knob 15).
2.2 Layer 2 — Distributional
- Report effect by cohort (§4 taxonomy), attributing each estimate to its source with its knobs.
- Normalize before comparing — CBO, JCT, and Treasury use different income concepts (knob 8).
- Watch incidence (knob 6), unit of analysis (knob 7), in-kind valuation (knob 10).
- Race/ethnicity or other narrowed-breakout cuts require the §4 provenance note and the
imputedflag when any part is imputed.
2.3 Layer 3 — Sectoral
- Apply the domain framework for the sector (health: HISIM2/HIPSM; tax: TPC/Tax Foundation; trade: Yale Budget Lab/PIIE; energy: NEMS/EIA; etc. — knob-baseline Part B).
- Watch discount rate (knob 3), elasticities (knob 5), geographic scope (knob 12), take-up/admin burden (knob 14), and unquantified effects (knob 13) as a first-class field.
Attribution discipline (absolute): every Q-track figure is attributed to a named source with its orientation label and knobs, or clearly labeled PU-produced (PolicyEngine, Phase 2) with its own knobs. Never present one source’s point estimate as “the” impact.
Section 3: The R-track (rights & liberties — the §1.11 doctrinal workflow)
For each affected right, compose the relevant legal-expert skill (resolve via make route) and run:
- Identify the right + governing doctrine.
- Classify the government action — content-based restriction, viewpoint discrimination, prior restraint, vote denial, vote dilution, disparate-impact vs. disparate-treatment, procedural vs. substantive due process, anti-commandeering, unconstitutional conditions.
- Map to the scrutiny tier / statutory test (strict / intermediate / rational-basis; the Gingles preconditions; the Brnovich guideposts) — declaring the precedent regime (knob 19). The same map “dilutes” or not depending on pre-Shelby / post-Brnovich / post-Callais doctrine.
- Assess chilling effects + burden incidence (knob 21) — who is deterred, which cohorts bear the burden, even where no one is directly prohibited.
3.1 Layer R1 — Rights-exposure
- Output: a rights-direction indication (
expands/contracts/mixed) with the right + doctrine named and the call attributed, orexposure-only(the recommended default) where a house-voice direction verdict is not defensible. - Plus a doctrinal-exposure rating:
settled/contested/likely-unconstitutional, **as a range with the governing precedent cited — never a verdict that the action is unconstitutional.** - Skills:
voting-rights-act-expert,voter-suppression-law,patriot-voting-research,election-law-and-administration(voting);first-amendment-legal-expert,patriot-press-freedom-tracker(speech/press);fourth/fifth/eighth/fourteenth-amendment-legal-expert(equal protection / due process);separation-of-powers-legal-expert,tenth-amendment-legal-expert(structure).
3.2 Layer R2 — Burden / chilling
- Map burdened cohorts (§4 taxonomy, incl. protected class, language-minority, immigration status).
- Record the chilling-effect field (knob 21):
estimated(with survey/behavioral basis) /acknowledged-narratively/ignored.ignoredandnot_disclosedrender as a visible credibility flag — the rights analog of “unquantified benefits.”
3.3 Layer R3 — Institutional severity
- Where the action touches oversight, independent bodies, surveillance, or checks (politicizes an independent agency, strips IGs, criminalizes protest, restricts judicial review, impoundment), score it with
democratic-health-monitoring(5-level Healthy→Critical) overrule_of_law/institutional_checks/civil_liberties, with escalation velocity. - Follow the democratic-health anti-patterns: score actions and outcomes, not rhetoric; require multi-source corroboration above severity 2; benchmark against comparative erosion, not domestic exceptionalism.
- Declare the precedent regime (knob 19).
Section 4: Cohort Taxonomy (define once, use everywhere)
Q-track / shared: income (SPM-relative + decile), age band, household composition, geography (state / county / metro-nonmetro / congressional district), race/ethnicity, disability, veteran status, industry/occupation, insurance status, housing tenure.
R-track adds: protected class (Equal Protection / Title VII / FHA classifications), language-minority status, immigration/citizenship status, and — where lawfully and reliably measurable — religion and LGBTQ status.
Provenance rule (mandatory): any cohort cut depending on a federal breakout imputed, narrowed, or discontinued in 2025–26 (SOGI items removed under EO 14168; CPI cells suspended; October 2025 CPS never collected) carries a visible provenance-and-coverage note per the ITI inferred-data-transparency standard, and sets imputed: true. Never present a demographic impact as observed when the breakout was imputed or discontinued. Prefer a policy-coding index (e.g. MAP) over an imputed survey estimate, and say so.
Section 5: The Knobs (every value ships with them)
No layer value is publishable without its knobs populated (methodology §3). not_disclosed is a first-class value and renders as a credibility flag. Minimum required knobs by layer:
| Layer | Mandatory knobs |
|---|---|
| Q1 Fiscal | 1 baseline · 2 horizon · 4 static/dynamic · 11 deficit-financing · 15 uncertainty |
| Q2 Distributional | 6 incidence · 7 unit · 8 income-concept · 9 cohort-def · 10 in-kind · 11 deficit-financing · 16 data-vintage |
| Q3 Sectoral | 3 discount-rate · 5 elasticities · 12 geo-scope · 13 unquantified · 14 take-up/admin |
| R1 Rights-exposure | 19 precedent-regime · 20 scrutiny-tier · 22 commensurability |
| R2 Burden/chilling | 21 chilling-estimation · 9 cohort-def |
| R3 Institutional | 19 precedent-regime · 15 uncertainty |
| Any cited source | 17 topic-selection · 18 funding/governance (of the source) |
Knob 22 is fixed at parallel-track. Monetizing a rights impact into a dollar bottom line (monetized-folded) is a defect.
Section 6: Divergence (Layer 6 — first-class product)
Whenever ≥2 credible sources disagree materially, build a divergence entry:
- Pair sources across orientation (knob-baseline Part B) — a left-of-center and a right-of-center estimate of the same provision. A one-sided table is invalid.
- Name the pivot knob — the one (or two) that produces the gap (corporate incidence 6, discount rate 3, precedent regime 19, static-vs-dynamic 4…).
- Cite the literature on that knob, not on the conclusion.
- Report the result under each assumption, side by side — never adjudicate to one number.
- Score each source via
source-credibility-scoring; flag any “supports X jobs / costs $Y” claim lacking a counterfactual or with an undisclosed multiplier configuration as a credibility red flag.
Section 7: Implementation, Realized & Self-Audit (Layers 4–5)
- Layer 4 (Implementation): did the rule get written, funds obligated, action enjoined/stayed/upheld, agency complying? Track via CourtListener/RECAP + USAspending. For rights actions, a court blocking the action is the realized impact. Use
policy-analyst-legislative-specialist. - Layer 5 (Realized): defaults to
too-early. A causal claim requires the §11 toolkit (difference-in-differences with staggered-adoption corrections, synthetic control, RDD) and an explicit attribution-limits statement. Report implementation fidelity separately from outcome change — conflating them is the most likely failure mode. - Self-audit: LIA publishes its own accuracy retrospectives on a fixed schedule (CBO’s forecast-accuracy practice; Yale’s one-year tariff retrospective). A “Score Audit” comparing predicted to realized is a distinct, high-credibility product.
Section 8: Output Format (Scorecard block)
Rendered into the legislation/executive-action analysis article as a two-track scorecard. The tracks are shown on one surface and explicitly refuse a single composite.
---
## Legislative Impact Assessment
**Action:** [official_title] ([identifier]) · **Status:** [status] · **As of:** [as_of_date]
**Rigor tier:** [Standard / Publication] · **Sources aggregated, not PU-produced** (unless a value is labeled PU/PolicyEngine)
> **Two tracks, never one number.** This action is scored on a quantitative track and a rights
> track that are reported separately and are never combined. It may point in opposite directions
> on the two.
### Q-track — Quantitative impact
| Layer | Finding (attributed) | Source(s) | Key knobs disclosed |
|-------|----------------------|-----------|---------------------|
| Fiscal | [e.g. "-$X deficit, FY26-35"] | CBO [cite] | baseline: current-law · window: 10yr · deficit-financing: not-distributed |
| Distributional | [effect by cohort] | TPC ↔ Tax Foundation | income-concept: expanded-cash · incidence: 75/25 |
| Sectoral | [sector effect] / not-applicable | [source] | discount-rate: 2% vs 7% · unquantified: [listed] |
### R-track — Rights & liberties impact
| Layer | Finding | Right / doctrine | Precedent regime (knob 19) |
|-------|---------|------------------|-----------------------------|
| Rights-exposure | [expands/contracts/mixed/exposure-only] · exposure: [settled/contested/likely-unconstitutional] | [right + governing test] | [pre-Shelby / post-Brnovich / post-Callais] |
| Burden/chilling | cohorts: [...] · chilling: [estimated/acknowledged/**ignored ⚑**] | | |
| Institutional | severity [1-5] on [indicators] / not-applicable | | |
### Divergence
**[Provision]:** [Left source] estimates [X] · [Right source] estimates [Y]. **Pivot knob:** [n — name].
[What the literature says about that knob.] Reported under each assumption; not adjudicated.
### Implementation & realized
**Implementation:** [status] — [litigation posture / obligation status]. **Realized:** [too-early / fidelity-only / outcome-estimated + method + attribution limits].
> **Framework disclosure:** The Legislative Impact Assessment applies a published, two-track
> analytical framework to a specific, cited piece of legislation or executive action. Quantitative
> figures are attributed to their source with the assumptions ("knobs") that produced them; they are
> not PU's own estimates unless explicitly labeled as such. Rights findings are evaluative judgments
> about the cited text and governing doctrine, not verdicts that any action is unconstitutional. The
> two tracks are reported separately and are never combined into a single score. Each value cites its
> source and the date it reflects; this assessment stales as baselines, amendments, data, and
> precedent change, and is subject to update.
Machine record conforms to docs/legislative-impact.schema.json and is saved as a sidecar at docs/lia-records/ (validated before publish). The Markdown block above is the authoring/preview form; on publish the scorecard renders as a self-contained HTML component (docs/legislative-impact-rendering-spec.md + theme/patriot-university-theme/assets/css/lia.css) — no composite score, color = track identity (never a good/bad valence), knob-chips with not_disclosed/ignored flagged, and a STATUTORY ACCESS badge kept distinct from any constitutional-right framing. Taxonomy home + publish flow: docs/legislative-impact-publishing.md (an EKB doc under context-and-backgrounders/contextual-analysis).
Section 9: Claims-Integrity & Defamation Guardrails
Reuses the financial-cost / DMA / SMA / DRA architecture: Truth (every Q figure cites a source + retrieval date; every R finding rests on verbatim statutory/EO text + cited doctrine); Fair-report privilege (public legislation, official scores, court dockets); Opinion privilege (Milkovich — exposure calls and severity scores are evaluations grounded in disclosed facts); Public-official doctrine (Sullivan — sponsors/issuers in office); good-faith uniform method (published framework, bipartisan source-pairing, no animus).
Absolute guardrails:
- Never present a single composite score netting the two tracks (knob 22).
- **Never assert an action is unconstitutional** — report
contested/likely-unconstitutionalas a range with the governing precedent, plus the actual disposition when one exists. - Never present an aggregated estimate as PU’s own unless genuinely PU-produced (PolicyEngine, labeled) — aggregated figures are attributed.
- **Never claim a realized outcome was caused by the action** without the §11 toolkit + explicit attribution limits.
- Never fold a rights impact into a dollar bottom line (knob 22).
- Flag every undisclosed knob as a visible credibility signal.
- Never render a roll-call scorecard inside an LIA block.
Publishing is human-gated; any likely-unconstitutional call or any PU-original (PolicyEngine) estimate additionally clears citation-checker / the Claims Ombudsman, and a rights verdict clears counsel.
Section 10: Composition with Other Skills
0. policy-analyst-legislative-specialist / federal-register-regulatory-analysis -> Layer 0 provenance + Layer 4
1. Q-track: source registry (aggregate + attribute) | PolicyEngine (Phase 2, PU-original, labeled)
2. R-track (resolve via `make route`):
voting-rights-act-expert / voter-suppression-law / patriot-voting-research -> R1 voting
first-amendment-legal-expert / patriot-press-freedom-tracker -> R1 speech/press
fourth/fifth/eighth/fourteenth-amendment-legal-expert -> R1 equal-protection/due-process
separation-of-powers-legal-expert / tenth-amendment-legal-expert
/ madisonian-constitutional-analysis -> R1 structure
democratic-health-monitoring -> R3 institutional severity
3. source-credibility-scoring (FRS/DCS) -> Layer 6 divergence credibility
4. constitutional-law-advisors / pu-advisor-* personas -> expert review lens (>=2 diverse)
5. citation-checker + Claims Ombudsman -> publish gate (runs last)
Interactions
- Reads:
docs/legislative-impact-methodology.md(canonical rules);docs/legislative-impact-knob-baseline.md(the 22 knobs + orientation-paired sources). - Receives from
policy-analyst-legislative-specialist: action identity, status, Youngstown/major-questions posture, compliance tracking. - Composes: the legal-expert skills (R-track),
democratic-health-monitoring(R3),source-credibility-scoring(divergence). - Hands off: roll-call ratings → separate scorecard namespace (never in an LIA block); statement veracity →
spokesperson-misinformation-assessment; rhetoric →dangerous-rhetoric-assessment.
Section 11: Anti-Patterns
What This Skill Is NOT
- NOT a single-number bill grader — the two tracks are never combined.
- NOT a producer of original fiscal numbers in v1 — it aggregates and attributes (PolicyEngine original estimation is a labeled Phase 2 layer).
- NOT a roll-call scorecard — vote ratings measure position-taking, not consequences.
- NOT a constitutional verdict machine — it reports doctrinal exposure + litigation status, never “this is unconstitutional.”
- NOT an ex-post causal claim generator — realized outcomes default to “too early” and require the §11 toolkit + attribution limits.
- NOT a person-scoring instrument (those are DMA/SMA/DRA).
Common Mistakes to Avoid
| Mistake | Correction |
|---|---|
| Netting Q and R into one “impact score” | Prohibited — two-track scorecard, never a composite (knob 22) |
| Reporting one source’s number as “the” impact | Attribute with knobs; show orientation-paired divergence |
| Presenting an aggregated figure as PU’s estimate | Attribute to source; only PolicyEngine values are PU-produced (labeled) |
| Omitting the baseline/window on a fiscal figure | Knobs 1–2 are mandatory on Q1 |
| A race/ethnicity cut with no provenance note | §4 provenance note + imputed flag mandatory when the breakout is narrowed/imputed |
| “This EO is unconstitutional” | Report contested/likely-unconstitutional as a range + the disposition |
| Chilling effect silently omitted | Record knob 21; ignored renders as a credibility flag |
| A one-sided divergence table | Invalid — pair across orientation |
| Rendering a vote scorecard inside the block | Separate namespace only |
Section 12: Examples
Example A — a tax reconciliation title (Q-track heavy, R-track empty)
- Provenance:
HR-1-119, enacted, Congress.gov cite. - Q1 Fiscal: CBO −$X over FY26–35 (baseline: current-law; window: 10yr; deficit-financing: not-distributed ⚑). CRFB flags a current-policy-baseline maneuver (knob 1) that would treat extension as costless.
- Q2 Distributional: TPC (expanded-cash income) shows top-decile gain; ITEP adds a race/ethnicity cut with a provenance note (imputation documented). Divergence with Tax Foundation on the growth offset — pivot knob 4 (static vs. dynamic).
- R-track:
not-applicable(no rights dimension). Recorded as not-applicable, not omitted. - Output: Q-track scorecard + divergence entry; R-track explicitly empty. No composite number.
Example B — an executive order targeting law firms (R-track heavy, Q-track thin)
- Provenance: EO number, Federal Register cite, status
enjoined. - R1 Rights-exposure:
first-amendment-legal-expert— viewpoint-discrimination + unconstitutional-conditions; scrutiny: strict; direction: contracts (attributed to the district court’s holding); exposure:likely-unconstitutionalas a range, precedent regime declared. - R2 Burden/chilling: burdened cohort = targeted firms and their clients; chilling: estimated (the order deters representation far beyond the named firms — knob 21).
- R3 Institutional:
democratic-health-monitoringseverity 4 onrule_of_law/institutional_checks, multi-source corroborated. - Layer 4 Implementation: enjoined by district court (docket cite) — the injunction is the realized impact.
- Q-track: thin /
not-applicable. - Output: R-track scorecard + implementation; Q-track empty. No composite; no “unconstitutional” verdict — exposure range + disposition.
Example C — refused request
- Request: “Just give me a 1–10 score for how bad this bill is.”
- Response: Refuse the single number; produce the two-track scorecard and explain the non-collapse rule (a bill can be fiscally favorable and rights-corrosive; one number would hide exactly the tension that matters).