Institutions
Reference file for real-world institutional facts that ground the model's assumptions, calibrate its parameters, and support the paper's applications (Equal Protection, Due Process). Organized by model component.
Stare decisis
Formal rules
- Horizontal stare decisis: courts should follow their own prior decisions. A policy, not a constitutional rule — the Supreme Court has repeatedly stated it will overrule precedents it considers wrongly decided. "Stare decisis is not an inexorable command."
- Vertical stare decisis: lower courts are strictly bound by higher-court holdings. Not the focus of this paper (see Jan 12 outline, point 9), but relevant as context for why horizontal stare decisis matters differently.
- Holdings are binding; dicta are persuasive but not binding.
- Overruling is possible but carries institutional costs.
Official stare decisis factors for overruling
When considering overruling a constitutional precedent, the Court invokes:
- Quality of reasoning — was the precedent well-reasoned or egregiously wrong?
- Workability — has the rule proven unmanageable?
- Consistency with related doctrine — is the precedent an outlier?
- Reliance interests — have people structured their lives around it?
- Changed facts or understanding — has the world changed in a way that undermines the rule?
These structure justification rather than dictate outcomes.
Real constraints on overruling
- Legitimacy: The Court has no enforcement power. It relies on compliance by other branches, public acceptance, and lower-court cooperation. Overruling too freely makes decisions look political and makes court-packing threats credible.
- Professional norms: Overruling must be framed as correcting error or restoring coherence. Openly saying "we overrule because we prefer a different policy" is normatively forbidden.
- Coordination costs: Constant revision of constitutional meaning prevents lower courts, legislators, and private actors from relying on doctrine.
- Reputational capital: Justices care about the reputation of the Court. Overruling too much depletes institutional capital and invites retaliation.
- Internal dynamics: Overruling requires a stable majority willing to bear dissent. Marginal justices often prefer narrowing or distinguishing. This creates inertia.
- Political feedback: Aggressive overruling intensifies confirmation battles and politicizes judicial selection.
Lower-court compliance (background)
Lower courts are formally bound but retain interpretive margin (what is the holding, how broadly does it apply, has it survived intervening change). When Supreme Court doctrine is unstable, lower courts rationally respond by narrowing holdings, delaying extensions, relying on signals rather than rules, and preserving issues for future review. Compliance remains formal but constraint softens. The authority of precedent depends on perceived durability, not just hierarchy.
Holdings
Definition
The holding of a case is the legal rule or principle necessary to decide the case, applied to its material facts. It is the binding part of the decision.
- Not just the outcome (who wins)
- Not everything the court says
- The narrowest legal proposition that explains the result given the facts
Holding vs dicta vs reasoning
| Component | Binding? | Model role |
|---|---|---|
| Holding (rule necessary for outcome) | Yes | Core constraint on admissible set |
| Reasoning / legal tests (frameworks, tiers) | Often treated as binding | Constrains the hypothesis class |
| Dicta (statements not necessary to outcome) | No, but influential | Signals future constraints |
| Concurrences | No (but Marks rule for pluralities) | Alternative admissible theories |
| Dissents | Never | Candidate models for future regime changes |
Breadth ambiguity
Holdings are inherently ambiguous in scope. The same decision can support narrow or broad readings. Brown v. Board illustrates this — the spectrum of plausible holdings ranges from "segregation in public schools violates EP" (narrow, fact-bound) through "state action conveying racial inferiority violates EP" (anti-stigma) to "any racial classification is presumptively unconstitutional" (colorblindness). The fight over which reading controls is a fight over how much of fact space the precedent constrains.
Courts generally resist broad holdings because they reduce future discretion, raise error costs, and make later courts feel trapped. Courts prefer narrow holdings + incremental extension, or standards rather than rules, unless coordination is fragile.
Overruling and distinguishing
Courts can: distinguish a precedent (facts differ in legally relevant ways), limit it (confine it to its facts), or overrule it (declare it wrong). All three reclassify how a prior data point constrains the admissible set.
The Marks rule
When no single rationale commands a majority in a plurality decision, the holding is the position taken by the Justices who concurred in the judgment on the narrowest grounds. Famously messy in practice.
Sanctions and enforcement
What actually happens to judges who defy precedent? This is the empirical grounding for the model's sanction parameter.
TODO: fill from legal scholarship
- Reversal: lower-court judges face reversal on appeal. What is the reversal rate? How does it vary by circuit, by doctrinal area? How costly is reversal for a judge's career?
- Reputation: professional reputation among peers, in the bar, in the academy. How is this measured? Is there evidence that reputation responds to doctrinal fidelity?
- Career consequences: promotion to higher courts, senior status, assignment to desirable cases. Is there evidence linking doctrinal compliance to career outcomes?
- Professional discipline: formal mechanisms (judicial conduct commissions, impeachment). How rare? Under what circumstances?
- Panel dynamics: how do co-panelists discipline each other? Do judges write separately to flag another judge's deviation? Does the threat of a concurrence or dissent constrain?
- Social sanctions: exclusion from professional circles, conferences, clerkship networks. The "dinner invitation" mechanism (discussed in meetings — low cost to exclude a tainted colleague).
What we know so far
- Holger emphasizes that binding force should arise endogenously (Dec 9 meeting)
- The Kandori-style enforcement idea: the "viral tweet moment" where deviation is made public serves as a marking device (see thinking.md)
- Bård Harstad's suggestion: cost of complaining depends on number who complain (1/n probability of testifying). Panel members are expected to call out wrongdoing; silence implicates them.
- Deciding in panels helps with monitoring. For trial judges, appeals and reversals play the same role. Single-judge decisions at the top (e.g., Brazil) are dangerous for the rule of law.
Court mechanics
How judges encounter cases and each other. Relevant for understanding how constraints accumulate and how the model's sequential structure maps to reality.
TODO: fill from legal scholarship
- Case assignment: how are cases assigned to judges/panels? Random assignment (most federal circuits), chief justice assignment of opinion-writing (SCOTUS). What is the actual randomization mechanism?
- Panel composition: three-judge panels in circuit courts, nine justices at SCOTUS. How are panels drawn? Rotation? Is there evidence of strategic panel manipulation?
- En banc review: when does a full circuit sit together? What triggers it? How does this relate to resolving intra-circuit doctrinal conflict?
- Opinion-writing process: majority opinion assigned by most senior justice in the majority. Opinion bargaining — how do concurrences and dissents shape the majority opinion? How does the need to hold five votes constrain holding breadth?
- Cert process: SCOTUS hears ~70 cases/year out of ~7000 petitions. The "rule of four" (four justices must agree to hear a case). How does case selection shape which constraints get added to the admissible set? Selection effects on doctrinal evolution.
- Amicus briefs and information: how do courts get information about facts beyond the case record? How does this affect the dimensionality of the fact space they reason about?
Overruling in practice
Empirical facts about how often and when the Court overrules itself. Relevant for calibrating the overruling cost parameter.
TODO: fill from legal scholarship
- How many times has SCOTUS explicitly overruled itself? (Congressional Research Service tracks this — roughly 150 times as of 2020)
- What is the distribution over time? By doctrinal area?
- What predicts overruling? Composition changes? Age of precedent? Number of concurrences/dissents at the time?
- How does "stealth overruling" (narrowing a precedent to irrelevance without formally overruling) compare in frequency?
- Is there empirical evidence on the relationship between overruling frequency and lower-court compliance?
Equal Protection doctrine
Institutional facts for the paper's EP application. Focus on doctrinal structure that maps to the model's constraint accumulation.
Doctrinal structure
EP asks whether the government has treated similarly situated persons differently without adequate justification. Three steps: (1) is there a classification? (2) what level of scrutiny applies? (3) does the classification survive that scrutiny?
Facial classifications explicitly distinguish between groups → trigger heightened scrutiny depending on the trait.
Facially neutral laws with disparate impact violate EP only if discriminatory purpose is shown (Washington v. Davis). Purpose inferred via Arlington Heights factors (historical background, sequence of events, procedural departures, legislative history). Must be "because of," not merely "in spite of" (Feeney).
Tiers of scrutiny
| Tier | Trigger | Test |
|---|---|---|
| Strict | Race, national origin, alienage | Compelling interest + narrowly tailored |
| Intermediate | Sex, legitimacy | Important objective + substantially related |
| Rational basis | Everything else | Rationally related to legitimate interest |
Post-SFFA (2023): race-based affirmative action in admissions effectively barred. "Rational basis with bite" (Cleburne, Romer): laws motivated by animus fail even rational basis.
Settled vs unsettled
Settled: tiered scrutiny framework; intent requirement for neutral laws; basic tests per tier; race → strict; sex → intermediate.
Unsettled: what counts as compelling/important; how narrow is "narrowly tailored"; how to infer intent from complex facts; how to classify new traits or contexts; how much weight to give stigma, history, or administrability.
Dimension dictionary (for coding EP cases)
- D1 Facial classification: explicit use of group membership in the rule (high/low)
- D2 Protected trait: race / sex / disability / sexual orientation / etc.
- D3 Intent evidence: evidence the government acted because of group disadvantage (high/low)
- D4 Interest strength: asserted governmental objective (weak → strong/compelling)
- D5 Means-ends fit: over/underinclusiveness; availability of alternatives; narrow tailoring (poor → tight)
- D6 Stigma / caste risk: expressive meaning and hierarchy reinforcement (low → high)
- D7 Institutional setting: wartime/emergency, schools, zoning, admissions, etc.
- D8 Precedent density / conflict: how "determined" the local doctrine is (low/high)
Coded cases
Brown v. Board (1954): D1 High; D2 Race; D3 not needed (facial); D6 High; D7 Public education; D8 High conflict (overturning "separate but equal").
Loving v. Virginia (1967): D1 High; D2 Race; D4 "racial integrity" = illegitimate; D6 High; D7 Marriage/criminal law.
Korematsu (1944): D1 High; D2 Race/ancestry; D4 Very strong (war/security); D7 Wartime; D5 Fit sharply contested.
Washington v. Davis (1976): D1 Low (neutral); D3 Low/unclear; D5 Impact insufficient; D8 Doctrinal gatekeeping.
Arlington Heights (1977): D1 Low; D3 Multi-feature (sequence, departures, history); D7 Zoning. Judicially endorsed feature-selection recipe for D3.
Feeney (1979): D1 Low; D2 Sex (effectively); D3 High threshold. Narrows admissible inference from impact to intent.
Reed v. Reed (1971): D1 High; D2 Sex; D4 Administrative convenience = too weak; D8 Early sex-equality marker.
Craig v. Boren (1976): D1 High; D2 Sex; D4 Traffic safety = important; D5 Fit poor (sex as noisy proxy).
United States v. Virginia / VMI (1996): D1 High; D2 Sex; D4 Tradition = insufficient; D5 Alternative program inadequate.
Cleburne (1985): D1 Medium; D2 Disability; D4 Skeptical; D6 Stigma present. Tier ambiguity: formally rational basis, functionally higher.
Romer (1996): D1 High-ish; D2 Sexual orientation; D4 Pretextual; D6 High.
Adarand (1995) → Grutter (2003) → Fisher II (2016) → SFFA (2023): Same doctrinal words, different model selection. Adarand: strict scrutiny for all federal race classifications. Grutter: diversity = compelling, holistic review = tight fit. Fisher II: no workable race-neutral alternative = decisive. SFFA: re-weights D5, reclassifies prior permissive readings.
Parents Involved (2007): D1 High; D2 Race; D4 Educational diversity/integration = contested (compelling?); D5 Fit contested (race as predominant factor vs individualized consideration); D7 K-12 public education; D8 High conflict (plurality + concurrence + dissent; Kennedy concurrence controls under Marks rule). Illustrates ideological disagreement within constrained F_t — Justices accepted the constraint structure (race is suspect, compelling interest + tailoring matter) but selected different admissible (w,c).
Cases that pinned down doctrinal rules
Tiers of scrutiny:
- Carolene Products fn4 (1938): some classifications merit heightened review; discrete and insular minorities.
- Korematsu (1944): race = "suspect," most rigid scrutiny.
- Craig v. Boren (1976): intermediate scrutiny for sex.
- Adarand (1995): strict scrutiny for all governmental race classifications.
Elements of the EP test:
- Washington v. Davis (1976): impact alone ≠ EP violation; purpose required. Hard exclusion rule.
- Arlington Heights (1977): non-exhaustive factors for inferring intent. Feature selection.
- Feeney (1979): intent = "because of," not "in spite of."
Structure of strict scrutiny:
- Grutter (2003): diversity can be compelling; holistic review can satisfy tailoring.
- Fisher I & II (2013, 2016): courts must verify no workable race-neutral alternatives.
- SFFA (2023): open-ended diversity appeals insufficient. Regime-shift case.
Rational basis with bite:
- Cleburne (1985): animus fails even rational basis. De facto sub-tier.
- Romer (1996): broad disabilities on a single group lack rational relation.
Due Process doctrine
Institutional facts for the paper's DP application. Focus on how competing interpretive frameworks coexist — illustrating the model's admissible-set ambiguity.
Text
Fifth Amendment (1791): "No person shall ... be deprived of life, liberty, or property, without due process of law." (Federal government.)
Fourteenth Amendment (1868): "... nor shall any State deprive any person of life, liberty, or property, without due process of law." (States.)
Procedural on its face. "Liberty" is undefined. The clause does not specify which liberties are protected or how to identify them.
Doctrinal evolution
Original understanding (19th c.): procedural — lawful procedures, fair notice, neutral adjudication. Liberty = freedom from physical restraint.
Substantive due process emerges (late 19th c.): some deprivations are unconstitutional regardless of procedure.
Lochner era (c. 1897–1937): SDP protects economic liberties, especially freedom of contract. Lochner v. New York (1905). Later repudiated.
Post-New Deal (1937–): Court abandons economic SDP. Doctrine moves to personal/family autonomy.
Privacy line (1960s–70s): Griswold (contraception), Eisenstadt, Skinner (procreation). Liberty = intimate personal decisions. Methodology is normative/functional, not historical.
Roe v. Wade (1973): liberty includes abortion decision. Relies on continuity of principle, not historical practice.
Glucksberg (1997): restrictive framework — unenumerated rights must be "deeply rooted in history and tradition" and "implicit in ordered liberty." Privileges historical evidence. Coexists uneasily with privacy cases.
Two incompatible models coexist (1997–2022): autonomy-based (Griswold → Roe → Casey) and history-based (Glucksberg). Which applies depends on the case.
Dobbs (2022): adopts Glucksberg as controlling, rejects Roe's autonomy framework. Text unchanged; interpretation of "liberty" changes.
Dimension dictionary (for coding DP cases)
- D1 Liberty interest type: economic / family-autonomy / bodily integrity / privacy-intimacy
- D2 Historical grounding: deeply rooted in history and tradition? (high/low) — the Glucksberg axis
- D3 Level of generality: how abstractly the right is characterized (narrow/broad) — determines whether a claimed right qualifies as "deeply rooted"
- D4 Government interest strength: weak → strong/compelling
- D5 Intrusion severity: how substantially the government action burdens the liberty (low → high)
- D6 Methodology: history-based (Glucksberg) vs autonomy-based (Casey/Griswold) vs balancing (Mathews)
- D7 Institutional setting: criminal law, family law, medical, economic regulation, welfare/administrative, end-of-life
- D8 Precedent density / conflict: how "determined" the local doctrine is (low/high)
Coded cases
Mathews v. Eldridge (1976): Procedural DP. D1 Property (disability benefits); D4 Government: administrative efficiency = moderate; D5 Loss of benefits = significant but not liberty/life; D6 Balancing (established the three-factor test: private interest, risk of erroneous deprivation + value of additional safeguards, government interest); D7 Welfare/administrative. The canonical procedural DP holding — imposes structural constraints on admissible (w,c) by requiring positive weight on each factor while permitting tradeoffs among them.
Lochner v. New York (1905): D1 Economic (freedom of contract); D2 Court treated as fundamental (later rejected); D3 Broad (general economic liberty); D4 Government: public health = moderate; D5 Moderate (maximum hours law); D7 Economic regulation; D8 Low (early SDP). Imposed constraints privileging economic liberty; later repudiated — constraint removal at institutional cost.
Griswold v. Connecticut (1965): D1 Privacy-intimacy (contraception); D2 Contested (penumbras, not historical practice); D3 Broad (marital privacy → zone of privacy); D5 High (criminal prohibition on contraception use); D6 Autonomy-based; D7 Family/reproductive; D8 Low (novel right). Introduced privacy as a constitutionally protected liberty interest through autonomy methodology.
Roe v. Wade (1973): D1 Bodily integrity / privacy (abortion); D2 Low (no historical tradition of abortion right); D3 Broad (right to privacy encompasses abortion decision); D4 Government: fetal life + maternal health; D5 High (criminal prohibition); D6 Autonomy-based, continuity-of-principle from Griswold; D7 Medical/reproductive; D8 Moderate (built on Griswold line). Extended autonomy constraints; trimester framework as constraint structure.
Planned Parenthood v. Casey (1992): D1 Bodily integrity / autonomy (abortion); D2 Low, but reliance interests high; D3 Reframed as "liberty" not "privacy"; D4 Government: fetal life legitimate throughout pregnancy; D5 "Undue burden" replaces trimester framework; D6 Autonomy-based + stare decisis analysis; D7 Medical/reproductive; D8 High conflict (plurality opinion). Replaced Roe's constraint structure with new one (undue burden) while preserving core holding. Central stare decisis case — explicitly weighed overruling factors.
Washington v. Glucksberg (1997): D1 Bodily integrity (assisted suicide); D2 High threshold applied — no deeply rooted tradition; D3 Narrow (Court insists on specific characterization: "right to commit suicide"); D4 Government: preserving life, preventing abuse = strong; D5 High (criminal prohibition); D6 History-based; D7 End-of-life/medical; D8 Low-moderate. Established restrictive framework for recognizing new unenumerated rights — requires "deeply rooted in history and tradition" + "implicit in ordered liberty." Coexisted uneasily with autonomy line.
Dobbs v. Jackson (2022): D1 Bodily integrity (abortion); D2 High threshold applied — no deeply rooted tradition of abortion right; D3 Narrow (specific right to abortion, not broader autonomy); D4 Government: fetal life + democratic self-governance; D6 History-based (adopts Glucksberg as controlling, rejects Roe/Casey autonomy framework); D7 Medical/reproductive; D8 High conflict (overruling). Regime-shift case: removes Roe/Casey constraints from F_t at cost C, adopts Glucksberg methodology as controlling. Text unchanged; interpretation of "liberty" changes. Canonical example of overruling reshaping the feasible set.
Why DP illustrates the model
The clause is abstract, "liberty" is underspecified, and the Constitution does not specify the method for identifying unenumerated rights. Multiple interpretive approaches (autonomy-based, history-based) can each be fit to the precedent set. The coexistence of these approaches for decades, followed by a sharp doctrinal reversal when court composition changed, is a canonical example of bounded disagreement within an admissible set followed by a regime shift.
The DP dimension dictionary makes this visible: D6 (methodology) is the axis along which the admissible set was most contested. Both the autonomy approach and the history approach were consistent with existing holdings, but they generated sharply different predictions for new cases (e.g., D3 level of generality is decisive under history but irrelevant under autonomy). Dobbs resolved the ambiguity by removing autonomy-based constraints, collapsing a large region of F_t.
Comparative stare decisis
How do other legal systems handle horizontal stare decisis? Relevant for understanding what is generalizable about the model vs specific to the U.S. system.
TODO: fill from legal scholarship
- England: House of Lords / Supreme Court formally bound itself until Practice Statement (1966) allowed departure. How often has it been used?
- Australia: High Court can overrule itself but rarely does. What is the norm?
- Canada: Supreme Court of Canada can depart from its own precedents. Frequency?
- Civil law systems: formally no stare decisis, but de facto judicial precedent plays a significant role (Germany, France, Brazil). How does this informal binding compare?
- International courts: ICJ, ECtHR, CJEU — different stare decisis norms. Any useful comparisons?
- Key question for the model: is the binding force of precedent primarily a function of formal rules, or does it emerge from the same coordination/sanction dynamics regardless of formal rules?