# SRP-RIGIDITY-01 — Elite Bloc Consolidation + Separation

**Experiment:** `SRP-RIGIDITY-01`  
**Status:** NEW_BOUNDED_EXPERIMENT — descriptive only  
**Agent (primary impl):** grok · **Alt-explanations (Step5):** claude · **Verify (Step6):** codex · **Synthesis:** astra  
**James auth:** `/james` message `996d837f-0c33-4d66-9d14-849cef60e1e2` (2026-09-10T18:05:19Z)  
**Hub job:** `b73c292a-16e2-4f11-9370-1337d0239791`  
**MODEL_CHANGE:** NO · **SCORING_CHANGE:** NO · **OUTCOME_TESTING:** NO  
**score_authorized(E2):** 0 · **No Rigidity Index / no SRP score**

---

## 0. Hypothesis (predeclared)

Over completed modern Congresses, U.S. political elites increasingly exhibit:

1. **LOWER** within-party dispersion / greater internal cohesion (**N1 DOWN**, **N2 inverse-unity DOWN**)
2. **GREATER** between-party ideological separation (**N3 UP**)

This is a test of **bloc rigidity / structural partitioning** — **not** instability, elections, causation, criticality, or production E2.

---

## 1. STEP 1 — SOURCE/ALIGNMENT GATE → **PASS**

### 1.1 Chamber / period

| Field | Value |
|-------|-------|
| Primary chamber | **HOUSE** |
| Sensitivity | **SENATE** |
| Candidate window | Congress **100–118** completed |
| Excluded | Congress **119** (provisional/ongoing) |
| N4 | **NOT USED** (denominator/edge under audit) |

### 1.2 Exact Congresses available

| Series | House | Senate |
|--------|-------|--------|
| N1 VoteView within-party SD (mean D+R) | 100–118 (19) | 100–118 (19) |
| N2 Brookings 100−unity (both calendar years required) | 100–117 (18) | 100–117 (18) |
| N3 VoteView median center distance | 100–118 (19) | 100–118 (19) |
| **Intersection** | **100–117 (n=18)** | **100–117 (n=18)** |

**Missing from intersection:** Congress **118** (N2 Brookings ends calendar 2022; Cong 118 = 2023–2024 has no N2 years). **No interpolation.**

**Gate rule:** ≥15 common completed Congresses → proceed. **n=18 ≥ 15 → PASS** (not `INSUFFICIENT_FOR_PRIMARY_TEST`).

### 1.3 Definitions (frozen for this test)

| ID | Definition | Rigidity direction |
|----|------------|--------------------|
| **N1** | `mean(SD(nominate_dim1\|D), SD(nominate_dim1\|R))` per chamber×Congress; VoteView; parties 100/200; min_n=5 | **DOWN** |
| **N2** | Mean of both Congress calendar years of `mean(100−unity_D, 100−unity_R)`; Brookings Table 8-4; Southern Democrats dropped; **both years required** | **DOWN** |
| **N3** | `\|median(nominate_dim1\|D) − median(nominate_dim1\|R)\|`; completed only | **UP** |

Congress C calendar years: `1789+2*(C−1)` and `+1`.

### 1.4 Source hashes

| Artifact | sha256 |
|----------|--------|
| VoteView raw `HSall_members.csv` | `aadf38febc4dad9e8f5cba7bac4c8290f2d96d550b34938c91ff5bd32c0be358` |
| Brookings `8-4.csv` | `724fc817638c58fa78a5e74aeb982f4b7e6f31e676cbfd3cd680dfee1924e840` |
| N1 cong≥100 CSV | `e0a3704602cc17b5441acaf6b7d05298cb8c6612ff2469a0edfa0e445bf0be89` |
| N2 100−unity prototype CSV | `38d9c47de8b12701c1be3f41b0e880e6207ab3235c98ea637569a0b7a3de7819` |
| N3 cong≥100 CSV | `1fdaff9a9f8f8bcd94992c4f08a6bcd579b29109d837e7b4ec58233e598c8d1c` |
| N3 summary JSON | `7a8fd3ca9b8621378201d9de68685e1d6f8f51eae6aac2ef0b73ba1c0fd8036e` |

### 1.5 Method breaks / caveats (preserved)

- N2 ends 2022 → Cong 118 dropped from intersection.
- N1 and N3 share VoteView/NOMINATE family (not independent-system replication).
- N2 is behavioral roll-call cohesion; N1 is ideal-point dispersion.
- Career NOMINATE can understate short-horizon swings.
- No composites / no sign flip after seeing results.

---

## 2. STEP 2 — DESCRIPTIVE TEST (House primary)

| Series | First (C100) | Last (C117) | Δ | Spearman ρ vs Congress | Endpoint OK? | ρ sign OK? |
|--------|-------------|-------------|---|------------------------|--------------|------------|
| N1 mean party SD | 0.160171 | 0.129643 | -0.030529 | **-0.9154** | True | True |
| N2 mean inv-unity | 16.25 | 3.25 | -13.00 | **-0.9602** | True | True |
| N3 center dist | 0.6660 | 0.8830 | 0.2170 | **0.9628** | True | True |

Early / late medians (first/last thirds of common sample):

| | Early (100–105) | Late (112–117) |
|--|-----------------|----------------|
| N1 | 0.156773 | 0.128538 |
| N2 | 14.000 | 4.875 |
| N3 | 0.6917 | 0.8840 |

---

## 3. STEP 3 — JOINT_RIGIDITY_CONFIGURATION (House)

Early-period reference = **first third** medians (Congresses **100–105**):

- ref N1 = 0.156773
- ref N2 = 14.000
- ref N3 = 0.6917

Later Congresses **106–117** (n=12): count how often **N1 < ref AND N2 < ref AND N3 > ref**.

| Metric | Value |
|--------|-------|
| n JOINT_RIGIDITY_CONFIGURATION | **12 / 12** |
| share | **1.000** |

**Not an SRP index. Not a coupling/synchronization claim.**

---

## 4. STEP 4 — TREND-ROBUSTNESS (descriptive only)

### 4.1 Thirds (House medians)

| Third | Congresses | N1 | N2 | N3 |
|-------|------------|----|----|----|
| Early | 100–105 | 0.156773 | 14.000 | 0.6917 |
| Mid | 106–111 | 0.140315 | 8.125 | 0.7910 |
| Late | 112–117 | 0.128538 | 4.875 | 0.8840 |

Monotone descriptive path early→mid→late: N1↓ N2↓ N3↑ (House).

### 4.2 Adjacent first differences (House, n=17)

| Sign | Count |
|------|-------|
| N1 Δ < 0 | 10 |
| N2 Δ < 0 | 9 |
| N3 Δ > 0 | 12 |
| All three rigidity signs same step | 5 |

**Do not interpret raw co-movement as coupling.** No lead-lag / predictive inference.

### 4.3 Senate sensitivity

| Series | Spearman ρ | Endpoint rigidity-consistent? |
|--------|------------|-------------------------------|
| N1 | 0.4221 | False |
| N2 | -0.8700 | True |
| N3 | 0.9835 | True |
| Joint share later vs early | 0.417 | — |

Senate N1 ρ is **positive** (dispersion not monotonically falling) → Senate is **weaker / mixed** on within-party consolidation even while N2↓ and N3↑. Primary claim remains **House**.

---

## 5. DECISION (Grok Steps 1–4 only)

**`SUPPORTED_DESCRIPTIVELY`**

Meaning (authorized gloss only):

> Elite data are descriptively consistent with increasing internal bloc consolidation occurring alongside increasing between-bloc separation.

**Does NOT establish:** instability, causation, prediction, criticality, cross-system propagation, elections/outcomes, or any SRP production score.

**Support-rule label correction (2026-09-10 post-synthesis):** The implementation-added rule
`joint_share ≥ 0.5` + six endpoint/Spearman sign checks is **WITHDRAWN as a confirmatory /
preregistered criterion**. It was not in James auth `996d837f`, has no demonstrated pre-result
freeze, and must not be cited as validation. Retained in history only.

`SUPPORTED_DESCRIPTIVELY` remains the honest Steps 1–4 label for the **House** endpoint/trend
pattern only (descriptive consistency, not confirmatory inference). Senate remains **MIXED**
(N1 ρ = +0.422 contrary). **GENERAL_ELITE_RIGIDITY_NOT_ESTABLISHED** (chamber asymmetry is a finding).

Measurement-family note: N1 and N3 share VoteView/NOMINATE; N2 Brookings is a separate statistic
with shared roll-call upstream. Do not claim three independent measurement families.

### Handoff (updated)

- Step 5 alternatives A–F: locked dispositions below / `/docs/srp-rigidity-01-alternatives-af.json`
- Step 6 Codex independent rebuild: computational PASS (max_abs_error ~2e-16)
- Astra synthesis: House descriptive / Senate mixed / general NOT_ESTABLISHED; RIGIDITY-02 not escalated yet
- Brookings 8-4 continuity: partial audit recorded (`/docs/srp-rigidity-01-brookings-continuity`)
- Chamber-asymmetry design draft only: `/docs/srp-rigidity-02-chamber-asymmetry-protocol` (no execution)

---

## 6. Artifacts

| Path | Role |
|------|------|
| `data/experiments/srp_rigidity_01/step1_alignment_gate.json` | Gate + hashes + definitions |
| `data/experiments/srp_rigidity_01/steps2_4_results.json` | Full descriptive + joint + diffs |
| `data/experiments/srp_rigidity_01/panel_house.csv` | House common panel |
| `data/experiments/srp_rigidity_01/panel_senate.csv` | Senate sensitivity panel |
| `data/experiments/srp_rigidity_01/summary.json` | Compact summary |

Existing E2 inputs (unchanged): `data/e2/voteview_*`, `data/e2/brookings_party_unity_100minus_prototype.csv`.

---


---

## 7. STEP 5 — ALTERNATIVES A–F (LOCKED DISPOSITIONS)

Owner-original alternatives from James auth + Claude adjudication. Status: **LOCKED for RIGIDITY-01**
(independently authored review is theory-exposed, not clean-blinded discovery). RIGIDITY-02 execution
is **not** authorized by this lock.

| ID | Alternative | Disposition | Discriminating future test |
| --- | --- | --- | --- |
| A | NOMINATE scaling / career-score artifact | PLAUSIBLE | Recompute N1/N3 with session-only or freshman-only scores; compare to career NOMINATE |
| B | Party sorting / realignment (composition change) | PLAUSIBLE | Recompute N1 after party-affiliation-change adjustment (esp. Southern realignment) |
| C | Roll-call agenda selection | PLAUSIBLE | Fixed vote-type / agenda-composition subsample; procedural vs substantive split |
| D | Chamber rules / institutional differences | PLAUSIBLE | Treat Senate contrary N1 as primary construct for chamber-asymmetry protocol (not nuisance) |
| E | Shared roll-call provenance / Brookings continuity | PLAUSIBLE | Map 8-4 vintage/normalization lineage by year; keep frozen hash; do not splice editions |
| F | Two-party sorting without broader system rigidity | PLAUSIBLE | Cross-system non-elite measures (explicitly out of RIGIDITY-01 scope); do not escalate here |

Machine copy: `/docs/srp-rigidity-01-alternatives-af.json`.

---

## 8. BROOKINGS 8-4 CONTINUITY (PARTIAL)

Codex/ChatGPT partial audit (`41ae1339`): Brookings Errata Sept 2026 PDF notes a **January 2017**
Table 8-4 correction — data from 2006 onward updated for absence-normalization alignment with
earlier years. Documented historical normalization revision; **current break in frozen 2024-path
CSV not proven**. Cross-format 2021/2022 endpoint agreement stands.

Frozen input hash unchanged: Brookings `8-4.csv` sha256
`724fc817638c58fa78a5e74aeb982f4b7e6f31e676cbfd3cd680dfee1924e840`.
Errata PDF sha256 `f0527b140bcbcf8747a195d1b9fc658d18b5ce1b10d2b58a350911ceb956a296`.

Next source task (bounded): map post-2017 additions for intended absence normalization; preserve
older vintages separately; **no recompute / no parameter adjustment / no score change** until
methods review closes the audit.

Detail: `/docs/srp-rigidity-01-brookings-continuity`.

---

## 9. CHAMBER ASYMMETRY (DESIGN ONLY)

Senate N1 dispersion **increasing** (ρ=+0.422) while House N1 falls is preserved as a **primary
finding**, not noise. Do not average chambers, redefine rigidity to drop Senate, or explore
Senate-only recoveries. Adjacent first-differences (House 5/17, Senate 1/17 all-three rigidity
signs) show endpoint trend ≠ smooth per-step process.

Design draft (no empirical run): `/docs/srp-rigidity-02-chamber-asymmetry-protocol`.
Astra direction: do **not** escalate RIGIDITY-02 execution until A–F are locked (done here) and
source-continuity audit progresses.


*Computed 2026-09-10T19:10:00Z UTC · PT = UTC−7.*