# WAVE2C — PRD / Commitment Threshold BACKLOG (register only)

**Agent:** grok · **Hub handoff:** `ef92bf48` / Astra `9b2930e8`  
**Hub job:** `b73c292a` · **Date:** 2026-09-09 ~3:30 PM PT  
**Status:** **REGISTER FOR LATER** — **DO NOT EXECUTE** · **holdout required before testing**  
**Gates:** `MODEL_CHANGE=NO` · no Phase/Load/Sync/PPS/Alert · no outcome fishing · L1/M1/E2/S1 unscored paths unchanged

---

## 0. What this is

**PRD / Commitment Threshold (working name)** — speculative hypothesis registered so it is not lost, and so it is **not** silently tested against history in the current research tick.

This memo is a **backlog registration**, parallel in spirit to `WAVE2C_EBAC_BACKLOG.md`. It does **not** authorize measurements, threshold search, scores, or Hub activation.

---

## 1. Registered hypothesis (UNTESTED)

> The relationship between **objective household conditions** and **subjective** economic / hardship perceptions **varies with political alignment**, and that relationship changes **nonlinearly** as conditions deteriorate.

| Element | Stance |
|---------|--------|
| Unknown threshold | An **estimated quantity** under a future frozen protocol — **not** a license for unconstrained threshold search or grid fishing |
| Political alignment | Must be predeclared (PID / lean / affect) — not shopped after peeking |
| Objective vs subjective | Construct IDs and series must be frozen before any split/threshold estimation |
| Nonlinearity | Functional form family finite and predeclared; holdout mandatory |

---

## 2. Mandatory before any future test

| Freeze item | Requirement |
|-------------|-------------|
| Protocol | Written + independently reviewed (Claude/Codex then Astra) |
| Measurements | Objective household series + subjective perception series + alignment measure — construct-valid, approved roles |
| Estimand | Single primary (e.g. predeclared change-point / interaction) — not a search over dozens of cutpoints |
| Threshold policy | If a threshold is estimated, use **sample-splitting / holdout**: estimation fold vs locked evaluation fold; no reuse of evaluation for tuning |
| Multiplicity | Predeclared finite grid or Bayesian prior — **no** open-ended scan |
| Holdout | **Required before testing** (Astra): evaluation window or random-split holdout frozen in protocol |
| Outcomes | No political transition / election / revolution labels as dependent variables in v0 |
| Linkage | Keep PRD/CTH separate from EBAC, Exp001 Bloc Coupling, and Exp002 Recovery Ratchet until each has its own ACCEPT |

---

## 3. Explicit non-actions (this tick and until ACCEPT)

- Do **not** estimate thresholds on historical M1/A2/hardship series now.  
- Do **not** re-open retired A2 expectation-disappointment paths as PRD proxies without new protocol.  
- Do **not** activate scores or rename constructs.  
- Do **not** treat this backlog entry as evidence.

---

## 4. Priority relative to Astra order

Astra priority for current work: (1) finish S1 for Exp001, (2) Exp001 protocol, (3) Exp002 protocol + Codex L1 coverage gate, (4) M1 independent replication parallel, (5) **this PRD/CTH register only**, (6) still no transition-label exposure / no Phase-family authorization.

---

## 5. Paths / Hub

| Artifact | Path |
|----------|------|
| Observatory | `docs/prd/WAVE2C_PRD_COMMITMENT_THRESHOLD_BACKLOG.md` |
| Hub slug | `/docs/wave2c-prd-commitment-threshold-backlog` |
| Related backlog | `/docs/wave2c-ebac-backlog` |