# L5 — Recovery data search (metadata only)

**Job:** `lane-L5-recovery-data-search` · Hub `b73c292a` · Wave `7bea403e`  
**MODEL_CHANGE=NO** · **NO recovery scoring** · **Do NOT rescue Exp002**

Machine-readable: `data/lanes/L5/recovery_candidate_series_v0.csv` · `.json`

## Purpose

Identify societal variables with enough **temporal density/depth** for honest shock-response / recovery *concepts* (half-life, elasticity, baseline displacement, damping, persistence). This is a **source/density search**, not a recovery model.

## Exp002 standing reminder

Exp002 (L1 trust individual polls) remains **INSUFFICIENT** for confirmatory recovery compute. Sparse irregular polls + methodology mix → interval-censored half-life is not honestly identified under prior gates. **L5 does not reopen or rescue Exp002.**

## Family scoreboard (density for recovery concepts)

| ID | Family | Frequency | Depth | Shock detect | Recovery observe | Min episodes (order) | Density disposition |
|----|--------|-----------|-------|--------------|------------------|----------------------|---------------------|
| RC1 | Consumer sentiment (SCA ICS) | monthly | 1952–2026 | HIGH | HIGH | O(10+) plausible | **STRONG** |
| RC2 | Consumer sentiment (Conference Board) | monthly | ~1967– | HIGH | HIGH if licensed | O(10+) | **STRONG_IF_ACCESSIBLE** |
| RC3 | Economic expectations (SCA components) | monthly | 1950s– | HIGH | HIGH | O(10+) | **STRONG** |
| RC4 | SCE expectations | monthly | 2013– | MED-HIGH | MED | <10 likely | MODERATE_SHORT |
| RC5 | Institutional trust (L1 individual) | irregular | 1958–2025 | MED | LOW-MED | LOW under gates | **WEAK** (Exp002 lesson) |
| RC6 | Political approval (Gallup hist) | era-varying | 1938– | HIGH in tracking era | HIGH within era | O(10+) stratified | **STRONG_WITHIN_ERA** |
| RC7 | Approval aggregators (Silver/538-class) | ~daily | ~recent decades | HIGH | HIGH recent | few/decade | **STRONG_RECENT** |
| RC8 | Generic ballot averages | campaign-dense | 1990s– | MED-HIGH | MED seasonal | low off-cycle | MODERATE_SEASONAL |
| RC9 | Protest events (ACLED) | daily | modern US | HIGH | HIGH (event damping) | multiple waves | **STRONG_EVENT_DAMPING** |
| RC10 | GDELT events/tone | subdaily/daily | 1979/2015– | HIGH raw | fragile validity | many | MODIFY_CONSTRUCT |
| RC11 | Media narrative tone | daily | varies | HIGH | HIGH narrative | many topical | MODERATE_CROSS_LAYER |
| RC12 | VIX/markets | daily | 1990– | HIGH | HIGH math | many | STRONG_BUT_CROSS_LAYER |
| RC13 | Google Trends | daily/weekly | 2004– | HIGH attention | MED ≠ preference | many | MODIFY_ATTENTION |
| RC14 | ANES FTs/trust | election-year | ~1978–2024 | LOW | INSUFFICIENT | too few | **INSUFFICIENT** |

## Concepts × best families (metadata mapping only)

| Concept | Best density families | Hard no |
|---------|----------------------|---------|
| Half-life after shock | RC1/RC3 monthly SCA; RC6 within-era approval; RC7 recent | RC5/RC14 sparse surveys |
| Elasticity (response amplitude) | RC1/RC3/RC6/RC7 | spliced multi-mode series without break model |
| Baseline displacement | RC1/RC6 long spans with era controls | RC4 short panel |
| Damping of event activity | RC9/RC10 | attitude series alone |
| Persistence | RC1/RC6 long memory windows | campaign-only RC8 |

## Methodology-break watchlist

1. **Gallup approval:** discrete ↔ daily tracking (2009–2017) ↔ weekly (2018) ↔ discrete (2019–) — stratify or model breaks.  
2. **SCA:** phone→web 2024; early sparse months.  
3. **L1 trust:** multi-pollster mix — already fatal for Exp002 recovery honesty.  
4. **Aggregators:** house-effect model versions need pins for PIT.  
5. **ACLED/GDELT:** codebook/source drift.

## PIT availability

- **Good:** SCA month stamps; Gallup field windows; market dates; ACLED event dates.  
- **Conditional:** aggregator as-of snapshots (revision risk).  
- **Weak for recovery grids:** ANES study-year; L1 years with zero polls.

## Minimum realistic episode counts (orders only)

Honest recovery research needs predeclared shock rules *before* counting. Orders of magnitude above are **capacity guesses**, not episode ledgers. Do not treat them as preregistered N.

## Explicit non-claims

No recovery scores. No half-life estimates. No Exp002 rescue. No correlations. No S1/E2 reopen.