# L4 — Candidate sensors (KEEP / MODIFY / INSUFFICIENT)

**Job:** `lane-L4-boundary-permeability` · RESEARCH-ONLY · MODEL_CHANGE=NO  
**CSV/JSON:** `data/lanes/L4/candidate_sensors_v0.{csv,json}`  
**Preserve:** E2 N4 offer/attract denominator audit — **no composite**.

Disposition ∈ {KEEP_CANDIDATE, MODIFY, INSUFFICIENT}. Trend direction is **irrelevant** to selection.

## Summary table

| ID | Family | Sensor | Source | Disposition | E2-N4 overlap |
|----|--------|--------|--------|-------------|---------------|
| L4-A1 | A legislative coop | CBS **offer** out-party original cosponsorship rate (chamber×Congress unweighted member-mean) | CEL CBS / landed E2-N4 | **KEEP_CANDIDATE** | **PRIMARY reuse** — do not recompute |
| L4-A2 | A legislative coop | CBS **attract** out-party original cosponsor share | CEL CBS / E2-N4 | **KEEP_CANDIDATE** (paired, separate) | **SECONDARY reuse** — never average with A1 |
| L4-A3 | A legislative coop | Congress.gov independent raw rebuild of offer/attract | api.congress.gov lineage A | **MODIFY** | Same construct; replication gate still `API_KEY_MISSING` |
| L4-A4 | A legislative coop | Bipartisan bill enactment / LES bipartisan pathway counts | LES / CEL adjacent | **MODIFY** | Related cooperation outcome, not N4 rates |
| L4-B1 | B bipartisan networks | Repeated cross-party cosponsorship **edge** networks (Congress-indexed) | Congress.gov / CBS micro | **MODIFY** | Overlaps N4 microdata; different aggregation (edges≠member-mean rates) |
| L4-B2 | B bipartisan networks | Voluntary caucus co-membership cross-party graphs | Caucus directories / Holman–Mahoney-type | **MODIFY** | Low overlap with N4 aggregates; panel continuity weak |
| L4-B3 | B bipartisan networks | Compulsory committee co-service as “network” | Committee assignments | **INSUFFICIENT** as permeability | Forced co-presence ≠ permeable choice; literature: often weak/negative vs voluntary ties |
| L4-C1 | C cross-party social | Cross-party friendship / discussion-partner items (Pew / YouGov / AFS-type) | Survey houses | **MODIFY** | **No E2-N4 overlap** (mass layer) |
| L4-C2 | C cross-party social | ANES social network / discussant party items (when present) | ANES CDF | **MODIFY** | Sparse / irregular; not S1 FT substitute |
| L4-D1 | D institutional reciprocity | Offer vs attract **kept separate** + documented reciprocity gap (attract−offer) as descriptive only | E2-N4 audit | **KEEP_CANDIDATE** (descriptive pairing rule) | **Is** the N4 audit — preserve contrary directions |
| L4-D2 | D institutional reciprocity | Selective-reciprocity lawmaking designs (Volden–Wiseman / Dobson-type) | Academic replication | **MODIFY** | Theory-adjacent; not a landed continuous SRP series |
| L4-D3 | D institutional reciprocity | GSS / generalized trust as elite reciprocity | GSS | **INSUFFICIENT** (wrong construct) | **S2 corroboration only** — forbidden as L4-D |
| L4-E1 | E civic interaction | Bridging civic org membership / cross-cutting associations | GSS/civic batteries; specialized bridging surveys | **MODIFY** | Density uneven; construct ≠ elite |
| L4-E2 | E civic interaction | Deliberative / bridging-program participation counts | Program admin data | **INSUFFICIENT** (coverage) | Selection into programs; not population panel |
| L4-X1 | guardrail | VoteView between-party center distance | E2-N3 | **INSUFFICIENT** for L4 | Polarization/separation — not permeability |
| L4-X2 | guardrail | Party-unity / within-party NOMINATE SD | E2-N1/N2 | **INSUFFICIENT** for L4 | Cohesion/dispersion LEVEL — not cross-boundary traffic |
| L4-X3 | guardrail | Out-party feeling thermometer / affective gap | S1 | **INSUFFICIENT** for L4 | Affect ≠ interaction |

## Notes on KEEP rows

1. **L4-A1 / L4-A2 / L4-D1** lean on the already-audited CEL CBS chamber series (`data/e2/cbs_chamber_congress_offer_attract.csv`) and `n4_definition_audit.json`.  
2. Zero-activity exclusions, original-only filter, unweighted member-mean, and **no party-opportunity reweight** stay as in the N4 audit.  
3. House offer ↓ with attract ↑/flat (98→118) is a **feature** for construct separation, not a bug to composite away.

## Explicit non-claims
No L4 score. No offer+attract index. No E2/S2/L1 score changes. No correlations with outcomes/transitions.