Log 018: The City Panel Protocol — a Pre-Registered Design for the Corridor's First Field Test¶
Mode: Applied Architecture / Pre-registration
Status: PROTOCOL — frozen before data contact
Date: 2026-07-06
Scope: Operationalizes the Viable Corridor's three constraint axes at city scale and pre-registers directional hypotheses, per the-city-as-deployed-intelligence and paper §5.4. No data has been examined against these hypotheses. The git commit introducing this file is the protocol's timestamp — the pre-registration is enforced by the repository's own provenance mechanism (Log 017, eating its own dogfood).
Depends on: papers/viable-corridor.md (§4, §5.4, P8), theory/human-organism-silicon-age/the-city-as-deployed-intelligence.md, 017_provenance-depth-and-the-verification-economy.md
Why pre-register¶
The corridor's Class B claims (real-world systems) carry a measurement debt the paper states openly: \(\gamma_{\text{eff}}\), \(K_{\text{eff}}\), and substrate headroom are barely operationalizable for a civilization (\(N = 1\)). At city scale they are measured municipal quantities across thousands of comparable instances. But observational urban data offers endless researcher degrees of freedom — proxy choice, panel choice, outcome choice — and a framework this flexible could be "confirmed" by accident. The registered-report discipline (Chambers, 2013) is the honest instrument: freeze the choices before touching the data, publish the freeze, and let the commit hash prove the order of events. What follows is the freeze.
Constructs and primary proxies¶
Each proxy is named with its known threat to validity; alternates are listed for robustness, not for post-hoc substitution.
| Corridor construct | Primary proxy (US panel) | Threat to validity |
|---|---|---|
| \(\gamma_{\text{eff}}\) — regulation strength | Land-use regulation stringency (Wharton Residential Land Use Regulatory Index; Gyourko et al. 2008/2021) | regulation ≠ homeostatic regulation; stringency can encode incumbent protection rather than anti-concentration |
| \(K_{\text{eff}}\) — coherence / coupling | Economic connectedness & civic engagement (Social Capital Atlas, Chetty et al. 2022, county level) | social capital is an outcome as well as an input; reverse causation with prosperity |
| Substrate headroom | Infrastructure utilization margins: water-system capacity vs. demand; grid reserve margin where published | municipal capacity data is patchy and heterogeneous; missing-not-at-random |
| Viability outcome (primary) | Shock-recovery half-life: years for employment and population to return to pre-shock trend after a FEMA-declared major disaster or a documented major plant closure | trend estimation choices; shock severity varies — severity controls required |
| Vital-floor breaches (secondary) | Boil-water advisories, blackout-hours, homelessness rate changes | reporting heterogeneity across municipalities |
The outcome choice is deliberate: the corridor predicts robust viability — return after perturbation — not growth. City-IP, so to speak: identity persistence of vital function under a kick. This is the intervention logic of the benchmark, found in the wild: shocks are the divergence queries that reveal which cities had the constraint architecture and which had eleven green quarters.
Pre-registered hypotheses¶
- H1 (the conjunction). Recovery is predicted by the minimum of the three standardized constraint scores better than by any single score or by their mean. Model comparison: min-model vs. best-single-axis vs. additive, by out-of-sample fit. Falsifier: a single axis or the additive model matches or beats the min-model — the conjunction claim fails at city scale.
- H2 (capability loading). Pre-shock growth rate interacts with substrate headroom: high growth on low headroom predicts slow recovery and more floor breaches; growth with headroom does not. Falsifier: growth uniformly aids recovery regardless of headroom — capability would not load the substrate axis in the field.
- H3 (the unwatched axis). Instrumentation proxies (smart-city technology adoption indices) predict recovery only conditional on the coherence score; the main effect alone is ≈ 0. Falsifier: instrumentation alone robustly predicts recovery — the Meridian warning would be wrong where it matters.
Panel and analysis plan (frozen)¶
US metropolitan areas / counties; shocks 2000–2020 with ≥ 3 post-shock years observed; target \(N \geq 100\) shocked units. Pre-specified form: recovery half-life regressed on the three standardized constraint scores, their minimum, the H2 interaction, shock-severity controls (FEMA damage class), region and year fixed effects, clustered errors. Robustness: the listed alternate proxies, substituted one at a time and reported all, not selectively. Multiple comparisons: three hypotheses, stated once, no garden of forking paths beyond the listed robustness set.
What this protocol does NOT claim¶
- No causal identification. Observational panel; associations only. Corridor-consistent associations would be survival of a falsification attempt, not proof; their absence would be genuinely damaging (that is the point).
- No proxy = construct identity. Every row of the proxy table is contestable; that is why it is frozen and public rather than chosen after seeing residuals.
- No commitment that we run it ourselves next week. The protocol is also an invitation: any group with panel-data experience can execute it, and the commit hash guarantees they would be testing our predictions, not our postdictions.
Failure conditions for the protocol itself¶
- Any analysis of these hypotheses on these data before this file's commit — there was none; the hash is the warrant.
- Proxy substitution after data contact presented as the primary result.
- Reporting only the robustness variants that flatter H1–H3.
Anchors. Chambers (2013), registered reports; Gyourko, Saiz & Summers (2008) and Gyourko, Hartley & Krimmel (2021), the Wharton index; Chetty et al. (2022), Social Capital I/II (Nature); West (2017), Scale. Internal: paper §5.4 (the measurement debt this addresses at its tractable scale), the city node, Log 017 (the provenance mechanism doing the timestamping).