# Requested Deliverables: Evidence and Remaining Gaps

Edition: September 11, 2026, review revision 2. The article is finalized for review; this does not mean production deployment or completion of every desired causal test. A missing measure is not a null result. The owner authorized commit and push, not merge or deployment.

| Requested deliverable | Current evidence | Status and limitation |
|---|---|---|
| 1. Source inventory | Hash-bound acquisition receipt, public inventory, bibliography, failed ERCOT request | Available; source-page redistribution rights are not assumed |
| 2. Feasibility matrix | This matrix and the six-hypothesis story contract | Available |
| 3. Reproducible pipeline | Bounded acquisition, parsers, model scripts, figure generation, tests, exact-hash recovery | Available; exact raw rebuild requires retained bytes if mutable upstream sources change |
| 4. Cleaned datasets | Census construction, county permits/QCEW, BEA, EIA, CSO, DESNZ, CBS, ABS and ONS context | Available; missing/suppressed values remain explicit |
| 5. Data dictionary | Field names, units, denominators, source and transformation rules | Available |
| 6. Methodology Markdown | Accounting, panel, matching, inference, source semantics and limitations | Available |
| 7. U.S. national analysis | Construction with/without DC; ownership shares; nominal BEA decomposition; volume/price comparison | Partial: matched floor area and realized permanent jobs per dollar unavailable |
| 8. U.S. county/metro econometrics | 694-county frame; 25 specifications; five dynamic profiles; two dated matched cases | Partial: observational, not identified causal effects; no exhaustive MW inventory or full requested control set |
| 9. Ireland deep dive | 2015-2025 metered electricity, non-DC remainder and revised dwelling completions | Available descriptive evidence; local connection/permit causal linkage unavailable |
| 10. UK deep dive | GB 2020-2024 meter-matched local/national shares; ONS investment classification review | Available descriptive evidence; enterprise centers and Northern Ireland excluded; DC investment not separately identifiable |
| 11. Germany/EU comparison | Frankfurt primary planning policy; EU reporting boundary; Netherlands dedicated-connection series | Partial: no harmonized German investment/electricity panel or policy discontinuity estimate |
| 12. Australia/Canada comparison | ABS building/equipment categories; AESO Alberta requested loads versus interim limit | Available bounded cases; neither is a harmonized national DC share or causal comparison |
| 13. 12-15 interactive figures | Fifteen native figures with source, period, units, data status, exact tables and CSVs | Implemented locally; rendered QA recorded separately |
| 14. Uncertainty and robustness | County-clustered intervals, pre-trends, weights, controls, period/mega-cluster sensitivity and event-date alternative | Partial: no valid scarcity interaction/IV, geographic spillover correction or exhaustive treatment-date robustness |
| 15. Bibliography | Primary statistical definitions, operator/municipal evidence, labeled secondary project date | Available |
| 16. Interactive publication-ready story | Native five-act article, local preview, URL state and research package | Local research edition; final runtime/accessibility/performance status in validation report. No deployment authorized |

## Hypothesis Coverage

| Hypothesis | Strongest available test | What remains unresolved |
|---|---|---|
| H1 Pure addition | Measured non-DC construction and non-DC electricity both remain positive/growing over selected periods | Their unobserved counterfactual path |
| H2 Construction crowding-out | County permits, contractor pay/employment, matched pre-2019 cases | Occupational worker transfers, equipment scarcity, local housing demand and all requested controls |
| H3 Grid crowding-out | GB local concentration and AESO reliability-based connection limit | Which other applications were delayed or displaced, full large-load queue time series |
| H4 Land crowding-out | Frankfurt's adopted planning response to commercial-site competition | Land-price panel, absent projects and causal policy discontinuity |
| H5 Capital crowding-out | BEA nominal asset decomposition, ONS ownership distinction | Loan-level allocation, lending constraints, alternative financed projects |
| H6 Complementarity | Power construction, EIA generating capacity and Midlothian utility works | How much additional capacity was caused by the DC investment and who pays |

## Non-substitutions

- The 2023 national-era profile is not a project-start event study.
- The two dated cases are descriptive matched histories, not an exhaustive natural experiment.
- A chain-volume investment index is not square meters of floor space.
- QCEW hosting establishments are not MW or physical campus counts.
- Contractor-industry pay is not the electrician occupation's wage.
- Commercial electricity is not entirely data-center electricity.
- A requested load is not operating consumption or a generation interconnection queue.
- An unavailable German, Australian or Canadian share is not zero.
- The pipeline does not identify a number of homes, factories or infrastructure projects forgone.

## Next Evidence That Would Change the Answer

Acquire a phase-resolved project panel with dated site work and energized MW; match it to utility load applications, delayed/withdrawn projects, consistent local construction values/area, occupations, population, income, unemployment, housing prices and land/zoning constraints. Preregister treatment and comparison rules before inspecting post-event outcomes. Evaluate spillovers and few-treated-cluster inference before using causal language.