FINANCEFIRST RESEARCH DATA KIT v1.1

Methodology and review status



What FinanceFirst calculated


We transcribed GAO’s aggregate Table 1 and checked every row and column total. The combined count is 12,797,264 + 241,976 = 13,039,240. Its share is 13,039,240 ÷ 16,029,433 × 100 = 81.3456%, shown as 81.3%. Keeping only the mapped-outside group in the numerator gives 79.8% of the same denominator. This sensitivity does not test a different flood model.


What the evidence cannot establish


We did not obtain First Street’s proprietary property records, reproduce its risk model, inspect homes or conduct interviews. The counts are model-derived properties, including residential and nonresidential, not current households. Model uncertainty remains even though the arithmetic uses integer counts. We have not constructed confidence intervals.


The NFIP coverage estimate is GAO’s rounded finding, not a FinanceFirst calculation. GAO’s Figure 1 note reports a 79.2% contract match rate and adjustment for unmatched contracts; Appendix I explains the matching stages and limitations. Private property-level coverage was unavailable. Different reference dates and incomplete coverage information prevent a current count of wholly uninsured households.


How the invented loss model works


For each insured category: payment = min(limit, max(eligible adjusted loss − deductible, 0)). Remaining cost = building replacement damage + belongings replacement damage + living costs − modeled payments. Replacement bills and eligible insured values are separate inputs. The formula is an educational simplification, not a complete implementation of the NFIP Dwelling Form or a private contract. See the exclusions next to the examples and the downloadable assumptions.


Review status


This report was prepared with AI assistance. No qualified human insurance or legal review has been recorded. The claims still requiring that review are the application of lending exceptions and prior-assistance obligations, and the interpretation of actual policy valuation, deductibles, exclusions and loss-settlement conditions. Code verification establishes arithmetic consistency, not how an insurer would settle a claim.




DATES
GAO source release: October 1, 2026. FinanceFirst first publication: November 11, 2026. Substantive revision and research check: November 11, 2026. These dates are distinct from observation periods.

REPRODUCTION
Extract research-data-kit.zip and run python reproduce.py with Python 3. Use this filename exactly; calculate.py is not required or included. No additional packages or network access. The script reads data/raw, compares calculations with data/derived, checks all six scenarios, and writes verification/results.json.

COLUMNS
See data/README.txt for units, zero/missing conventions and definitions. Raw means transcribed GAO aggregate counts or invented inputs, not underlying proprietary records. Numerical scenario inputs did not change from v1.0.
