Methodology
What is observed, what is modeled, and why the site refuses to call modeled ZIP estimates “actual averages.”
1. Published premium anchor
The current seed uses the NAIC Homeowner Property Insurance Market Dynamics Report's 2024 average premium per policy by NAIC zone: Northeast $1,396, Midwest $1,476, Southeast $1,818, and Western $1,600. Those are published regional averages, not state or ZIP averages. Where a state-specific official observed premium feed is explicitly wired in, such as Texas TDI, that observed statewide figure becomes the state anchor while remaining clearly labeled.
2. Housing-value adjustment
Each state benchmark adjusts its regional anchor using median owner-occupied home value from the latest available ACS 5-year data. The exponent is deliberately sublinear (0.38) and bounded. Locality pages then use the same bounded adjustment relative to their state.
3. Severe-weather context
NOAA Storm Events records are summarized over the rolling configured window. State event frequency is a bounded secondary model adjustment. County data, when available, can slightly adjust local benchmarks. Storm records are not insurance claims.
4. FEMA hazard context
County-level FEMA National Risk Index scores and ratings are displayed as descriptive public risk context for composite risk and individual hazards. FEMA NRI is not used as a claimed carrier rating factor.
5. Property calculator scenarios
Home value, roof age and deductible settings apply transparent illustrative adjustments. They are not reverse-engineered insurer rating factors and should never be represented as a guaranteed quote or savings estimate.
6. THICI
THICI = state modeled benchmark ÷ population-weighted national modeled benchmark × 100. A score of 100 is the national model baseline.
7. Affordability and cost per $100K
Affordability = modeled premium ÷ ACS median household income. Cost per $100K = modeled premium ÷ median home value × 100,000.
8. Page-quality thresholds
To reduce thin programmatic pages, county, place and ZCTA pages are only generated above configurable population thresholds. Search engines receive crawlable static HTML and chunked XML sitemaps.
9. First-party premium data
User premium submissions are collected separately from public benchmarks. A SUBMISSIONS_CSV_URL can be configured in GitHub Actions to import an anonymized storage feed. Until sufficient samples exist, the site does not publish consumer medians.