Methodology
What this site cannot tell you
Nothing about any individual property. Every figure here is an area-level median or a hazard rating for an area. It is not an appraisal, not insurance advice, and not a valuation. It says nothing about whether an area is a good or bad place to live.
ZIP codes and ZCTAs
We display “ZIP code” because that is what people type. We key on the Census ZIP Code Tabulation Area. ZIP codes are US Postal Service delivery routes, not areas; ZCTAs are census-tabulated areas that approximate them. They do not always share boundaries, and some ZIP codes have no corresponding ZCTA.
The crosswalk
FEMA publishes the National Risk Index at county and census tract grain. The ACS publishes at ZCTA. ZCTAs do not nest inside counties — 10,187 of them, 30.2%, touch more than one, and four touch six. Tracts do nest inside counties, and tracts and ZCTAs are both built from 2020 census blocks, so we join on tracts.
The Census relationship file gives the land area of each ZCTA-tract intersection and of the whole tract. It does not give the population of the intersection. We estimate it:
part population = tract population × (part land area ÷ tract land area)
That assumes population is spread evenly across a tract’s land. It is not, but tracts are drawn to hold roughly 1,200 to 8,000 people and are subdivided where density is high, which is what makes the assumption tolerable at this grain. It would not be tolerable at county grain. Every row records which method produced it.
Composite score is a population-weighted mean. Hazard ratings are categorical, and we report the rating held by the plurality of the ZCTA’s estimated population. We do not average ratings and we do not convert a weighted score back into one: FEMA does not publish its bin thresholds with the data, so deriving a rating would mean inventing it. Every rating shown is one FEMA actually assigned to part of that ZCTA.
Connecticut
Connecticut replaced its eight counties with nine Council of Governments planning regions as county-equivalents, and its census tracts were re-coded. The 2020 relationship file carries the old tract identifiers; the December 2025 NRI carries the new ones. The two sets describe the same 879 tracts and share no identifiers at all, so a direct join silently drops every Connecticut ZIP code. We bridge through 2022 county subdivisions. Those rows are marked with their own crosswalk method.
Uncertainty
ACS values are survey estimates and carry margins of error. We print them. In the 2020–2024 release, 56.5% of ZIP-code home-value estimates carry a margin of error greater than 10% of the estimate, and 7.8% carry one greater than 50%. A figure shown without its margin of error would misrepresent how well it is known.
Top-coded figures
The ACS caps median home value at $2,000,000 and median real estate taxes at $10,000, and reports the cap as an exact-looking number with no margin of error. Where that happens we show “$2,000,000 or more” rather than the literal, because the literal would state a precision the source does not have. Effective rates derived from a capped input are bounds, not point estimates.
What is missing, and why
A ZIP code is published only when it has enough data to be worth reading. Below that threshold the page still renders, says what is missing, and is withheld from search results. Puerto Rico’s 132 ZIP codes have ACS tax and value figures but the NRI reports insufficient data for every hazard there, so none is published as complete.