Four analyses of US public housing-risk data, each built from named federal files and published with the figures behind it. They exist because the questions were ours before they were anyone else's: we kept needing numbers that nobody had put together, so we put them together.
Whether US homeowners insurance premiums track the modelled risk of the county they cover.
Across 3,093 counties, modelled hazard explains part of what people pay, but two counties facing the same modelled risk can differ by more than double.
Flood is excluded from standard homeowners policies. How many people bought the separate one?
In the typical county, roughly one home in seven inside a mapped flood zone carries an NFIP policy, and take-up does not track flood risk.
The flood map decides who is required to buy insurance. What happens to everyone else?
More than one paid flood claim in four came from outside the mapped high-risk zone. In Texas it was about half.
What the national dam inventory records about condition and emergency planning, read together.
Of 17,049 dams classified high hazard potential, about one in six carries a condition rating of poor or unsatisfactory. 636 of those also have no emergency action plan on file.
What the United States actually does with a flooded house, across 32,779 federally funded decisions.
Demolition beats elevation 2.7 to 1. Five states do nearly all the raising, and only in those five does the foundation under the house change the outcome.
Half the housing in two North Texas counties went up inside one twenty-year window, in the heaviest hail corridor in the country.
Screening all 100 counties for both concentrated construction and heavy hail leaves three standing, and two of them share a border. Roofs there do not age one house at a time.
Allegheny County records more severe weather than almost anywhere, and insures for less than almost anywhere. Why those are the same fact.
Across the 100 most populous counties, recorded storm frequency and the share of households paying top-band premiums are effectively unrelated (Spearman -0.101). Allegheny is the extreme case: 3rd highest storm count, 6th lowest share paying over $3,000.
Every study publishes its county or state table as a CSV, its full figures as JSON, and an interactive county lookup you can embed on your own site. All of it is free to use with attribution, and the embeds carry their own credit line, so nothing needs to be asked for.
Each study also states what it cannot tell you, and where its own totals do not reconcile. If a number here does not match something you have, that section is the first place to look, and if it still does not add up we would rather hear from you than not.
Corrections, questions and data requests: hello@beforeregret.com.