National analysis · 32,779 federal decisions, 1996–2025

Raise or remove

When a house floods badly enough for the federal government to intervene, there are two outcomes: it gets lifted, or it gets bought and knocked down. Across 32,779 single-family homes, demolition won 2.7 times out of every one elevation. Which one your house gets is decided first by your state, and only then by what it stands on.

The record

Two outcomes, one lopsided ledger.

FEMA publishes every property its hazard mitigation grants have paid to alter or remove. Filtered to single-family houses facing water — where the choice was acquisition and demolition, or elevation — the file holds 32,779 decisions across 1,196 counties and 4,093 ZIP codes, spanning fiscal 1996 to 2025.

73.3% of them ended with the house gone. That is the first thing worth sitting with: the federal response to a repeatedly flooded home is usually not to make it survivable. It is to remove it and leave the lot empty.

The obvious explanation, and why it fails

Basements look decisive until you check whether it is really the basement.

Sorted by foundation, the national numbers are tidy. Homes with basements were demolished 93.4% of the time; homes already standing on piers or piles, 46.3%. It reads as engineering: you cannot easily lift a house with a hole under it, so those houses go.

That reading does not survive a simple check. Louisiana has pier-built housing and raises houses. Iowa has basements and demolishes them. If the states that build one way are also the states that choose one way, foundation would look decisive while explaining nothing.

So the same question has to be asked inside states rather than across them — and doing that splits the country before it says anything about foundations at all.

Five states, and everyone else

Elevation is not a national policy. It is a regional one.

Ranked by how often they demolish, the 25 states with at least 300 decisions do not form a spectrum. They form two groups with an empty gap between them: five states sit between 22.6% and 51.2%, then nothing until 71.5%, after which twenty states run up to 100%.

Highlighted rows are the 5 states that use elevation. The dividing line is drawn at 60%, and it can be moved anywhere between 55% and 70% without a single state changing group.

Eight states out of twenty-five demolish more than 97% of the time. Kansas and Oklahoma have never raised a single-family home with these grants — not rarely, never.

What the foundation decides

It matters enormously, and only in five states.

Share of homes bought and demolished rather than raised, by foundation type, within each group of states.

The order matters for anyone reading this about a specific house. Geography is the first filter and it is close to binary. Construction is the second, and it only gets a turn if the first one allows it.

Method

What was counted.

Source is FEMA's Hazard Mitigation Assistance Mitigated Properties file, 99,927 records, downloaded in bulk from the OpenFEMA API. Rows are kept where the structure is single-family and the action was either acquisition and demolition, or elevation — 32,779 of them. Safe rooms, wind retrofits, seismic retrofits, floodproofing, vacant-land purchases and non-residential structures are excluded: none of those is a choice between saving a house and removing it, and counting them would inflate the sample while blurring the question.

A state is classed as using elevation when it demolishes in fewer than 60% of its decisions, and only states with at least 300 decisions are ranked. Foundation shares are computed within each group rather than nationally, which is the whole point of the exercise. Fiscal year is the grant's year, not the year a house came down.

Limitations

What this does not show.

These are federally funded decisions only. States, counties and cities run their own buyout programmes, and homeowners elevate houses privately every year; none of that is here. A ZIP with no record is not a ZIP where nothing happened.

Nor is this a sample of flooded homes. A property appears because somebody applied for a grant and the application succeeded, so every figure describes the population of funded decisions and not the population of houses at risk. The direction of that bias is unknown.

Foundation type is blank or recorded as "other" on 2,518 rows, 7.7% of the set; those rows are excluded from the foundation tables and included everywhere else. A payment amount is present on only 47.8% of rows, which is why no cost-per-decision figure appears anywhere in this study.

One reading of the gradient is engineering rather than preference: lifting a house over a basement may simply be impractical, in which case the number describes a constraint and not a choice. This data cannot separate the two, and nothing here should be read as saying a state preferred to demolish a particular house.

FEMA buyouts are not an unexamined subject — the Natural Resources Defense Council and Climate Central have both published analyses of acquisitions. What appears to be new here is the cut by foundation type and the two-regime split, not the existence of the programme.

Appendix

Every ranked state.

<iframe src="https://www.beforeregret.com/research/raise-or-remove/embed/"
        width="100%" height="420" style="border:1px solid #e0dcd3"
        title="FEMA home buyouts and elevations by ZIP code"
        loading="lazy"></iframe>