Methodology

Every figure on LandBrief traces back to a county record or to a computation described on this page. This is the description: where the data comes from, what we do to it, and — just as important — what we refuse to publish and why.

Where the data comes from

The records are county assessor and GIS exports — parcel boundaries with their attribute tables — delivered county by county. Each county’s dataset carries the date of the assessor roll it was cut from, and that date is published as the “data as of” vintage wherever the county’s records appear. Some counties deliver fresh data every year; a few of the rolls we hold date back years. The data sources page shows, live, how many counties each state has on file, how many parcels that is, and how recent the sources are.

County names and boundaries come from the Census

County deliveries arrive under folder and file names typed by people, and names typed by people drift. Every county name is therefore canonicalised against the Census Bureau’s TIGER database before anything loads, and a name TIGER does not recognise is refused rather than guessed at. County outlines on maps come from TIGER as well. This is also why Virginia’s independent cities appear under their real names — Norfolk city, not the Norfolk County that Virginia abolished in 1963.

How parcels get their city

Nearly half of the counties in our archive record no city on their parcels at all — the field simply is not in the county’s export. For those counties we place each parcel by its geometry: a point-in-polygon test against Census incorporated-place boundaries assigns the parcel to the city it physically sits in. A parcel inside no incorporated place is listed under Unincorporated, and on this site that word means something: it is a measured answer, not a default we fell back to.

Two kinds of acreage

A parcel’s size can come from two places: the acreage the county recorded, and the area of the mapped boundary itself, computed from the geometry. Where counties publish both, the two agree closely — across counties that record acreage, the median ratio of computed to recorded is almost exactly 1.0 — but they are not the same number and we never silently merge them. A deed measured to the centreline of a road is larger than the mapped polygon; a condominium’s units each sit on the same footprint. Wherever a size is computed from the boundary rather than recorded by the county, it is labelled that way.

When a county’s own figures are impossible

Some counties record acreage figures that cannot be true. The clearest case: Travis County, Texas — home of Austin — records a combined 6,048,943,111 acres across its parcels. The county physically contains about 636,360 acres; the recorded total is roughly 9,500 times the land that exists. This is not a units error that a conversion would fix — the inflation is uneven across parcels, so no constant factor recovers the truth.

We check every county’s recorded acreage against the Census-measured land area of the county itself. Where the recorded total exceeds three times the physical land area, we withhold recorded acreage entirely rather than publish it: those counties rank their parcels and owners by parcel count instead, and parcel sizes shown there are computed from boundaries. At the time of writing this affects 78 counties. The three-times threshold is a judgment call and it deliberately errs toward publishing — there are counties whose totals are merely improbable rather than impossible, and those still publish, flaws and all, because the alternative is us silently editing county records.

Partial data is treated as partial

A county that records acreage for only some of its parcels poses a subtler trap: a ranking drawn only from the covered parcels looks complete and is not. Clark County, Nevada records acreage for about one parcel in seven — a “largest parcels” list built from that subset would omit the county’s actual largest tracts. For any county whose acreage coverage is materially incomplete, rankings blend the county’s own figure where it exists with the boundary-computed area where it does not, and every computed row is flagged as computed.

When one parcel ID is many parcels

Hundreds of counties publish parcel identifiers that repeat — one county in South Carolina files more than 82,000 parcels under 648 identifiers. Where an identifier is shared, its page lists every parcel that carries it and each parcel gets a stable key derived from its own geometry, so distinct properties stay distinct even when the county’s numbering does not distinguish them.

How pages are produced

LandBrief’s pages are generated by software from the records and computations described above: the numbers on a county page are queries against the county’s data, not text anyone typed. The explanatory prose — this page, the about page, the descriptions that frame the numbers — is written and reviewed by people. No statistic on this site is generated by a language model, and no page invents a figure its underlying records cannot support: where data is missing, the page says so or omits the section rather than filling the gap.

Corrections

If a number here disagrees with the county’s official record, we want to know — the contact page reaches us. What happens next is described in the editorial policy.