Civic Data Works

Civic Data Works · the law · undeveloped land

What counts as “undeveloped land” under Sec. 104?

The statute requires every CDBG grantee to publish a searchable database of all parcels of undeveloped land it owns, effective October 1, 2026. It never says what “undeveloped” means. This is the first question every city hits, and the answer decides which parcels go on the list.

The words Congress used

“…the grantee maintains, on a publicly accessible website, a searchable database that identifies all parcels of undeveloped land owned by the grantee.”

Housing and Community Development Act of 1974 §104(b)(7), 42 U.S.C. 5304(b), as added by Public Law 119-101 · read the enrolled bill

That is the whole requirement. There is no definition of “undeveloped,” no field schema, no minimum parcel size, and no exclusion list. As of mid-2026 HUD has published no guidance, no rule, and no model database. The provision is self-executing, so the reading is the grantee’s to make — and to write down.

Why the definition is the whole job

Almost every city already holds both halves of the answer. The county assessor knows which parcels are recorded in the city’s name, and the same records usually carry some signal of whether anything is built on them. What is missing is the intersection — a published list of the parcels that are both city-owned and undeveloped.

We checked 240 CDBG entitlement cities against the four statutory tests, including complete censuses of Texas, Ohio, and Florida. None of them clears all four. In 137 of those 240, the specific gap we recorded includes the undeveloped-or-vacant dimension: the city publishes what it owns, but nothing in the public view tells a resident which of those parcels are empty. Every per-city finding is published, with its method, on the compliance tracker.

Three concrete shapes of that gap, all verified against the live data:

In all three cases the honest description of the work is publish what you already have — not build a database from scratch.

The three signals cities actually have

There is no national vacancy dataset. In practice a grantee has one of three signals, and each is wrong somewhere specific.

Improvement value = $0 The assessor’s improvement (building) value on the parcel. Where the county carries improvement values on tax-exempt property — Texas and Florida both do — a zero is a strong undeveloped signal, and it is the cleanest test available. Fails in states that zero out improvement values on exempt government parcels. Ohio does this: the test flags every city-owned parcel as undeveloped, which is worse than no test at all because it looks like it worked.
Land-use / use code A coded classification such as “VAC” or “00” on each parcel. Cheap when it is real. Fails when the code list is assumed rather than checked — legends are mislabeled, codes get reused, and a value that reads as “vacant” in one county means something else in the next. We published one such claim ourselves and had to retract it. Query the actual distribution of values in the live layer before you rely on any code.
Structure footprints Intersect each parcel with a building-footprint layer — FEMA’s USA Structures is national and free — and treat zero structures as undeveloped. This is the fallback where assessor values are useless, and it is how we handled Ohio. Requires parcel polygons. Testing a centroid’s bounding box instead of the real polygon produced 23 false positives out of 25 in our own first pass. Use the geometry, not a point.

Categorize, never omit

The statute says all parcels. The tempting move is to drop the awkward ones — right-of-way slivers, parcels with no coordinates, parcels whose status you cannot determine — because they make the list look messy. That is the one mistake that can make an inventory statutorily insufficient. Our own first build dropped 39% of qualifying parcels this way before we caught it.

The defensible pattern is to publish every parcel and label why it is on the list:

A reader can then see the whole universe and the reasoning. A reader looking at a filtered list cannot tell the difference between “we own nothing else” and “we left things out.”

Writing a reading you can defend

Because there is no federal definition, the deliverable that matters in a monitoring visit is a short written statement of the standard you applied. In practice that is four sentences:

  1. The ownership test — how you decided a parcel is owned by the grantee. Note that §104(b)(7) binds the grantee, so leased, trustee-held, and authority-held parcels need an explicit call.
  2. The undeveloped test — which signal you used and why it is appropriate in your county.
  3. The exclusions — what you categorized rather than published as undeveloped, and the rule you applied.
  4. The refresh cadence — how often the published database is rebuilt from source.

Write it once, publish it beside the database, and keep it with the CDBG certification file. If HUD later defines “undeveloped,” a documented reading is a paragraph to update rather than a rebuild.

Related