Houston's 1947 Zoning Plan, Then and Now
Houston never adopted zoning. Comparing the city's 1947 plan to 2026 land use, more than half of the land it zoned for homes and for industry is used that way today, and business districts match three times as often as chance would predict.
Scroll to explore ↓
The plan
On 15 December 1947, the City of Houston published a Zoning District Map laying out ten cumulative districts, A through J — from First Dwelling to Heavy Industrial. Each district permitted every use allowed in the districts before it, plus more.
Houston never adopted this plan. Voters have rejected zoning multiple times since, and the city remains the largest in the United States without a conventional zoning ordinance.
Decoding the map
The districts are marked only by printed screen patterns — stipples, hatches, crosshatches — not color. We classified every part of the scan automatically, matching each patch's texture to one of the ten legend patterns with a random-forest classifier.
Checked against an independent, blind visual reading of 270 sample points, the automatic classification is correct for … of the mapped area at the Residential / Commercial / Industrial group level.
Houston today
To compare, we use parcel-level land use from the Harris County Appraisal District's tax-year 2026 records (valued as of 1 January 2026), grouped by the City of Houston into categories such as single- and multi-family residential, commercial, office, industrial, public and institutional, and park. It records what the appraisal district had on file for each parcel, which can lag changes on the ground.
Check our reading of the 1947 map
Left: the original 1947 map. Right: the districts our classifier read from its printed patterns. Drag the handle to see how well they line up. The reading is right for about 86% of the area at the Residential / Commercial / Industrial level. It is weakest on the business districts, where dense crosshatches are often confused with the hatches of apartment and light-industrial districts.
Plan versus reality
Drag the handle to compare the 1947 plan with 2026 land use anywhere in the study area.
Where the plan matches, and where it doesn't
Click anywhere in the study area to see the 1947 district and 2026 use at that spot.
The numbers
How much of each 1947 group's land matches its intended 2026 use — and how that compares with chance.
Match by group: intended use and permitted use
Intended use (solid dot) counts land whose 2026 use is what the district was meant for: homes in residential districts, businesses in commercial ones, industry in industrial ones. Permitted use (ring) also counts any use the cumulative 1947 ordinance allowed there, such as homes or shops in an industrial district. Whiskers are 95% bootstrap confidence intervals, the hollow grey marker is the chance baseline, and the shaded band is the range from 400 random shifts of the 1947 map.
Permitted use only beats chance for residential districts. Commercial and industrial districts allowed almost every use, so nearly all of their land "conforms" no matter where the districts are drawn.
Where 1947-zoned land ended up
Flows from each 1947 group to its 2026 land-use group, by acreage.
Match by district, A–J
Each of the ten 1947 districts, from First Dwelling (A) to Heavy Industrial (J). Solid bars: intended use (strict). Light bars: permitted use (cumulative).
How close is close?
If we deliberately shift the georeferenced 1947 layer sideways, agreement should fall — and it does, monotonically, which is independent evidence the registration is sound.
Method & caveats
- The 1947 districts were digitized automatically from a scanned paper map by texture classification, then checked against a blind visual reading of 270 stratified sample points (made by a single rater, an AI vision model directed by the analyst): about 86% correct at the group level — high for residential and industrial, but only about 42% of the area mapped as commercial is truly commercial on the 1947 map, so treat the commercial figures with more caution.
- The scan was georeferenced with 32 ground-control points; the registration's check-point error is about 54 ft, with roughly 30% of the study area lying outside the control-point hull and therefore less certain.
- Undeveloped, unknown, and vacant parcels and rights-of-way are excluded from every percentage's denominator.
- Every reported group difference beats 400 random spatial shifts of the 1947 layer.
Full method: methodology.md · Results detail: results.md · Code and data: github.com/jseppi/houston1947
Credits
- 1947 map scan: Zoning District Map, Houston, Texas, December 15, 1947 (City of Houston). David Rumsey Map Collection, David Rumsey Map Center, Stanford Libraries (CC BY-NC-SA).
- 2026 land use: Harris County Appraisal District (HCAD) tax-year 2026 appraisal land-use codes (valuation date 1 January 2026), as grouped by the City of Houston (Enterprise GIS) in its Land Use map service; retrieved 26 September 2026.
- Basemap: OpenFreeMap, © OpenStreetMap contributors (ODbL).
- Ground control: street centrelines from OpenStreetMap (ODbL).
- AI assistance: analysis code, map processing and this page were built with the help of Claude (Anthropic); see the AI-use statement in methodology.md.