Site sourcing and screening — Texas

What screening actually looks like in Texas

The structural constraint is identical to California's: a developer almost never closes on land before the subsidy award exists, because the 9% credit — or the 4% credit paired with a Texas Bond Review Board (TBRB) private-activity-bond reservation — is the capital that pays for the land. So the instrument is a long-dated option or contract with an extension ladder, not a 30-60 day escrow, exactly as in California. What's genuinely different is the menu of instruments Texas will accept, and how narrow it is. 10 TAC Section 11.1(122) defines Site Control as enforceable ownership or contract rights good enough to develop, operate, and record a LURA against the property, and 10 TAC Section 11.204(9)(B) lists exactly three acceptable forms — a recorded deed (or a lease with 45 years remaining), a lease-option with a minimum 45-year term, or a contract or option to purchase. That 45-year minimum lease term is a flat, Texas-specific number; California's instruments key to "the regulatory period," a variable. Texas has no equivalent to California's chain-of-agreements or eminent-domain site-control categories, and no HCD-style DDA/ENA instrument type carved out in the rulebook.

The distribution across those steps is as uneven as California's, just shaped differently. TCEQ's hazard layers and FEMA flood are queried in seconds. Zoning is not a data-freshness problem at all in a large share of Texas — it's a category question, since a real fraction of Texas land has no zoning classification to look up, by design, not by data gap. And the underwriting-adjacent numbers — this cycle's cost-per-square-foot thresholds, the income-and-rent tables — are refreshed annually against TDHCA's own moving calendar, the same panic-in-the-two-weeks-after-publication pattern California's brief documents.

Two decisions come out of this phase, same as in California, and Texas adds its own twist to the second one. The first is whether to pursue the site. The second is what price goes into the option or contract — and in Texas, that price does not just inform underwriting later, it becomes the underwriting number directly. Under 10 TAC Section 11.302(d)(1)(A), for a non-identity-of-interest acquisition, "the underwritten acquisition cost will be the amount(s) reflected in the Site Control document(s)" — no appraisal required to set it. That means the software-relevant check in Texas isn't a date-window problem the way it is in California; it's a data-consistency problem: does the Development Cost Schedule's acquisition line match the number actually written into the option or contract.

One more Texas-specific timing fact worth building into any Texas-facing calendar: the QAP itself is not fully settled until early December of the prior program year. The 2026 QAP was approved by the TDHCA Board on November 6, 2025, and approved by the Governor — with a substantive modification — on December 1, 2025. A mid-cycle correction in any given year is plausible, not hypothetical.

Zoning is not a code to read — it's a question of whether one exists

Texas counties have no general zoning authority. Zoning is a municipal-only power, and a permissive one: Local Gov't Code Section 211.005 says a city's governing body may divide the city into zoning districts — it never requires it. A large share of Texas land — every unincorporated acre in all 254 counties, plus every acre inside a city that has chosen not to zone — has no zoning classification to look up, by design. A Texas parcel screen's first zoning question isn't "what does the code say," it's "does zoning apply here at all," and treating a null answer as missing data rather than the correct answer is a real, first-order mistake.

Houston is the standing example, and it isn't trivia. It is reportedly the largest city in the United States with no zoning ordinance — voters have rejected it in three referenda, reported as 1948, 1962, and 1993. Land use inside Houston runs instead on the city's own subdivision/development code (setbacks, parking, lot size, plat approval) and on private deed restrictions, which by some published estimates cover something like a quarter to nearly half of the city's land area depending on definition — a figure no single source reconciles cleanly. In a no-zoning jurisdiction, a deed restriction can be the entire land-use control regime for a parcel, discoverable only by reading recorded instruments, not by pulling a zoning map. TDHCA's own rulebook treats this as a live threshold item, not trivia: 10 TAC Section 11.204(10) requires every Application to submit one of four zoning-status letters, including a "No Zoning Ordinance in Effect" letter for jurisdictions like Houston's — and that letter has to be refreshed annually, versus a six-month shelf life for the other three letter types.

Extraterritorial jurisdiction (ETJ) adds a wrinkle with no California parallel. A municipality can require plat and subdivision compliance in a ring around its borders — historically up to five miles for large cities — without having any zoning authority there at all. A parcel can sit inside a city's platting jurisdiction and simultaneously outside its zoning jurisdiction, and conflating the two is an easy, real mistake. Senate Bill 2038 (2023) added a further wrinkle: residents or a majority-in-value of landowners in a small area can now petition or vote to release land from a municipality's ETJ outright.

The one genuinely new fact for 2026 is Local Gov't Code Chapter 218, added by two 2025 bills effective September 1, 2025. It applies to any municipality with population over 150,000 that sits wholly or partly in a county with population over 300,000 — in practice, essentially every large Texas city. Where it applies, the city must allow multifamily and mixed-use residential as-of-right in any zoning classification that already permits office, commercial, retail, warehouse, or mixed-use development, and it may not impose density below 36 units per acre, height limits below 45 feet, parking above one space per unit, or a rezoning/variance/conditional-use step — approval has to be administrative, not a council vote. No dataset anywhere flags which zoning classifications, in which cities, now qualify; determining eligibility for a given parcel means reading the local zone code's permitted-use list and checking that none of the statutory exceptions (heavy industrial proximity, airport or military-base buffers, clear zones) apply.

The parcel layer is 254 appraisal districts, not one

Texas parcel records are held by a Central Appraisal District (CAD) — a special-purpose political subdivision distinct from county government, not a department of it. There are 254 counties and 253 CADs (Potter and Randall counties share one joint office), each with its own GIS maturity, and there is no state-mandated parcel schema and no single statewide parcel API.

One genuine positive surprise: Texas owner-of-record data is both legally and practically more open than California's. Tax Code Section 25.025 confidentially protects only a narrow, named, opt-in list — peace officers, jailers, and judges, among a few others — and only if that individual affirmatively elects confidentiality. Outside that list, a CAD's appraisal roll, including owner name and mailing address, is standard public information on essentially every county's own site, no records request needed.

The hazard layers are open here; the site-feature gate is a termination trigger, not a checklist

This is a genuine inversion of California's pattern, and it should be said plainly rather than assumed away. TCEQ's contamination layers — leaking petroleum storage tanks, Superfund, brownfields, active landfills, voluntary cleanup sites — are open ArcGIS REST services with CSV, shapefile, and GeoJSON downloads, with no bot-block. FEMA flood is the same federal layer California already uses, fully portable. That is a real structural advantage over California, where the equivalent state agencies (DTSC, SWRCB) return HTTP 403 to a scripted client.

But the layer that actually kills a Texas deal isn't a contamination database — it's written directly into the QAP itself, and it is stricter than California's process in a specific way. 10 TAC Section 11.101(a)(2), Undesirable Site Features, and Section 11.101(a)(3), Neighborhood Risk Factors, work as clear/mitigated/hard-exclusion gates, same as California's SB 35/AB 2011 exclusion list — except an undisclosed Undesirable Site Feature that staff finds after submission doesn't go to a discretionary mitigation process. It terminates the Application outright, with no Executive-Director-level judgment call available the way California's process allows.

Neighborhood Risk Factors are socioeconomic rather than physical, and one of them names a specific commercial data vendor directly inside the regulation: a census tract's Part I violent crime rate above 18 per 1,000 people "as reported on neighborhoodscout.com" for new construction in an Urban Area. The poverty-rate threshold is 40% (55% in TDHCA Regions 11 and 13). A third factor flags a school rated "Not Rated: Senate Bill 1365" by the Texas Education Agency — and the 2026 QAP's own text freezes that reference year at 2022, which may be a deliberate carry-forward or simply stale drafting; either way, it's worth flagging rather than treating as current on its face. All three factors are curable with documented mitigation, the same conditional-not-absolute pattern California's own screen follows.

What this means in practice: a clean TCEQ hazard screen answers a different question than a clean Undesirable Site Feature screen, and presenting one as proof of the other overstates what's actually been checked. Texas Historical Commission Atlas data on National Register listings and cemeteries appears open; whether precise archaeological site locations are separately gated wasn't independently confirmed, and habitat/listed-species data wasn't checked at all — the same honest gap California's own brief leaves for its own unpublishable layers.

De-concentration: the collision check no California screen needs

This is the single biggest structural difference between screening a site in Texas and screening one in California, and it operates as outright ineligibility, not a scoring adjustment. 10 TAC Section 11.3 codifies six distinct de-concentration rules, and a developer who signs an LOI without checking them is gambling on a fact that TDHCA's own application log already shows in black and white each cycle.

TDHCA's own 2026 cycle shows exactly how this resolves: McCart Meadows Senior Living beat View at Alta Mesa in Tarrant County at a tied score of 170; Jordan II Apartments beat Residences@Springdale in Travis County, 173 to 170; Vista at Culebra beat NHH Culebra in Bexar County at a tied score of 170. In every one of the three real conflicts, the loser didn't survive to compete further in the round at all. A developer who mapped competitors' pending sites before signing an LOI would have seen every one of these coming.

Where the screen stops, and what it cannot tell you

Construction cost is the same hole it is in California: there is no free, authoritative, unit-level Texas construction-cost dataset, and the federal BLS Producer Price Index gives escalation, not a level. What Texas has instead of California's Threshold Basis Limits is a scoring incentive, not a hard cap — 10 TAC Section 11.9(e)(2), Cost of Development per Square Foot, indexed annually to CPI-U.

Jurisdiction behavior is a genuine, structural absence, not a data gap to route around. California's HCD Annual Progress Report exists because state law requires every jurisdiction to report permits issued against a housing allocation, every year, in a standard format. Texas has no equivalent mandate and consequently no equivalent dataset — there is nothing to compute a jurisdiction-friendliness scorecard from, and inventing a proxy without saying so would misrepresent a real regulatory absence as a mere data gap. Per-project entitlement duration remains what it is in California: folklore, living in city permit portals and in land-use counsel's heads.

A parcel carrying an agricultural, open-space, or timberland special valuation triggers a rollback tax on change of use — the chief appraiser assesses the difference between taxes actually paid and what market value would have owed, plus interest. Since House Bill 1743 took effect September 1, 2019, that look-back is three years at 5% annual interest, down from five years at 7%. It's a real, common, entirely foreseeable acquisition-budget line item for exurban and rural Texas sites, and it surfaces late because it lives in the CAD's own valuation history, not in a title exception.

Water and sewer service runs on a different mechanism than California's priority-duty statute. A Certificate of Convenience and Necessity (CCN), issued by the Public Utility Commission of Texas under Water Code Chapter 13, gives one retail utility the exclusive right to serve a defined territory. The question for a Texas site isn't whether the utility will prioritize an affordable project — it's which CCN holder has the legal right to serve this specific tract, whether that holder has built capacity nearby, and whether closing the deal requires a CCN amendment or decertification proceeding at the PUCT, which is a state administrative process with its own timeline, not a request letter.

The genuinely good news, and worth stating because it cuts against the rest of this section: Texas's own award data is better than California's. TDHCA publishes its current-cycle 9% competitive award and waiting-list results as a real XLSX file, and the Texas Bond Review Board publishes its private-activity-bond reservations — the 4%/bond side, which is most of the Texas market — as a 27-sheet, multi-year workbook with archived history back to 2023 and project-level dollar figures. Neither is a cumulative, all-years comps table the way CTCAC's own List of Projects file is for California — TDHCA's 9% file covers one round at a time, with no parcel-identifying field — but both are real, structured, current data, not PDF scans.

The honest limit is the same one California's own guide draws. Whether the seller will actually extend the option, whether a deed restriction bars multifamily use, whether the CCN holder will agree to serve the site, whether a competitor is about to file within two miles — none of that is a dataset, and none of it should be synthesized. What screening can and should do in Texas is make sure the arithmetic questions are answered before the phone call: whether zoning even applies, whether a de-concentration conflict already exists on the map, whether the site clears the Undesirable Site Feature buffers, and what the acquisition price actually locks in once it's written into the site control document. EZFeasi has no Texas parcel, zoning, hazard, or award data loaded today — this entire phase is greenfield product work, not a refinement of an existing Texas feature.