Sources and provenance

Compiled for Levenhall LLC. Every substantive figure in the atlas traces to one of the sources below. Where a source is a primary state publication, that is noted; where a claim rests on secondary reporting or on a program administrator's own marketing, that is noted too, because the two do not carry the same weight.

Tier 1 — primary state publications

Sales tax — extraction method and validation

Parsed programmatically from the OTC quarterly COPO chart. The chart is a multi-column layout in which the rightmost column of the final pages carries *lodging* tax rather than sales tax; the parser is column-position aware so that lodging figures do not overwrite sales figures. 77 county rates and 567 municipal rates extracted; 520 matched to municipalities in the atlas. Municipalities without a matched rate do not levy a city sales tax — per the chart's own rubric, "if a city or county is not listed, they do not have a sales or use tax."

Combined rate is computed as state 4.5% + city rate + county rate, where the county component is included only when the chart marks the city ** (county tax also due). Spot-validated against independently known figures: Broken Arrow 8.417%, Tulsa 8.517%, Oklahoma City 8.625%, Norman 8.75%. Four municipalities are matched through a name alias where the OTC chart and the census name differ — most importantly Pryor Creek → Pryor, which is MidAmerica Industrial Park's municipality.

Rates change quarterly. This set is 3Q 2026 (July–September). Re-extract each quarter.

NAICS eligibility schedules

Encoded from the qualifying-industry schedules in the Commerce guide for all three Quality Jobs programs, plus derived mappings for the non-Quality-Jobs programs that are keyed to industry. Matching is by prefix in both directions, so a 6-digit user code matches a 2-digit statutory sector and vice versa. The schedules differ materially between the three programs — a code listed under one does not imply the others, and the engine reports each separately.

Wage tables — extraction method

Both county wage tables were extracted programmatically from the published PDFs rather than transcribed. All 77 counties reconciled against the independent county list with zero discrepancies. The one formatting artefact — the guides print "Le Flore", the county list uses "LeFlore" — was normalised. Tables are effective 1 February 2026 to 31 January 2027.

CFIUS distances

Great-circle distances computed from installation centroid to site centroid. These are indicative only. The regulation measures from the *installation boundary*, so real figures are shorter than those shown. Tulsa at ~93 miles from Tinker AFB sits close enough to the ~100-mile threshold that only a precise, parcel-level measurement will settle it. Do not rely on the screen for a transaction decision.

Tier 1b — per-state programme sources (national layer)

Every state's flagship programme catalogue is sourced from that state's own economic development agency and revenue department, plus the governing statute where one was identified. Oklahoma remains the only jurisdiction at full coverage — primary-source catalogue, county wage tables, municipalities, sites and a hand-built evaluator. Every other jurisdiction is at headline coverage: flagship programmes only, with financial figures modelled from published mechanics rather than from a determination.

Catalogue depth — two tiers, counted separately

Oklahoma carries 82 programmes because it was built from primary sources. Every other state carried between five and ten: the flagship programmes an agency leads with, not its catalogue. California, the world's fourth-largest economy, carried nine against a published state list of sixty-nine.

Every one of the 51 jurisdictions has now been worked through against the administering agency's own published programme list. The atlas carries 1,348 state and territory programmes plus 10 federal — 458 full profiles and 890 catalogue records. The thinnest jurisdiction carries 14 and the median is 21; the counts are what each agency actually publishes, not a quota. Where an agency's site refused to serve a scripted request (Arizona, Michigan, Ohio, Vermont, Colorado, Massachusetts, Alaska), the programme list was recovered through a domain-scoped search of the same site rather than left short.

Programmes are therefore held at two tiers, and the audit counts them separately on every build so the depth of coverage stays visible rather than being implied:

A catalogue record carries no mechanics by construction, and tools/audit.js fails the build if one ever does — so deepening the catalogue can never quietly change a modelled figure. The California merge was verified against this directly: the state went from 9 programmes to 70 with the best lawful stack unchanged at $37.81M, the same five programmes in the same order.

Two editorial rules govern what is carried:

• A programme belongs if a company siting or expanding a facility could plausibly use it. Pennsylvania's Department of Community and Economic Development publishes 139 funding programmes; 64 are carried. Housing, homelessness, school facilities, residential weatherization, dam safety, sewage, tourism marketing and the expired COVID-era relief programmes are not — a business establishment atlas listing a vaccine outreach grant is worse than one that does not.
• Repealed programmes are not carried. Georgia's Port Activity Tax Credit was repealed effective 1 January 2026 by the Georgia Economic Growth and Tax Relief Act (HB 463) and is excluded rather than listed as available. Alaska's tax credit index likewise carries fourteen credits marked expired or repealed — the Cook Inlet jack-up rig credit, the film production credit, the frontier basin credit and others; the ten still in force are carried and the fourteen are not.

Rebuild a state's catalogue with python3 tools/merge_catalogue.py <code> <source.json>. The merge never modifies or removes an existing programme, and adds a record only where the name matches nothing already present.

The merge's name key is deliberately loose but it is not a synonym dictionary: it cannot see that "New Economy Jobs Program" and "New Economy Jobs Tax Credit" are one Delaware programme, or that North Carolina's "Utility Account" and "Industrial Development Fund — Utility Account" are the same grant. Every pair of programme names within a state was therefore compared token-by-token after the merges and reviewed by hand; 82 records were removed with python3 tools/drop_programmes.py <ids.json> — 81 duplicates under a second wording, and Georgia's Port Activity Tax Credit, which was carried as a modelled full profile despite the rule below. Removing it lowered Georgia's modelled best stack from $6.53M to $5.94M. Every other state's figure was unchanged, which is what the two-tier rule is for.

Workforce — BLS QCEW

The atlas priced incentives but said nothing about whether a project could be staffed or what labour would cost — which is the factor state recruitment packs lead with, ahead of incentives. That layer is now built from the Quarterly Census of Employment and Wages, 2025 annual averages, published by the Bureau of Labor Statistics on 26 May 2026.

Kalawao County, Hawaii is the single county with no figures. Its population is under a hundred and QCEW does not disclose employment that would identify an individual employer. It is shown as undisclosed rather than as zero, and the county count in the panel says so.

Two cautions the panel states on its face. Average annual pay is across every industry and level, so it is the going rate in that county and not the wage for any particular role. And employment concentrates far harder than pay does — a large pool is a hiring market and a competitive one, because every employer in it is bidding for the same people.

The project brief

A state economic development agency will not engage on a description. It sends a project information form and asks for it back: jobs and average wage, capital investment split four ways, site and building specifications, utility demands, rail and truck movements, a timetable, and the factors driving the decision.

The form's shape is the convention rather than one state's preference. It was taken from the Maryland Department of Commerce's *FDI Project Information 2026*, and independently confirmed against Walmart's own supplier site-selection guide, which advises companies to prepare the same categories before approaching any agency.

The guided plan now ends with that form, filled as far as the atlas can fill it — operation type, NAICS, entry route, preferred county, new jobs, average wage, derived annual payroll, total capital, rail and port needs taken from the constraints the reader flagged, the modelled incentive value, and the states weighed against this one, which the compare tray on the national map has already recorded.

Fields the atlas cannot know are printed as outstanding rather than guessed. Fifteen of them, consistently across all fifty-one: the company and contact block, the capital split, the building and site specification, utility volumes, the year-by-year hiring profile, financing and board approval, and the timetable dates. That is the point of the document as much as the filled half is — it is the checklist of what to settle before the first call rather than during it.

What states say about themselves

State economic development agencies publish recruitment packs. They are written to recruit, and the rankings in them are chosen by the state. Where the atlas carries one, it is reproduced as the agency's claim, attributed and dated, never as a finding of this atlas — and where a claim touches a figure the atlas independently holds, that figure is printed beside it so the reader can see whether the two agree.

A check resolves one of three ways, because a claim can be wrong, right, or right and incomplete:

Held for all 51 jurisdictions, 242 claims in total, each sourced from that state's own agency and dated; the file names the document and its URL. tools/emit_edo.py regenerates them from one dict, so adding or refreshing a state is a data change rather than a code change.

Twenty-nine of those claims touch a figure the atlas independently carries and are checked against it at render time. Twenty-two agree. Two differ — Georgia states a corporate rate of 4.99% where the atlas holds 5.19%, and Pennsylvania quotes 7.99% where the atlas holds the current 7.49%, both cases of an agency quoting a target or a superseded vintage. Five are true but incomplete, and those are the useful ones:

There is no common format, and no fifty-state equivalent of a data book. Three groups: a structured pack (Maryland, Texas, Florida), a page carrying real figures (Utah, Nebraska, Massachusetts, Michigan, Illinois, Indiana and others), and agencies whose site publishes nothing consolidated, where the claims had to be gathered from the agency's own statements and press material. Claim counts therefore run from two for Oklahoma to eight for Maryland, and the panel shows what that state actually publishes rather than padding to a fixed number.

Three obstacles are recorded because they will recur on any refresh. The substance is often not linked from the page that advertises it — Florida's own "why Florida" page carries two rankings while why-florida-fast-facts.pdf sits unlinked on the same server. Arizona, Michigan, Illinois and New York return 403 or a Cloudflare challenge to a fetcher and to a headless browser alike; their claims were recovered from the search index, which carries the page content the site refuses to serve directly. Minnesota redirects automated requests to a bot-validation endpoint.

One entry is unusual and worth flagging. Oregon's card carries the agency's own account of the state's weaknesses — 3,000 semiconductor jobs shed in 2025 and candid remarks about cost of living and perceptions of Portland — taken from Business Oregon's semiconductor ecosystem review rather than a recruitment page. It is included because it is the same agency speaking, and a pack that only ever flatters would be a worse guide.

Tier 2 — reference datasets

Population and county-assignment data are convenience reference. For a filing or an incentive application, use the figure the administering agency uses.

Tier 3 — program administrators and secondary sources

Used where no primary state publication covers the point. Verify directly with the administrator before relying on any of these.

• Enterprise Zone doubling of the Investment/New Jobs Tax Credit to 2% of investment or $1,000 per employee — Oklahoma Department of Commerce, *Investment/New Jobs Tax Credit Package*. Also applies where investment exceeds $40M. This is the single most consequential addition to the catalogue that the main Commerce guide does not spell out.
• MidAmerica Industrial Park — scale (9,000+ acres, ~80 employers, 4,000+ employees), Quick Action Incentive Fund, Enterprise Zone status, on-site Union Pacific rail and switchyard, on-site airport, customized training — MidAmerica's own site-selection materials.
• Iron Horse Industrial Park / Citizen Potawatomi Nation — ~700 acres on trust land, A-OK Railroad connection, FTZ #106 magnet site, Honoring Nations award — CPN materials and KGOU reporting (Dec 2025).
• PartnerTulsa — TIF districts (Downtown, 11th and Lewis, Peoria Mohawk Business Park, 36th St North at MLK), infrastructure grants, five-year abatements.
• Greater Oklahoma City / City of OKC — Strategic Investment Program thresholds of 50 full-time jobs or $1.75M new annual payroll. The city's published wage example is out of date; confirm the current benchmark directly with the city.
• Lawton — land grants, fast-track permitting, forgivable job loans, Southwest Rail Industrial Park (2,500+ acres) — local EDO materials.
• Ardmore Development Authority — discretionary grants, loans and land based on jobs and capital.
• Choctaw Nation Business & Economic Development — Choctaw Development Fund (created 2017), 10½-county territory (growchoctaw.com).
• Cherokee Nation — TERO (tribal resolution October 1983, ordinance 11 August 1984); Tax Commission business licensing, which requires Cherokee citizenship *and* ownership or lease of restricted Indian land *and* location within the Nation's jurisdictional boundaries. Commerce Services: SSBCI award of $86M+ for Native-owned small business lending (max loan $5M, project cost ceiling $20M); $500,000 USDA Rural Business Grant. Contact sbac@cherokee.org, (918) 453-5536.
• Muscogee Nation — Department of Commerce; Contracting and Employment Support Office (CESO), created by the Contracting and Employment Support Act of 2019. A TERO-equivalent constituted by tribal statute rather than carrying the TERO name — do not assume a conventional TERO fee model.
• Chickasaw Nation — Chickasaw Business Network: Small Business Development Center, Apelachi Business Incubator, Business Directory, Preferred Vendor Program, Bid Board. Preferred Vendor Program established consistent with U.S. DOT DBE regulations.
• Cheyenne and Arapaho Tribes — Law and Order Code Title 1, Ordinance 6147001. The only Oklahoma nation in this atlas publishing an explicit TERO fee basis: covered employers other than construction contractors, with 20+ employees on tribal-jurisdiction lands or gross sales of $100,000+, pay an annual fee of 0.5% of employees' annual payroll — assessed on payroll, not on contract value. TERO office: P.O. Box 67, Concho, OK 73022, (405) 422-7564.
• Comanche Nation — TERO vendor licensing administered through the Tax Commission, not a standalone TERO office. 584 NW Bingo Rd., Lawton, (580) 492-3240.
• Osage Nation — Osage LLC, holding company for multiple 8(a)-designated, HUBZone-certified tribally owned businesses. Note that Osage County is coterminous with the reservation and the Osage Mineral Estate is held in trust, severing mineral from surface rights county-wide.
• Citizen Potawatomi Nation — CPCDC tribal CDFI lending to the Nation's citizens nationwide and to all American Indians in Oklahoma; FireLake enterprise group; A-OK Railroad.
• Opportunity Zones — Oklahoma's 117 current QOZs including 66 rural; 413 eligible "on-list" tracts of which 188 rural; nomination window opening 1 July 2026 with designations effective 1 January 2027 — Oklahoma Department of Commerce and Oklahoma Municipal League.
• McGirt line of cases — *McGirt v. Oklahoma* (2020); *Stroble* (Okla. 2025, holding McGirt did not displace state income tax authority); U.S. Supreme Court denial of certiorari (April 2026).
• Sovereign immunity — *C & L Enterprises, Inc. v. Citizen Band Potawatomi Indian Tribe of Oklahoma*, 532 U.S. 411 (2001).
• STEP Fund — up to $24,000 per company under STEP 12, for activities between 1 October 2024 and 29 September 2026. This grant period is expiring; confirm the successor round.

The NPV model

assets/model.js produces indicative figures for ranking alternatives. Its assumptions, stated plainly because they matter:

• Everything is reported as net present value at a discount rate the reader sets, defaulting to 8%. Incentives do not arrive as a lump sum — Quality Jobs pays quarterly for up to ten years, an ad valorem exemption saves tax each year for five — so each future payment is divided by (1 + rate) once per year of waiting before being summed. The rate is the reader's own cost of capital, and it is an input rather than a constant because it decides which programme wins: between 8% and 20% a programme paying over ten years loses about 38% of its modelled value while a front-loaded one loses about 25%, so a high rate favours cash now over credits later. The same Oklahoma project is worth $12.91M at 0%, $9.93M at 8% and $7.12M at 20% — identical statute, identical eligibility, only the assumed value of time changed. Both the national map and the state model default to the same 8%, so a project priced on one is priced on the other.

• Quality Jobs family is modelled at the statutory maximum rate (5%, 6% with veteran uplift, 10% for 21st Century). The actual net benefit rate comes from a Commerce cost/benefit analysis and is frequently lower. Treat modelled QJ figures as a ceiling.
• Ad valorem exemption is modelled flat on initial investment at a user-set assessment ratio and millage. In reality the base declines with depreciation, so the model overstates later years. It is discounted from year 2, reflecting that the exemption period begins 1 January following the year the facility becomes operational.
• Manufacturers sales tax exemption is modelled as a one-off saving on a user-set share of capex treated as exempt-eligible equipment. Default 60%. The exemption does not reach administration, sales, distribution, transportation, site construction or site maintenance.
• Tax credits are modelled at face value. A credit is worth nothing without Oklahoma income tax liability to absorb it. The model does not test for liability — that is a company-specific fact.
• Stack enumeration walks every subset of eligible programs, discards any containing a pair joined by an edge in the exclusion graph, ranks the survivors by NPV, and suppresses stacks that are strict subsets of a higher-ranked one.

Known correctness detail: Quality Jobs + Investment Tax Credits is encoded as excluding *both* standalone Quality Jobs and the standalone Investment/New Jobs Credit, because it is the combined programme replacing them rather than an addition to them. Without that edge the engine double-counts.

Tribal profiles — the data honesty rule

data/tribal.js carries a status on every one of the 38 nations:

• sourced (10 nations) — fields taken from the nation's own published material or a named secondary source.
• contact-only (28 nations) — a verified route to the office is held, but programme detail has not been verified. Fields are absent rather than guessed.

This distinction is deliberate and is surfaced in the interface, including a "sourced profiles only" filter. TERO fee schedules and tribal business-licensing procedures are frequently unpublished and obtainable only by telephoning the office. For a product being licensed to clients, a visible gap is worth more than a plausible invention — a wrong TERO fee quoted into a bid is a real loss.

Where a fee schedule is genuinely published it is shown as such and quoted; where it is not, the interface says so explicitly rather than falling back on the 1–5% industry range.

Competitor state comparison

data/compete.js compares Oklahoma against Texas, Arkansas, Kansas and Missouri. Every row carries a confidence marker rendered in the interface:

• verified — from a named primary source
• reported — credible secondary source; rates move, re-check before relying
• structural — a feature of the regime rather than a number

The strongest genuinely defensible argument the comparison surfaces is structural rather than numerical: Oklahoma's Quality Jobs family pays cash, while Kansas HPIP and Arkansas Advantage Arkansas are income tax credits. A new entrant, or a foreign-owned entity with little in-state taxable income, extracts full value from the former and often very little from the latter. Headline percentage comparisons systematically miss this.

Known gap: right-to-work status and foreign land ownership rows for the peer states are marked structural and are directionally correct but were not verified statute by statute for each state. Missouri's 2018 referendum rejection of right to work is well documented and is the one that matters.

Map geometry and chart colour

County boundaries are derived from the us-atlas counties-10m TopoJSON (US Census cartographic boundary files), filtered to FIPS 40* and projected to a local equal-aspect plane scaled by cos(mean latitude). All 77 counties present; Le Flore normalised to LeFlore to match the rest of the dataset. Total path payload ~20KB, so the map ships inline with no tiles and no network calls.

Chart colour follows the data-visualisation method rather than the brand palette, deliberately:

• Choropleth uses a single-hue sequential blue ramp, six classes, light→dark for low→high on the light surface and dark→light on the dark surface — so in both modes the lightest-contrast step means "near zero" and recedes toward the surface. Six classes stays under the ~7-class limit beyond which adjacent bins blur.
• Bar comparisons are single-series magnitude — one colour for every bar with the subject emphasised, never a value-ramp across nominal categories (which would double-encode length as hue).
• Every chart has a table-view twin already in the atlas: the county table for the map, the program-by-program list for the model chart, the side-by-side table for the power chart.
• The brand teal is reserved for interface chrome and emphasis, so a data colour never impersonates a UI state.

Flags, emblems and branding

Levenhall marks are the firm's own assets, taken from the Levenhall Website repository (levenhall-logo-full.png, levenhall-wordmark-white.png, favicon) and used under the Levenhall Brand Guidelines v2026.06. The atlas palette and type stack now follow that document: Deep Navy #0F1F3A and Gold #C9A961 as the master pair, Burgundy #8B1538 as the editorial accent, Fraunces for display, Source Sans Pro for body, JetBrains Mono for data. Cyan #38D9C9 is reserved for the AIGIS product family and is deliberately unused here.

Oklahoma state flags are original vector art drawn from documented descriptions, not traced reproductions. The 1911–1925 original (scarlet field, white star outlined blue, "46") is accurate and simple. The current flag is an explicitly *simplified* rendering for interface use — the Osage shield, seven feathers, calumet and olive branch are stylised. For any published or official use, obtain the standard from the Oklahoma Historical Society.

Tribal flags and seals are deliberately absent. Three reasons, recorded in data/flags.js:

1. Each of the 38 nations is a sovereign government; its flag and seal are official emblems of that

government, not decorative assets.

1. The Cherokee, Chickasaw, Choctaw and Muscogee nations, among others, publish policies expressly

restricting reproduction of their seal or flag without written permission. A licensed commercial product needs written permission, not fair-use reasoning.

1. Rendering 38 sovereign emblems from memory would produce inaccurate versions. An inaccurate

sovereign emblem is a worse outcome than none.

In their place the atlas uses a neutral typographic medallion carrying the nation's initials, which is not derived from any nation's emblem. If the flags are wanted in the product, the path is to write to each nation's communications or executive office and request the emblem plus permission for the specific use. That is a correspondence exercise, not a research one.

Municipal flags. Only Oklahoma City (redesigned 2021) and Tulsa (1973, George Vlosich) are recorded as having official flags. The overwhelming majority of Oklahoma's 597 municipalities have never adopted one — "every city's flag" is not a set that exists. Both are subject to their city's own reproduction policy and are described rather than reproduced.

Known limitations

1. Federal Indian country tax provisions are expiring extenders. Accelerated depreciation under

IRC § 168(j) and the Indian Employment Credit under IRC § 45A have repeatedly lapsed and been retroactively reinstated. Both are flagged in the catalogue. Confirm enactment for the applicable tax year before modelling either.

1. Oklahoma's foreign land ownership regime is live law. SB 1582, SJR 31 and SB 1672 in the

2025–2026 sessions have variously proposed removing exemptions, adding forced-divestiture timelines, and applying restrictions uniformly. Take current advice.

1. Automatic 5% counties and Opportunity Zones are reviewed annually. The atlas describes the

statutory tests rather than listing current designations, because a list would go stale. Verify any specific address against the Commerce Business Incentives map at okcommerce.gov/maps-geographic-data.

1. Municipal requirements are not catalogued. Oklahoma has 597 municipalities, each setting its

own occupational tax, licensing and zoning rules independently. The atlas tells you the category of obligation exists and that you must clear it locally; it does not enumerate 597 ordinances.

1. Covered ports. CFIUS jurisdiction attaches to real estate that is, is within, or will

function as part of a covered port, independent of installation proximity. Whether the Tulsa Port of Catoosa or Port of Muskogee is a listed covered port has not been confirmed here.

1. 38 versus 39 tribal nations. 38 is the count of federally recognised tribes headquartered in

Oklahoma. Sources reaching 39 generally add the Euchee (Yuchi), associated with the Muscogee Nation and not separately federally recognised. The atlas lists 38.

Not legal or tax advice

The controlling authority is always the statute and the administering agency. Incentive eligibility turns on facts this atlas cannot see, and several of the largest programs are mutually exclusive in ways that make a wrong assumption expensive. Verify before you rely.

Data vetting — what was checked, and what was found

Every jurisdiction, county and programme record was put through an automated integrity audit and a link check. This section records what the audit tests, and the defects it found and fixed, because a claim of accuracy is worth nothing without the method behind it.

What the audit checks, on every build

tools/refcheck.js runs inside build.py and stops the build if any name used as a reference fails to resolve: a municipality naming a county that does not exist, a council or a county naming the other, a county naming a nation absent from the roster, a duplicate or mis-prefixed FIPS code, a county count that disagrees with the registry, or a municipal sales tax whose parts do not sum to its combined rate. These never raised an error before — a broken name silently dropped the row, which is how Le Flore County came to sit in no council.

Content is checked separately by rendering all fifty-one atlases and reading what actually appears in every panel, rather than by reasoning about the templates. Oklahoma's statutes, agencies, programme names, phone prefixes and characteristic figures are searched for in the fifty pages that must not contain them.

Defects found and fixed

Tribal roster provenance

The roster is parsed from the Bureau of Indian Affairs notice *Indian Entities Recognized by and Eligible To Receive Services From the United States Bureau of Indian Affairs*, published 30 January 2026 (document 2026-01899). The parse yields 573 distinct entities. The notice itself states 575; the difference is four cross-reference entries — Arctic Village, Village of Venetie, St. Paul Island and St. George Island — which the notice lists as pointers to a parent tribal government that is itself listed. They are excluded rather than double-counted.

Each entity is assigned to the state of its tribal headquarters. Where the Federal Register entry names the state, that name governs. Where it does not — 103 entities — the assignment follows the BIA Tribal Leaders Directory. Entities whose name begins with a state word that belongs to the tribe rather than its location (the Delaware Tribe of Indians, headquartered in Oklahoma; the Sac & Fox Nation of Missouri, located in Kansas and Nebraska) are assigned explicitly rather than by pattern.

Link health

All distinct programme source URLs are fetched on demand. Those that return 403 to an automated client — Arizona Commerce, georgia.org, Idaho Commerce, maine.gov, Empire State Development, michiganbusiness.org, Justia and the Tennessee revenue help centre — are behind bot-protection and were confirmed to render in a real browser. No source link in the catalogue is known to be dead.

Deepened states — September 2026

States taken from headline to extended coverage: state-specific entry pathways, compliance calendar, headline tax facts and agency contacts read from the state's own agencies; appraised sites with a CFIUS screen; and a programme catalogue deepened from the state's official incentive index. The financial model still values these programmes from published mechanics — extended is not the hand-built, statute-verified evaluator Oklahoma carries, and the masthead tier says so. Every figure below traces to the URL beside it.

Texas (TX)

Texas — sources and provenance

Compiled for Levenhall LLC, US Business Establishment Atlas, Texas deepening. Every substantive figure in data/states/tx/pathways.js, data/states/tx/sites.js and data/states/tx/incentives.js traces to one of the sources below. Each record carries the URL actually read as source and the retrieval stamp 2026-09-18. Where a primary page did not publish a figure, the record describes the point qualitatively and carries verify: true.

##### Tier 1 — primary state and federal publications

##### Tier 2 — official pages of public site owners, EDCs and tribes (owners' own statements)

##### CFIUS distances — method

Great-circle distances from an approximate installation centroid to an approximate site centroid, computed locally. They are indicative only: the regulation measures from the installation boundary, Joint Base San Antonio is a multi-site installation, and Sherman (≈130 miles from Tinker AFB) and the Corpus Christi inner harbour (≈130 miles from JBSA) sit close enough to the threshold that only a parcel-level measurement will settle them.

##### Known limitations

• Registry gap. The 2025 CFR edition of Appendix A lists Fort Cavazos, Killeen, TX in Part 1; the atlas registry's Texas cfius list omits it. OK_CFIUS.installations carries it with a note; data/_registry.js was not edited.
• statutes.capitol.texas.gov now serves a JavaScript shell to non-browser clients, so Tax Code §§ 11.251, 11.43, 22.23, 151.318, 151.359, 151.429, 171.002 and Gov't Code chs. 2275 and 2303 were not read from the statute site. Every figure attributed to them was taken instead from the enrolled bill text at capitol.texas.gov, the Comptroller's own forms and pages, or the Governor's programme pages, which quote the statute. Section numbers are cited as those documents cite them.
• Sales tax filing-frequency thresholds (dollar tests for monthly / quarterly / yearly) were not found on a Comptroller page; the calendar states only that the Comptroller assigns frequency by letter.
• Franchise-tax economic nexus threshold for out-of-state entities was not read from a primary page and is not stated.
• Workers' compensation. The text of Labor Code § 406.033 (common-law defences) was not read from a primary page; the pathway cites the section as TDI-DWC's information sheet does and describes the consequence qualitatively.
• Form 301 instructions do not mention a home-state certificate of existence; the absence is reported with verify: true rather than asserted as a rule.
• Individual income tax and right-to-work entries cite the Governor's recruitment material and the registry respectively, not the Texas Constitution or Labor Code, and carry verify: true.
• Tribal pathway. County locations for the Alabama-Coushatta and Ysleta del Sur Pueblo reservations, the Restoration Act history and the 2022 Supreme Court decision were not confirmed on a primary page within this pass (the opinion URL tried returned 404) and carry verify: true. No TERO ordinance or fee schedule is published by any of the three tribes.
• TexAmericas Center legal form (a legislatively created authority) was not confirmed from statute; the site record carries verify: true.
• Sherman Progress Park acreage (3,300) comes from TxEDC's regional page, not SEDCO's own site, which lists the parks without acreage.
• JETI wage test. The statute sets 110% of the industry-sector average annual wage (QCEW); the engine's minWage supports only a county-average kind, so the record approximates and is flagged verify: true.
• R&D credit. The engine models a flat 8.722% of the research-spend assumption; the statutory base is incremental. Flagged in modelNote and verify: true.
• Texas Tax Code § 23.03 and Economic Development & Diversification In-State Tuition (already present under a shorter name) and JETI (already present under a different name) were left out of /tmp/tx-catalogue.json so the merge tool would not create duplicates; § 23.03 is a reporting requirement, not an incentive.
• Site certification. Texas operates no state certified-sites programme; the six sites are public-owner or Type A EDC properties selected from the owners' official pages, TxDOT port profiles and the FTZ record, not from a state certification list.

Florida (FL)

Florida — sources and provenance

Compiled for the Levenhall US Business Establishment Atlas, data/states/fl/ (pathways.js, sites.js, incentives.js). Every URL below was fetched and read on 18 September 2026. Where a page could not be reached (noted under Known limitations) the fact was either omitted or carried qualitatively with verify: true.

##### Tier 1 — primary state and federal publications

##### Tier 2 — utility site-readiness programmes, regional EDOs and mirrors

##### Known limitations

• Distances in sites.js are indicative. Great-circle miles from approximate installation and site centroids; the CFIUS rule measures from the installation boundary, which for Eglin AFB and Cape Canaveral SFS lies far from the centroid. Do not rely on them for a transaction decision.
• FTZ site coverage. Only NFMIP's FPL sheet states a zone number ("Foreign Trade Zone #64"). For the other six sites the zone grantee serving the county is named in stack and ftz is left null; the Board's OFIS database is a JavaScript application that could not be read.
• Name reservation. The Sunbiz fee schedule lists reservation fees (corporation $35, LLC $25) but the availability of reservations by entity type was not confirmed on a Division page; flagged verify.
• Local business tax. Chapter 205 authority is cited; county and municipal rates and receipt procedures are not carried. Industry licensing (DBPR, FDACS, DEP, DOH) is named but not sourced; the state's Open My Florida Business portal did not render for the fetcher.
• HIPI status is ambiguous. Section 288.108 is in force and FloridaCommerce lists the programme, but EDR's 2026 evaluation groups it with the 2023 repeals/realignments and no award range is published. Carried with verify: true and a placeholder amount.
• QRT and IWT per-trainee figures in mechanics are illustrative placeholders inherited from the headline catalogue; CareerSource Florida sets the QRT amount per application and publishes only the IWT $100,000 per-grant ceiling and 75% rate.
• Job Growth Grant Fund annual appropriation is not carried (set in each General Appropriations Act); the earlier "$74 million" claim was removed as unsourced.
• Miccosukee Tribe government website (tribe.miccosukee.com) did not resolve and miccosukee.com returned 403; contact details are flagged verify and only the BIA HEARTH Act entry is relied on. Neither nation's TERO position could be confirmed from pages read.
• SB 264 litigation status is described qualitatively without a case citation because no court document was fetched; flagged verify.
• EB-5 amounts: the USCIS Policy Manual page read states $1,050,000 / $800,000 with automatic increases from 1 January 2027; a search summary asserted an earlier inflation adjustment, which was not confirmed on a USCIS page and is not carried.
• Exploration Park acreage is not published on the Space Florida pages read and is omitted.
• Workers' compensation subsection numbering: two reads of s. 440.02 place "Employment" at subsection (20); older secondary material cites (17). The file cites the section and describes the subsection as "(20) in the current online statutes".
• Property tax millage is not carried for Florida; abatements and county rates are outside this dataset.
• eCFR blocked the fetcher (redirect to unblock.federalregister.gov); Appendix A was taken from the govinfo CFR annual edition and the Federal Register final rule, and the two definitions from the LII mirror.

Colorado (CO)

Colorado — sources and provenance (for SOURCES.md)

Compiled for Levenhall LLC, 18 September 2026, for data/states/co/pathways.js, data/states/co/sites.js and the deepened data/states/co/incentives.js. Every substantive figure in those files traces to a source below. Colorado's state web properties (coloradosos.gov, tax.colorado.gov, cdle.colorado.gov, famli.colorado.gov, dpt.colorado.gov) refuse scripted fetch clients; pages were retrieved with a browser user-agent and converted to text. leg.colorado.gov bill pages were read through a second fetch client after returning HTTP 406 to the first. Where a page could not be reached at all it is listed under Known limitations rather than relied on.

##### Tier 1 — primary state, federal and tribal publications

##### Tier 2 — regional economic development organisations and other secondary sources

##### Known limitations

• Individual and corporate rate for tax year 2026 is not yet published by the Department of Revenue; the atlas shows the 2025 rate (4.4%) and notes the § 39-22-627 mechanism that temporarily cut 2024 to 4.25%.
• CFIUS distances are place-centroid to Colorado Springs city centroid (Census 2024 Gazetteer), not installation-boundary measurements. Peterson SFB lies at the city's eastern edge and Schriever SFB further east; the Census military-installation gazetteer is no longer published, and the Space Force fact-sheet pages were unreachable. Boulder (~84 mi) and Windsor (~112 mi) are therefore marked as requiring precise measurement; Windsor additionally sits in a county carrying Part 3 township-and-range areas that were not resolved to parcel level (verify: true).
• Enterprise Zone coverage of individual sites (Peak Innovation Park, Colorado Air and Space Port, Great Western Industrial Park) was not resolved against OEDIT's interactive map, which is not text-readable; each record says so.
• Southern Ute TERO fee rate is set by tribal resolution and published as a "Quarterly Fee Update" that was not retrieved; the step is marked verify: true.
• EB-5 investment amounts were not carried: the USCIS programme page returned pre-2022 figures inconsistent with the EB-5 Reform and Integrity Act; the step carries the 10-job requirement only and verify: true. L-1 is described without figures.
• Skill Advance Colorado (Colorado First / Existing Industry) publishes no per-trainee cap on OEDIT's page; the previously modelled $1,200 per job was withdrawn and the programme is now listed, not modelled.
• DPT's exemptions index page still prints the $52,000 threshold; the declaration-schedules page states $56,000 for tax year 2026 and is used. DPT pages render Drupal warnings but the statements are intact.
• FTZ 112 (Colorado Springs) is recorded as Not Active on the FTZ Board's OFIS database; the site records say so. No claim is made about reactivation.
• PEDCO acreage for the three city-controlled parks is not published on the pages read; PEDCO also announced its President & CEO's resignation in May 2026 and, per secondary reporting not relied on, a change in its contract with the City of Pueblo — confirm the current counterparty.
• Two OEDIT catalogue records carried from August 2026 — "Advanced Industries Business Accelerator Grant" and "Small Manufacturer's AdvantEDGE Training" — do not appear on OEDIT's September 2026 programme index and are flagged verify: true rather than removed, per the no-removal rule. Two pre-existing duplicate records for the Advanced Industry Investment Tax Credit and two for Skill Advance/Colorado First are likewise retained and cross-noted.
• data/_registry.js for Colorado lists seven CFIUS entries (Parts 1 and 2). The new Part 3 county areas (F.E. Warren missile field) are carried in sites.js and pathways.js but the registry was not edited.

Maryland (MD)

Maryland — sources and provenance

Compiled for Levenhall LLC, 18 September 2026, for data/states/md/pathways.js, data/states/md/sites.js and the deepened data/states/md/incentives.js. Every figure in those files traces to a URL below; the source field on each record names the one actually read. Where a fact could not be traced to a primary source it was omitted or marked verify: true (listed under Known limitations).

##### Tier 1 — primary state and federal publications

##### Tier 2 — county economic development organisations and official county/planning sources

##### Known limitations

• CFIUS Appendix A: the govinfo XML fetch returned five Maryland Part 1 entries and both Part 2 entries; the LII mirror returned all eight Part 1 entries but no Part 2 entries. Joint Base Andrews, NSF Carderock and NSF Indian Head therefore rest on the LII text and the atlas registry only and are flagged verify: true in OK_CFIUS. No great-circle distances were computed; every site carries miles: null and a qualitative proximity note.
• JCTC statute: Commerce prints "Maryland Code, Tax – General § 6-301"; the credit's programme provisions sit in the Economic Development Article, Title 6, Subtitle 3. The record quotes Commerce and asks for the article to be verified.
• Individual income tax brackets: the December 2025 Tax Alert and the 2026 withholding memo state the 6.25% bracket thresholds differently ($250,001–$500,000 vs $500,001–$1,000,000 single). Both are reproduced and the item is verify: true. The 2% capital-gains surtax threshold was not read.
• EB-5 investment amounts: the USCIS page as fetched shows the vacated 2019 figures; amounts are omitted and the step is verify: true.
• Workers' compensation: the WCC Q&A states "with few exceptions" every employer with one or more employees must insure; the exceptions were not enumerated.
• Local licensing: Maryland has no statewide business licence; trader's licences are county/clerk matters. The domestic steps 9–10 and the foreign/international local steps are verify: true.
• Tribal: what state recognition confers in Maryland procurement is not stated in any source read; step 3 is verify: true.
• Sites: Riverside Research Park (Frederick) and Fort Detrick acreage/workforce, Aberdeen Proving Ground workforce, Fort Meade acreage, and any Tradepoint Atlantic Enterprise Zone status were not found on an official page and are omitted; those sites carry verify: true. Harford County states the Aberdeen/Havre de Grace zone at 9,738 acres while the Commerce 2025 report prints 9,797; both are shown. The Aberdeen/Havre de Grace zone's published expiry (14 June 2026) has passed relative to some lease horizons — redesignation must be confirmed.
• Maryland has no certified-sites programme; the sites file says so and describes nothing as certified.
• Incentives: md-cybersecurity-investment-incentive-tax-credi, md-cybersecurity-investor-incentive-tax-credit, md-rise-zone-tax-credit, md-businesses-that-create-new-jobs-tax-credit, md-hire-our-veterans-tax-credit, md-maryland-industrial-partnerships-program, md-tedco-seed-investment-funds and md-maryland-economic-development-corporation-fi are verify: true (superseded, transferred, or no official page located). More Jobs for Marylanders is closed to new enrolment and its mechanics were removed so the model no longer counts it. PWQ keeps a placeholder $1,000-per-trainee assumption, labelled as such.
• The 2017 Commerce brochure remains the only source for several catalogue-only names; it is nine years old.

California (CA)

California — sources and provenance

Compiled for the Levenhall US Business Establishment Atlas, data/states/ca/ (pathways.js, sites.js, incentives.js). Every substantive figure traces to one of the sources below. Pages were fetched and read on 18 September 2026; where a page could not be retrieved (HTTP 403) that is recorded under Known limitations and the affected record carries verify: true.

##### Tier 1 — primary state, federal and statutory publications

##### Tier 2 — regional economic-development organisations and secondary compilations

##### Known limitations

1. No CFIUS distances. The Oklahoma module carries great-circle miles from installation centroid to site. For California those figures would have had to come from memory, so miles is null on every site and inRange is a qualitative judgement. Otay Mesa's position relative to the Chocolate Mountain Aerial Gunnery Range extended range is recorded as "uncertain". A parcel-level measurement from the installation boundary is required before any transaction decision.
2. Three official pages returned HTTP 403 to every fetch method available in this session: mojaveairport.com (East Kern Airport District), cityoflompoc.com, and nahc.ca.gov. Mojave facts were taken from the FAA licence and FAA Environmental Assessment instead; Lompoc/Vandenberg facts from REACH Central Coast and the Commercial Space Master Plan; the NAHC role from Pub. Res. Code § 21080.3.1. The Victorville page was readable only through WebFetch; curl was refused.
3. Secretary of State fee schedule. The only downloadable schedule is dated June 2018; the current individual form pages confirm $100 / $70 / $25 / $20, and the $10 name-reservation fee is taken from the 2018 schedule with verify: true.
4. New Employment Credit. The statute and FTB both state availability for taxable years beginning before 1 January 2026. AB 2205 (analysed by FTB 19 February 2026) would extend it to 2031; its enactment status was not confirmed. mechanics have been removed from ca-nec and the record carries status and verify: true.
5. California Competes Grant. Thresholds are from the 2023-24 and 2021-22 FAQs; no FY 2026-27 grant appropriation was found on GO-Biz's programme page. Marked verify: true.
6. Film & Television Tax Credit and Economic Development Rate. The Film Commission's page yielded no rate or allocation text; the utility EDR percentages are not published by GO-Biz. Both entries keep prior values only as descriptions and carry verify: true; EDR is not modelled.
7. CAEATFA cap. The $100,000,000 annual figure comes from a GO-Biz footnote; the $15 million individual cap from CAEATFA's FAQ. Pub. Res. Code § 26011.8 could not be displayed by leginfo (multiple-section selector) and was not read.
8. EB-5 investment thresholds are not quoted anywhere in the California module because the USCIS page read does not state them.
9. Tribal profile depth. Consistent with the brief, California's 109 tribes are not profiled individually; TERO and vendor-preference obligations are marked verify: true because no primary source lists them statewide. The registry's tribes count (108) differs from the Judicial Council's 109 and the BIA Pacific Region's 104/105; the pathway quotes the sources rather than reconciling them.
10. Local incentive programmes (City of San Diego Business Incentive Program, Chula Vista, Kern County AdvanceKern) are described only as the regional EDCs describe them; programme guidelines were not read.
11. Catalogue merge. Seven records were added from the GO-Biz Business Incentives List at confidence: "reported" with no mechanics; the CIIP row was already present. Rows that would have duplicated existing entries under a different name (Film & TV 4.0, Opportunity Zones 2.0, CALeVIP window 3) were deliberately not submitted.

Alabama (AL)

Alabama — sources and provenance

Compiled for the Levenhall US Business Establishment Atlas, data/states/al/ (pathways.js, sites.js, incentives.js). Every substantive figure traces to one of the sources below. Where a source is a primary state or federal publication, that is noted; where a claim rests on a local economic development organisation's own page, that is noted too, because the two do not carry the same weight. All retrieved Sep 2026 (18 September 2026).

Access notes. Justia and the ALISON code browser refused automated retrieval (HTTP 403 / JS shell); statutory text was therefore taken from the enrolled act as hosted by the Alabama Department of Revenue and from the Legislature's own enrolled-bill PDFs. The eCFR appendix was read through the eCFR versioner API (XML) for the same part; the human-readable URL is cited. PDFs that the fetch tool could not render were downloaded and text-extracted locally with pdftotext.

##### Tier 1 — primary state and federal publications

##### Tier 2 — local EDO, grantee and tribal publications (noted as such)

##### Known limitations

• Statutory text. Justia and the ALISON code viewer blocked retrieval, so Jobs Act sections are quoted from the enrolled HB241 (2023) PDF hosted by ALDOR, whose OCR is imperfect; the act number (2023-34 per ALDOR's file name) is not confirmed against the enrolled header. Post-2023 amendments (e.g. Act 2026-533, cited on ALDOR's Growing Alabama page) were not read.
• Certificate of existence. The current SOS forms rely on the signatory's certification of existence rather than an attached home-state certificate; the SOS foreign-corporation page does not state an age requirement. Marked verify: true.
• EB-5 amounts. USCIS's EB-5 page as fetched still presented the 2019 rule's $1.8M / $900,000 figures, which were superseded by the EB-5 Reform and Integrity Act of 2022; because the two readings conflicted, no dollar amount is stated and the step is marked verify: true.
• Bank onboarding. No Alabama agency publishes bank account-opening requirements; the step is descriptive and marked verify: true.
• UI rate vintage. The 2.70% entry rate is from the 2021 Employer Handbook; the $8,000 wage base is confirmed on the current FAQ. No current-year rate schedule page was found on workforce.alabama.gov (the legacy UC employer pages return 404).
• E-Verify. The general employer E-Verify mandate (Ala. Code § 31-13-9 / -15) was not read in statute; the atlas cites ALDOR's requirement of E-Verify enrolment documentation for Chapter 9B filings.
• Tribal land base. The Tribe's pages do not publish reservation acreage, an enrolment count, a full list of trust parcels or a TERO fee schedule; the tribal pathway is marked verify: true at those points.
• AdvantageSite inventory. advantagealabama.com is JavaScript-rendered and could not be read; the site list was assembled from state, city and county-EDO announcements. Trussville acreage (74.5 acres in secondary reports) is not stated on the EDPA page read and is flagged. Barton Riverfront's AdvantageSite status was not stated in the state article (advantageSite: null).
• FTZ coverage of Etowah and Jefferson parcels. FTZ 98's service area was not read; ftz: null with a note.
• CFIUS distances are great-circle from approximate centroids and indicative only, as in the Oklahoma module. Whether the Port of Mobile is a 'covered port' under § 802.210 was not verified against the DOT/BTS list.
• Property-tax abatement model. The 75% / 20-year figure in al-abatements.mechanics is a modelling assumption for 'all non-educational millage', not a published number; local millage splits are not in this dataset.
• Sales tax combined average. ALDOR (≈8%) and the registry's Tax Foundation figure (9.46%) differ in method; both are shown.
• Registry. data/_registry.js for al carries a mangled CFIUS label ("Andersen Air Force Base Yigo, Guam. Anniston Army Depot Anniston") and a one-line foreign-land summary ("prohibits foreign individuals and entities from acquiring agricultural land") that does not match HB379's foreign-principal definition. Neither was edited here; sites.js supplies a corrected CFIUS block that overrides the registry at load time.
• Manifest. data/_manifest.js lists al without pathways and sites; the loader will not pick up the new modules until build.py restamps the manifest (not run, per instructions).

Virginia (VA)

##### Virginia (VA) — sources for the deepened pathways, sites and incentives layer

Compiled 18 September 2026 for Levenhall LLC. Every URL below was read on that date. VEDP's own site (vedp.org) refused automated retrieval (Cloudflare) from every tool available; its programme pages were read from Internet Archive captures of the same URLs, and each such citation carries the capture date. Where a fact could not be tied to a primary source it is omitted or marked verify: true in the data files.

###### Tier 1 — primary state and federal publications

###### Tier 2 — locality and regional EDO publications (each the owner's own page)

###### Known limitations

• VEDP pages read from archive captures, not live. Live vedp.org blocked every fetch. Captures are from May–August 2026; the data files say so on each citation. Re-read live when access is possible.
• Stock-corporation annual registration fee conflict. The SCC's published schedule dated 02/26 shows $100 (1–5,000 shares) to $1,700 (over 270,000); the text of § 13.1-775.1 on LIS reads $50 plus $15 per 5,000 shares to $850. Both are quoted; the item is marked verify.
• Business Ready Sites tiers. Only MAMaC states its VBRSP tier (Tier 4) on its own page. VEDP's certified-sites list gives parcel sizes and zoning but the tier column was not in the text read. Tiers for the other five sites are null and flagged. The governor's August 2024 and August 2025 VBRSP award releases returned 404.
• Progress Park detail. The Joint IDA's pages returned 404; only the county's page was read. Utility capacities, rail and available lots are not stated in the data file.
• Distressed-locality designations (which drive the COF 25/15-job tiers, VIP/Talent Accelerator 85% wage rule and the data-centre $70M/10-job tier) are published by VEDP as a map that could not be read; the engine applies only the general thresholds.
• Machinery and tools tax and BPOL are local; no Virginia locality rates are carried, so property-type incentives and the M&T burden are described, not modelled.
• Port incentives are recorded as sunset per the statutes as read (taxable years before 1 January 2025; EID grant applications by 31 December 2024). VEDP's index still lists them; any 2025–2026 re-enactment must be checked on LIS.
• CFIUS distances are place-centroid to place-centroid from the Census Gazetteer, indicative only; Virginia has no Part 2 installations so only the one-mile test and the covered-port rule apply. The registry's Virginia CFIUS list lacks Fort Gregg-Adams, which the current eCFR Appendix A includes.
• Registry foreign-land flag. The registry records Virginia as "Prohibits foreign ownership"; the statutes read show a permissive general rule with a foreign-adversary agricultural-land ban only.
• Tribal layer. No TERO ordinance, preferred-vendor programme or tribal incentive could be sourced for any Virginia tribe; the pathway says so rather than importing Oklahoma assumptions. Pamunkey's federal acknowledgment is cited to the BIA's final-determination record (80 FR 39144, 8 Jul 2015); the frequently quoted 28 Jan 2016 effective date was seen only in search summaries and is not asserted.
• VJIP and Talent Accelerator per-job values in mechanics are internal placeholders; VEDP publishes neither.
• Immigration and EIN steps are federal and quoted from USCIS/IRS pages; the E-2 "typical investment" range used in the Oklahoma file was not reproduced because USCIS publishes no figure.