What the additional source checks change

These notes record a source review completed on 28 October 2026. Existing data strengthen some arguments in the seven-area strategy and qualify others. Growth is concentrated, adoption depends on the task, and an active roster is not enough evidence of delivery.

The plan uses a small number of findings that bear on decisions. These notes keep the supporting evidence so the chapters do not have to carry every statistic. They do not assume any partner acceptance, baseline or budget.

Research investment and business participation

Arkansas nominal academic R&D rose 33.9% in FY2019–2024. Fayetteville and UAMS account for 99.7% of the net increase. The eight other matched reporters grew 0.94%.

UAMS's FY2024 total is imputed by the source and represents 46.1% of the state total. Excluding UAMS in both years changes growth to 23.7%. Division of Agriculture work is reported under Fayetteville, so these institutions do not map all work geographically.

Business-funded academic expenditure fell from $14.624m to $12.966m. These findings strengthen the case for shared access and wider participation. They do not explain why funding changed.

SBIR/STTR records show 32 source identities in each of 2015–2019 and 2020–2024. Of these, 17 appear in both windows and 15 appear only in the later one. Observation amounts rose 86.6%. Excluding each window's five largest identities leaves growth of 105.2%. The pattern shows turnover and greater funding depth while counts stayed flat. It says nothing about firm founding or commercialization.

Classified NSF crop commitments fell from $9.106m in FY2016–2020 to $1.551m in FY2021–2025. Excluding three older titles without obvious crop relevance and the largest earlier rice award still gives a 30.1% decline. These exclusions are title-level sensitivity checks and are not authoritative recodes.

Growth in food-related NSF commitments depends heavily on Cultivate IQ Phase 2, which supplies 84.9% of the later bucket. Neither grouping measures annual expenditure or proposal success.

Data files: Institution/cohort sensitivity · Funding sources · SBIR concentration · Identity turnover · Influential awards

Food and farm applications

In the 2023 County Business Patterns (CBP), Arkansas food manufacturing averages 178.4 jobs per establishment, compared with 54.8 nationally. In Arkansas, 14.5% of these establishments have 500 or more employees, compared with 2.24% nationally. This favors arrangements that both large plants and smaller businesses can use. Establishments are not parent companies.

The 2022 Census of Agriculture reports precision practices on 7.05% of all Arkansas farms, compared with 11.90% nationally. Applying US rates for each acreage size class to Arkansas farm counts implies 13.05%, so acreage mix alone does not explain the difference. Crop suitability, economics and returns are not measured.

Irrigation covers 34.61% of Arkansas farm acreage, compared with 6.24% nationally. This is a separate acreage measure. It is not evidence about precision-tool use by irrigators or about water withdrawals.

Farm Internet access is 81.91% in Arkansas, compared with 78.69% nationally. Access prevalence does not establish field coverage, speed, reliability or affordability. The adoption differences support testing with distinct groups of farms. They do not provide a count of customers or identify a single cause.

Data files: Plant structure · Farm measures · Acreage-size standardization

Operating context and reach

The 2022 benchmark in version 6 of the Freight Analysis Framework (FAF6) attributes 22.38% of Arkansas-origin and Arkansas-destination truck tonnage to food and agriculture commodities. A separate comparable FAF5 series shows cross-border truck tonnage falling 3.54% in 2018–2023, while within-state tonnage rose 1.18%. The two versions are not spliced together.

Scale and connections favor concrete shipment tests. Freight flows do not reveal empty miles, delays or demand for research.

Employment in selected hardware industries fell 8.9% in 2022–2025. Broad industrial electricity retail sales rose 21.1%, and the solar share of generation rose from 1.12% to 6.90%. These are different activities. Together they support investigating grid and control questions. They do not point to a particular buyer or a diagnosed bottleneck.

Primary-metals employment grew, while other material-producing industries moved differently. The metals measure does not describe the full strategic area.

Among clinical studies that have ever listed an Arkansas facility, 55.3% list Little Rock and no other Arkansas city. The same concentration appears among studies that recruit globally. Registry locations describe a clinical-research footprint. They do not measure rural access, current enrollment or patient need.

Data files: Food/farm freight · Comparable freight trends · Electricity changes · Industry output · Clinical geography

People and delivery

Among Arkansas employer businesses not planning to use AI in the next six months, 61.5% cite non-applicability, 21.6% privacy or security, 21.5% a lack of knowledge of AI's capabilities and 6.4% skilled workforce. These are conditional, multiple-response results. They favor checking for useful tasks and secure data before assuming that staffing is the answer. They cannot explain the statewide adoption gap, and they cannot be applied to all businesses or farms.

College data show existing routes in computing and in adjacent electronics and manufacturing fields. General-computing master's credentials rose from 217 to 778 in 2020–2024, and Southern Arkansas University accounts for 84.1% of the increase. This is a broad computing proxy and does not measure the availability of AI specialists.

The 2024–2034 occupational projection used here has 39 annual openings for electrical and electronics technicians: 37 from exits and transfers and two from growth. Credentials are awards and forecasts are projections. Neither establishes current vacancies or retained workers. Setting course content or hiring numbers would require a review of the competencies actually needed.

In the 2021–2025 researcher collaboration network used for the plan, 38 of 71 links between groups with at least two mapped researchers on each side still depend mainly on one researcher. Only one of nine clusters that meet the thresholds spans at least three institutions, and it has a materials and biomedical core. Coverage limits and thresholds leave out some staff and proprietary work.

These results support shared tasks and planned handovers. The rule of three institutions and two people is a policy choice. It is not an estimated optimum or a census of feasible teams.

Data files: AI uses and barriers · Projected openings · Industry/talent findings and definitions · Network checks

Selection and measurement

The joint removal tests use the fixed published baseline cells and the existing rule that an area needs two supported passes. Removing both the provisional economic-alignment criterion and the peer-plan context leaves food, crops and health eligible. Removing other pairs yields other portfolios.

These are chosen stress cases. They are not probabilities, and they do not propose a new rule. All seven areas remain strategic. Capability, demand and readiness each need their own evidence.

Other sources qualify broad accounts of progress. In the Business Dynamics Statistics (BDS) for 2015–2023, manufacturing employment rose 5.65% while the number of firms fell 4.99%. In the Business Formation Statistics (BFS) for 2024–2025, total applications rose, but high-propensity and wage-planned applications fell. Jobs and applications do not establish that the employer base is diversifying.

The Annual Business Survey (ABS) measure of product or process innovation covers 2021–2023. It reports innovation at 17.2% of covered Arkansas companies, compared with 23.3% nationally. The 2021–2023 window follows the source's own table titles.

In FY2025, Medicaid alone represents 56.4% of prime grant net obligations to Arkansas recipients. Contracts that the source codes as R&D (product and service code A) are a narrow, separate procurement subset. Broad federal totals are not research funding.

Compared with 2024, the 2025 Capital Scan reports 47% more venture capital deals but 33% fewer dollars, with 83% in five deals. The report retrospectively removed zero-dollar accelerator deals and reclassified sectors. The comparison therefore uses that edition's own history, since older editions cannot be spliced into it. None of these business, finance or patent measures supplies a commercialization success rate.

Data files: Joint selection cases · Employer breadth · Application composition · Business innovation · Capital by stage · R&D-coded contracts

Methods and coverage

This is a bounded review of the sources used in the plan. It is not an exhaustive analysis. Every source keeps its native period, geography, unit, suppression and revision status.

Commitments, expenditures, obligations, award observations and private financing cannot be added together. Any funding total for an area should be read alongside its classified and unclassified shares. Unknown results are treated as unknown, not as zero.

The 20-source depth ledger records how each source was used before, the depth this review added, and where further independent work is possible. The integration record shows which findings entered the main text and which remain supporting evidence. Source hashes and calculations are kept with the review.