Tarrant, San Bernardino, and Riverside counties pair the biggest veteran pools with above-average manufacturing/logistics job bases

Summary

The corpus's one table that isolates recently separated (Gulf War II / post-9/11) veterans at county grain — census.acs_veterans_by_period — is currently broken (every query returns an S3 404; confirmed on 3 separate calls and logged as a defect). Working around that gap with the best available proxy — total ACS veteran population by county, cross-referenced against BLS QCEW manufacturing + transportation/warehousing employment share — the counties that best combine a large veteran pool with a labor market genuinely built on manufacturing and logistics skills are Tarrant County, TX (Fort Worth), San Bernardino and Riverside Counties, CA (Inland Empire), Dallas County, TX, Duval County, FL (Jacksonville), and Wayne County, MI (Detroit) — all with 60,000+ veterans and a manufacturing+logistics employment share above the 14.4% national baseline (16.9%-24.1%). The counties with the single largest raw veteran totals (Los Angeles, Maricopa, San Diego) do not clear that bar — their job bases skew toward services, tourism, and defense/aerospace R&D rather than production and distribution work.

Two different things this question could mean, and why the answer differs

"Recently separated" and "total veteran population" are not the same population. A large share of the biggest total-veteran counties (Maricopa, Los Angeles, Clark, San Diego) are longstanding retirement destinations whose veteran counts skew toward Vietnam- and Gulf War I-era retirees, not people who left service in the last few years. Recently separated veterans concentrate much more heavily, in relative terms, near the installations that process the most separations every year — Fort Cavazos (Bell County, TX), Fort Liberty (Cumberland County, NC), Fort Campbell (Montgomery County, TN), and Camp Lejeune (Onslow County, NC) — where veterans are 18-22% of the adult population versus 3-10% in the big metro counties.

But those installation counties have thin manufacturing/logistics job bases of their own (11.2%, 12.9%, 13.1%, and 4.7% of private employment respectively — at or below the 14.4% national baseline). The manufacturing- and logistics-heavy labor markets that would actually employ those skills sit elsewhere: the Dallas-Fort Worth metroplex, the Inland Empire, metro Detroit, metro Houston, and Jacksonville. This is a well-documented dynamic — veterans separating near a base commonly relocate for work rather than staying in the installation county, and Dallas-Fort Worth in particular markets itself explicitly to transitioning service members for this reason.

The strongest overlap counties

National baseline: manufacturing (NAICS 31-33) plus transportation/warehousing (NAICS 48-49) is 14.4% of private U.S. employment. That figure is drawn from econ.national_wages for year 2025 specifically — an initial query against that table for year 2024 returned zero rows with a high-severity low_coverage diagnostic, because econ.national_wages currently has only the 2025 vintage loaded despite its schema declaring 1989-2025 as its window. The 2025 figure (133.2M total private jobs; 12.58M manufacturing + 6.62M transportation/warehousing) is used as the baseline instead, and is one year newer than the county-level QCEW figures (2024) it is being compared against — a one-year vintage mismatch that is immaterial at this scale (national mfg+logistics share moves by tenths of a percentage point year over year) but is disclosed here rather than blended in silently.

Data limitation and what would close it

census.acs_veterans_by_period is the only table in this corpus carrying period-of-service detail (Gulf War II/post-9/11 vs. earlier eras) at county grain, from ACS table B21002. Every query against it — including a plain SELECT * LIMIT 1 and describe_table itself — returns The specified key does not exist (S3, 404), reproduced three times. This is a genuine defect (the table is declared and cataloged as loaded, 2009-2023, but its underlying data file is missing from the object store), logged to the defect queue in this session. Separately, the Census Bureau's own ACS API requires an API key for this variable set that this session does not have, and the VA's own "Post-9/11 Veterans by County" heat-map dataset (data.va.gov) is published only as an .xlsx binary this session's tools cannot parse. Once acs_veterans_by_period is restored, the analysis above should be rerun substituting actual post-9/11 veteran counts for the total-veteran proxy used here — the ranking would likely tighten toward the installation-adjacent counties for pool size, while the manufacturing/logistics-share comparison would be unaffected.

What This Report Does Not Answer

• Identify counties with a large pool of recently separated veterans: The only county-grain table carrying period-of-service detail (census.acs_veterans_by_period, ACS B21002) is broken — S3 404 on every query, reproduced 3x and logged as a defect. Substituted total veteran population (all eras) as a scale proxy, and separately identified the installation counties (Fort Cavazos, Fort Liberty, Fort Campbell, Camp Lejeune) known to process the highest annual separation volumes as a qualitative cross-check.

Every query behind this report

7 warehouse calls ran in this session, in order. Each is reproducible against the same snapshot.

SELECT * FROM census.acs_veterans LIMIT 3

SELECT geo_name, county_fips, veterans, civilian_population_18_plus,
  ROUND(100.0*veterans/NULLIF(civilian_population_18_plus,0),1) AS veteran_pct
FROM census.acs_veterans
WHERE "year"='2023' AND geography='county'
ORDER BY veterans DESC
FETCH FIRST 25 ROWS ONLY

SELECT geo_name, county_fips, veterans, civilian_population_18_plus,
  ROUND(100.0*veterans/NULLIF(civilian_population_18_plus,0),1) AS veteran_pct
FROM census.acs_veterans
WHERE "year"='2023' AND geography='county'
  AND county_fips IN ('37051','48027','48099','21015','47125','37133','37141','40109')
ORDER BY veterans DESC

SELECT DISTINCT industry_code, agglvl_code FROM econ.county_qcew WHERE "year"='2024' AND area_fips='48027' AND own_code='5' AND (industry_code LIKE '31%' OR industry_code LIKE '32%' OR industry_code LIKE '33%' OR industry_code LIKE '48%' OR industry_code LIKE '49%' OR industry_code='10') ORDER BY industry_code

WITH vet AS (
  SELECT geo_name, county_fips, veterans
  FROM census.acs_veterans
  WHERE "year"='2023' AND geography='county'
    AND county_fips IN ('04013','06037','06073','48201','48029','32003','17031','06065','48439',
      '06059','12057','53033','08041','48113','12031','04019','53053','06071','12103','06067',
      '26163','12009','51059','12099','15003','48027','40109','37051','47125','37133','48099',
      '21015','37141')
),
qcew AS (
  SELECT area_fips,
    SUM(CASE WHEN industry_code='10' THEN annual_avg_emplvl END) AS total_emp,
    SUM(CASE WHEN industry_code='31-33' THEN annual_avg_emplvl END) AS mfg_emp,
    SUM(CASE WHEN industry_code='48-49' THEN annual_avg_emplvl END) AS logistics_emp
  FROM econ.county_qcew
  WHERE "year"='2024' AND own_code='5'
  GROUP BY area_fips
)
SELECT v.geo_name, v.veterans, q.total_emp, q.mfg_emp, q.logistics_emp,
  ROUND(100.0*(COALESCE(q.mfg_emp,0)+COALESCE(q.logistics_emp,0))/NULLIF(q.total_emp,0),1) AS mfg_logistics_pct
FROM vet v LEFT JOIN qcew q ON v.county_fips = q.area_fips
ORDER BY v.veterans DESC

SELECT own_code, industry_code, annual_avg_emplvl
FROM econ.national_wages
WHERE "year"='2024' AND own_code='5' AND industry_code IN ('10','31-33','48-49')

SELECT own_code, industry_code, annual_avg_emplvl
FROM econ.national_wages
WHERE "year"='2025' AND own_code='5' AND industry_code IN ('10','31-33','48-49')