The US research workforce is moderately concentrated geographically, and concentration has risen — driven mainly by patenting activity pulling toward California, not by academic R&D dollars

Summary

The US research workforce is moderately concentrated geographically, and that concentration has been rising over the past 15 years — though the two halves of "research workforce" move quite differently. Using BLS occupational data for scientists, engineers and software developers (2025), the 10 largest states hold 54.8% of the 5.78 million-person workforce (HHI = 491, Herfindahl "unconcentrated" band; Gini = 0.54). California alone holds 14.2%, well above its 11.6% share of the US population. Looking at how this has moved over time via two 2010-2024 proxies: academic R&D spending concentration has stayed essentially flat (Gini 0.554 to 0.566, HHI ~484 to ~488), tracking population fairly closely, while patenting activity — the clearest proxy for where invention actually happens — has concentrated sharply, with its HHI rising from 901 to 1,273 (crossing from "unconcentrated" into DOJ/FTC's "moderately concentrated" band) and California's share of US patent-inventors climbing from 25.3% to 32.6%, nearly triple its population share and still rising every year in the series.

Why three different sources, and what each one is actually measuring

No single table in this corpus carries a true research-occupation headcount with a multi-year history, so the concentration level and the change over time come from different, complementary sources — each disclosed here rather than blended silently:

• Current snapshot (level): BLS Occupational Employment and Wage Statistics (OEWS), state x occupation, for architecture/engineering (SOC 17-0000), life/physical/social science (19-0000), and software developers (15-1252) — the closest available proxy for "research workforce headcount." This table is loaded for 2025 only, so it answers the level question but not the trend question on its own.
• Trend, input side: NSF's Higher Education R&D (HERD) survey, academic R&D expenditure by state, 2010-2024 — a dollar measure, not headcount, but the standard NCSES series for where research spending (and by extension research staffing) sits.
• Trend, output side: USPTO PatentsView inventor locations joined to patent grants, 2010-2024 — inventor-credited patents by state and year. This captures where invention is actually happening, industrial R&D included, but omits research staff who do not appear as named inventors (many academic researchers, most support/technician roles).

All three agree on the state that leads (California) and on the direction implied by comparing the current snapshot's above-population-share concentration to the two trend series moving upward or flat — a reasonable convergence check given that no one table can do this alone.

The current level: moderately concentrated, but not extreme

By this snapshot, the S&E workforce is spread more evenly than, say, a typical concentrated industry — no single state dominates outright — but it is still meaningfully more concentrated than population, with California, Texas and New York alone accounting for over a quarter of the total.

The trend: concentration up, but only on the invention side

Academic R&D dollars (HERD) show a remarkably stable geography across 2010-2024: HHI moved only from 483.5 to 488.4, Gini from 0.554 to 0.566, and the top-5-state share from 39.0% to 40.0%. This is a slow drift, not a structural shift — funding follows the existing footprint of research universities, which does not change quickly.

Patent-inventor activity tells a different story. Its HHI rose from 901 in 2010 to 1,273 in 2024 — crossing the DOJ/FTC threshold from "unconcentrated" into "moderately concentrated." Its Gini rose from 0.671 to 0.712. The entire move is disproportionately a California story: California's share of US patent-inventors rose from 25.3% to 32.6% over the same window, while every other large state's share was comparatively stable or shrinking in relative terms. Because software and computing patents (increasingly concentrated in California-headquartered tech firms) have grown as a share of all US patenting over this period, this looks like an industry-composition effect as much as a pure geography effect — the workforce that files patents has shifted toward technology-sector R&D, which is itself geographically concentrated.

Caveats and excluded units

Every HHI/Gini computation in this report drops the four US territories (Puerto Rico FIPS 72, Guam 66, US Virgin Islands 78, Northern Mariana Islands 69) and American Samoa (60) to keep the comparison to the 50 states + DC, and drops rows with a null state code (e.g. foreign or unresolved patent-inventor locations). Naming the effect: including the three territories that actually carry S&E employment data (Puerto Rico 22,800; Guam 2,490; US Virgin Islands 410 — 25,700 combined, 0.44% of the US total) changes the 2025 workforce HHI only from 491.2 to 487.0 and California's own share from 14.17% to 14.11% — a negligible shift, because the excluded units are all far smaller than any state. The patent-inventor HHI/Gini series (2010-2024) does NOT exclude DC or territories — only null-state rows — so its n of 52-55 already includes them; territories are a trivial share of total US patent-inventor activity in every year, so this inclusion vs. exclusion choice does not materially change that series' reported trend either. In both cases the exclusion/inclusion choice moves the headline numbers by well under half a percentage point, so the reported figures are robust to it.

Every query behind this report

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

SELECT state_fips, SUM(employment) AS emp
FROM econ.state_occupation_employment
WHERE year = 2025 AND state_fips IN ('66','78','72','69','60')
GROUP BY state_fips

SELECT state_fips, SUM(employment) AS research_employment
FROM econ.state_occupation_employment
WHERE year = 2025
GROUP BY state_fips