This site exists because 2003 was the year I stopped being able to find work in my own field. That's a personal claim, not a data point — so this report keeps the two separate. What BLS's own labor force numbers show: women's participation peaked in April 2000 and, as of August 2026, sits below where it was in 1990. Men's participation has been falling for decades. The H-1B cap tripled to 195,000 right as the dot-com bust hit U.S. tech workers, and a federal watchdog was already flagging the tracking gap in September 2003. And from a $34 million settlement in 2013 through a $3.2 million one in August 2026, DOJ and federal courts have repeatedly found specific companies did exactly what displaced workers have described for two decades: preferred visa holders over Americans.
I've been documenting the guest-worker visa system since I couldn't find work in my own field anymore. 2003 is the year that started for me. I can't turn that into a statistic — nobody tracks "the year an individual American stopped getting called back" — and I'm not going to pretend the aggregate numbers below prove my specific experience. What they do show is that 2003 wasn't a quiet year in the data either: it was the last year of a tripled H-1B cap, the same year GAO itself published a report because it couldn't tell whether the H-1B program was displacing U.S. workers, and roughly the year women's labor force participation — which had been climbing since the 1960s — turned and started coming back down. Draw your own conclusions about the overlap. I've laid out exactly what's verifiable below, and separated it from what's just my own account.
BLS's Labor Force Participation Rate (LFPR) measures the share of the civilian population 16 and over that is either working or actively looking for work — a broader, longer-running measure than the monthly employment-level swings covered in our companion report on the "98 of 100 jobs" claim. Pulling the full seasonally adjusted series (LNS11300001 for men, LNS11300002 for women) back to 1990 shows two distinct, decades-long trends — not a single-year event:
Two things worth being precise about. Men's participation has been declining continuously since long before 2003 — it was already down 3.2 points from 1990 to 2003, and kept falling at almost exactly the same pace afterward. That's a well-documented, decades-long trend with multiple established drivers (an aging population, rising disability rates, earlier retirement, longer time in school) that predates any single policy change and isn't something this report claims to fully explain. Women's participation is the sharper break: it rose steadily from the 1960s through 2000, then reversed. As of August 2026 it sits at 56.5% — below the 57.7% it stood at back in 1990, meaning the entire late-1990s gain has been given back and then some. The reversal's timing — a multi-decade rise that tops out and turns in exactly the 2000–2003 window — is worth noting precisely because it's unusual, not because this report can prove what caused it.
The American Competitiveness in the Twenty-First Century Act (AC21, signed December 2000) tripled the annual H-1B cap from its historic 65,000 to 195,000 for fiscal years 2001 through 2003 — a limit already raised once before, to 115,000, for FY1999–FY2000. This wasn't a quiet technical adjustment; it was Congress deliberately opening the door wider right as the labor market was about to turn.
The cap reverted to 65,000 starting FY2004 not because it had been maxed out — demand had already fallen well below 195,000 by FY2002 — but because AC21's temporary increase simply expired and Congress didn't renew it. The practical effect for anyone job-hunting in U.S. tech from 2001 to 2003: even as the dot-com bust triggered some of the largest domestic tech layoffs in a generation, H-1B approvals never dropped below roughly 1.2× the old 65,000-a-year baseline. The cap and the layoffs were running at the same time, not offsetting each other.
GAO's own headline finding in that report wasn't a smoking gun — it was an admission of a data gap: neither DOL nor the (then-new) Department of Homeland Security had the tracking in place to say how much the H-1B program was actually displacing U.S. workers, in either direction. That gap is exactly why this report leans on the DOJ and federal-court section below for anything it treats as proven rather than merely coincident in timing.
This is the part of the report that doesn't depend on timing or correlation. These are specific companies that DOJ or a federal court found — through settlement, jury verdict, or judicial ruling — had discriminated against U.S. workers or defrauded the visa system in favor of foreign labor. Four cases, thirteen years apart at the ends, spanning outsourcing, hardware, and now AI:
| Company | Year(s) | Finding / Status | $ / Outcome |
|---|---|---|---|
| Infosys | 2013 | DOJ settlement: systemic misuse of B-1 (visitor) visas for skilled/unskilled labor that should have required H-1B status; concealed this from clients while billing for "offshore" work actually done onshore; widespread I-9 compliance failures. | $34,000,000 |
| Cisco Systems | 2020–ongoing | California Civil Rights Dept. sued Cisco alleging it enabled caste-based discrimination against a Dalit engineer by upper-caste Indian-American managers. CRD dropped claims against the two individual managers, but the case against Cisco itself remains active; a federal court upheld California's authority to litigate caste discrimination in July 2025. | Ongoing |
| Cognizant Technology Solutions | 2017–ongoing | Palmer v. Cognizant (2:17-cv-06848, C.D. Cal.): a federal jury found Cognizant engaged in intentional discrimination against non-South Asian (race) and non-Indian (national origin) employees terminated "from the bench," meeting the standard for punitive damages. A December 2025 court order separately found Cognizant's "Visa Readiness" and "Visa Utilization" policies had a disparate impact on the same groups, class period Dec. 2016–Oct. 2022. Damages phase pending. | Liability found |
| OpenAI / Statsig | 2026 | DOJ Civil Rights Division settlement: fewer than 10 PERM (green-card) positions not posted to the public careers site, applications required by mail while other roles took electronic submissions, discouraging late-night radio ads — practices DOJ said discouraged U.S. workers from applying so visa holders could be preferred. | $3,200,000 |
OpenAI/Statsig: $1.2M in civil penalties, $2M to a back-pay/victim fund, three years of DOJ monitoring. Infosys: $10M civil forfeiture + $24M civil penalty, plus continued audit of its B-1/H-1B and I-9 practices. Sources linked below.
This investigation is produced independently — one person, 23 years of data work, living on $1,419/month Social Security. If this work matters to you, a small donation makes a real difference.
♥ Donate via Square More Reports