Research · working paper
Backlog Gradient
The question. Does an Indian H-1B worker's chance of ever entering the employment-based green-card queue depend on where the employer sits in the staffing chain — a direct end-client, a prime vendor, or a sub-vendor? This page tests the employer-structure part of that question with public data alone. It is a record, not a verdict on any firm.
The answer. No. On the sponsorship margin, the data do not support it. Across the four complete years with clean data (FY2020–FY2023), an Indian worker's chance of the employer starting a PERM green-card case per H-1B approval shows no Tier 1 > Tier 2 > Tier 3 gradient — and in FY2021–FY2022 the ordering reverses, with sub-vendors sponsoring more. The prediction, registered before this analysis, is not met.
This concerns the entry (sponsorship) margin only. The exit margin (EB-1C/EB-5) is pending a USCIS I-140 FOIA return.
Pre-registration
Predicted direction. Sponsorship rate and exit share for Indian workers ordered Tier 1 > Tier 2 > Tier 3.
Supportive result. Tier 1 rate ≥ 2× Tier 3, bootstrap 95% interval on the ratio excluding 1.0, in ≥ 4 of the recent complete years.
Refutes it. Tiers within 1.25× of each other, or the ordering reversed, in a majority of years.
Controls. Employer size (log H-1B approvals), NAICS (4-digit), state, PERM education mix; gradient reported with and without.
Fixed verbatim from the project brief, authored before any data was ingested; committed to version control at docs/backlog-gradient/04-preregistration.md. Analysis run 2026-09-06.
The full analysis — result, chart, method, data & downloads
Everything below the pre-registration: the per-tier sponsorship rates and bootstrap intervals, the chart, the exit-share section, the tier-rule method, the dataset vintage, the limitations, and the CSV downloads.
*We're putting the finishing touches on the detailed analysis. Everything above is free to use right now — no account, nothing you enter is stored. Tapping the link above just tells us it's wanted (an anonymous tally — no email, nothing about you).
Educational research summary, not legal or immigration advice. A record, not a verdict. Report an error: [email protected].