What actually wins a tariff exclusion
Base rates from 52,746 decided Section 301 exclusion requests filed 2018–2020, computed per tranche, cross-checked against the published GAO figure, and corrected for a concentration problem that distorts every headline number published so far.
Read this before the numbers. As of 2026-09-09, no live US tariff action accepts product-specific exclusion requests with published outcomes. Every process that produced the rates below has closed. USTR was asked to create a new one in July 2026 and declined.
These are historical base rates under closed processes. They are not a forecast, and there is currently nothing live to forecast. See the live process map.
Grant rate by tranche
Rates are computed over decided requests and reported per tranche. They are never averaged across tranches, because each tranche ran a different process — a different filing system, a different form, and a different volume of requests.
| Tranche | Requests | Granted | Denied | Grant rate | Filing window |
|---|---|---|---|---|---|
| List 1 $34B action (July 6, 2018) |
10,814 | 3,656 | 7,158 | 33.8% | Jul 11 2018 – Oct 9 2018 |
| List 2 $16B action (August 23, 2018) |
2,869 | 1,074 | 1,795 | 37.4% | Sep 18 2018 – Dec 18 2018 |
| List 3 $200B action (September 24, 2018) |
30,283 | 1,497 | 28,786 | 4.9% | Jun 30 2019 – Sep 30 2019 |
| List 4A $300B action (September 1, 2019) |
8,780 | 575 | 8,205 | 6.5% | Oct 31 2019 – Jan 31 2020 |
The collapse between List 2 and List 3 is a process boundary, not a product story. Lists 1–2 were filed as documents to regulations.gov. Lists 3–4A were filed through a separate USTR portal with a longer form, and drew roughly three times as many requests. Reading the fall from ~35% to ~5% as evidence that later products were weaker would be a mistake.
The correction nobody has published
Request counts are extremely concentrated. The single largest filer accounts for 19.4% of all 52,746 requests on its own; the ten largest account for 32.7% and the hundred largest for 53.9%. Of 4,787 distinct filers, 1,967 filed exactly one request.
That means the widely-quoted headline rate is dominated by a handful of very large filers and does not describe the position of an importer filing a small number of requests. Both views:
| Tranche | All filers | Excluding filers with >100 requests | Requests in the corrected view |
|---|---|---|---|
| List 1 | 33.8% | 33.1% | 7,641 |
| List 2 | 37.4% | 30.2% | 1,920 |
| List 3 | 4.9% | 9.1% | 12,878 |
| List 4A | 6.5% | 7.6% | 5,225 |
| All four (cross-tranche) | 12.9% | 16.9% | — |
Counted per filer rather than per request, the median filer's grant rate was 0% and the mean was 17.6%. 26.2% of filers won at least one request.
On the last row. The cross-tranche figures blend four different processes and are shown only because they are the numbers in public circulation. Prefer the per-tranche rows for any real use.
Where to go next
- Live processes — what is open today, what closed and when, and the two dates that matter.
- By HTS heading — 706 four-digit headings, each with its own per-tranche rates.
- What importers argued — the 2024–25 machinery docket, the only Section 301 exclusion corpus whose argued grounds are public.
- Methods & sources — provenance, the GAO cross-check, and what this data cannot tell you.