Reconciliation
Four numbers are called "the US tariff rate". They measure different things. What each one is, its latest stated value, and why the gaps exist. Updated every print.
| Source | What it measures | Cadence | Latest stated |
|---|---|---|---|
| ETRX | Realized: calculated duty assessed at entry as a share of customs value, by origin and HS chapter, imports for consumption, HS 98 and 99 excluded. First print settles. | Monthly, on the Census FT-900 date | 7.16% (August 2026) |
| Penn Wharton Budget Model | Realized effective rate from USITC DataWeb customs data, national aggregate with partner and product breakdowns. Research output, revised as published. | Irregular updates | 7.1% for June 2026 (post of August 10, 2026; China 23.2%, steel and aluminum 40.9%, vehicles 13.2%) |
| Yale Budget Lab | Statutory: the average tariff rate implied by policy as written, applied to trade weights, including actions announced but not yet collected. Measures the policy, not the border. | Updated with each policy action | budgetlab.yale.edu (statutory, not comparable one to one) |
| Treasury customs receipts | Cash deposited into the Treasury as customs duties, net of refunds, on a cash-timing basis (Monthly Treasury Statement). | Monthly | fiscaldata.treasury.gov (dollars, not a rate) |
Why the numbers differ
- Paper versus border. A statutory rate counts every duty the law imposes. A realized rate counts what was assessed on the goods that actually entered: exemptions, USMCA carve-outs, exclusions and trade shifting away from the highest rates all lower it. The gap is information, not error.
- Calculated duty versus cash. ETRX uses the duty fixed at entry. Cash receipts arrive later and are reduced by refunds; after the Supreme Court struck down the IEEPA tariffs in February 2026, a refund process exceeding $100B makes cash-based measures hard to read for years. Calculated duty is immune to it.
- China, same month, two shops. ETRX-CN (China, all chapters) printed 23.47% for June 2026; Penn Wharton states 23.2% for the same month from the same USITC data. 0.3 points apart, by two independent computations.
- Baskets. ETRX steel (HS 72 and 73, all origins) printed 35.87% in August 2026; Penn Wharton groups steel with aluminum. ETRX publishes vehicles per origin (EU, Mexico, Canada); Penn Wharton for all origins. Same data, different cuts.
- Revisions. Census revises trade data. ETRX never restates a first print and publishes revisions separately; research outputs generally republish the revised figure.
- Coverage. ETRX uses imports for consumption and excludes HS chapters 98 and 99 and AD/CVD duties. Aggregates that include them differ by construction.
External figures are quoted with their publication date and never restated by ETRX. Reproduce any ETRX value on the verify page.