Trimtabist

The US Trade Ledger — every US container gateway, every partner country, every month — and what each one actually paid at the border.

The "dodge premium" mostly is not real — and the metric that says it is, lies

published null · 2024-01 → 2026-06 · verdict: HYPOTHESIS FAILED — PUBLISHED AS A METHODS LESSON

What was registered, before the data was touched

The naive claim: trade migrates from China to Vietnam in proportion to the tariff-rate spread between them. Registered test: bucket shared HS6 codes by 2026 rate spread; the claim requires Vietnam-growth to rise monotonically with spread. Registered 2026-08-09.

What the record shows

Vietnam grew fastest (+170%) where the spread was smallest (<5 points) and China fell 38–53% in every bucket regardless of spread — the monotone relation the hypothesis requires is absent. Diagnosis: the biggest China→Vietnam moves (laptops, phones, telecom) happened in goods where China was barely tariffed at all — anticipation and supply-chain hedging, not collected duty. And the metric itself misleads: rate-on-dutiable is computed on a rounding error whenever the duty-free share is high, producing plausible-looking spreads on codes where almost nothing pays duty. A genuine rate-driven dodge exists in a smaller, duller set (cables, valves, transformers). The lesson is printed as method rule: burden and rate-on-dutiable are different instruments; quoting the wrong one manufactures findings.

China−Vietnam rate spreadHS6 codesVietnam growthChina change
under 5 points65+170%−53%
5–12 points88+6%−47%
12–20 points26+52%−52%
over 20 points120+29%−38%
If the dodge-premium story were true, the third column would rise down the table. It peaks in the first row — the biggest moves happened where the collected spread was smallest. Source: TT-010, 2024 vs 2026 trailing windows.

What would prove this wrong

A positive monotone spread→growth relation would have confirmed the original hypothesis; the measured relation is non-monotone with its largest value in the lowest bucket.

Cite this

In print: “The "dodge premium" mostly is not real — and the metric that says it is, lies” — a pre-registered event study by the Trimtab US Trade Ledger, ruled 2026-08-06, from US Census port records and 174,440 vessel stays (NOAA AIS, 2018–2024) — trimtabist.com/atlas/dodge-premium-methods
For the record: Event Atlas · dodge-premium-methods · HYPOTHESIS FAILED — PUBLISHED AS A METHODS LESSON · ruled 2026-08-06

This study is a frozen artifact: the ruling and its exhibits do not change when new data lands. If the verdict is ever revised, the revision is printed here with its reason, and the original stays on the record.

← all studies · sources: TT-008 port ledger · method & rules