Use official sources
Schedules, legal instruments and implementation notices are taken from official customs, tariff, trade-ministry or agreement sources. The latest confirmed material applicable to 2026 is used.
Calculations
Averages and AVEs
Both datasets use the same sequence, while the evidence available for each schedule may differ.
From a tariff line to a published average
1. Read the payable duty
Use the published percentage or interpret the full non-AV duty formula and legal conditions.
2. Convert non-AV duties
Calculate an AVE only when units, currency, customs value and quantity are compatible and sufficient.
3. Aggregate without changing weights
HS6 and product-group results preserve the contribution of the national tariff lines underneath them.
AVE in plain language: 100 × the duty payable on the imported quantity ÷ the customs value. When official tariff-line trade is unavailable, a documented fallback may use another supported year or trade-based unit value; the method and limitation remain attached to the estimate.
Exact calculation reference
Calculation formulas
These are the original formulas used in the MFN and Preferential methodology sources, reproduced without changing their notation. They remain separate where the two products apply different legal or publication rules.
MFN methodology
MFN formulas and released examples
Official tariff-line trade is used first. The original fallback and quota calculations are shown with the released examples used to explain them.
AVE calculation using official tariff-line trade
Duty-free and positive ad-valorem rates enter directly. For the formulas below, V is customs import value, Q is the matching quantity in the legal duty unit, d is the specific duty amount per q0 units, f converts the duty currency into the value currency, and t is an ad-valorem percentage. Every published operator is applied in its legal order.
Specific duty
The duty is converted into the value currency and divided by the exact customs value represented by the matching quantity.
Released example — Japan 110814091: the tariff is ¥119/kg. Customs recorded ¥33,805,000 and 255,000 kg in 2025, so 100 × 119 × 255,000 ÷ 33,805,000 = 89.7648%.
Compound duty
An ad-valorem component and a specific component are both payable, so their percentage incidences are added.
Accepted example — Canada 19012015: 8.5% + C$0.1193/kg; C$106,723,587 and 25,920,890 kg give a specific incidence of 2.8975%, hence 8.5 + 2.8975 = 11.3975%.
Alternate duty
“Whichever is higher” uses the maximum; “whichever is lower” uses the minimum. A compound branch is assembled before comparison.
Released example — Austria 08081010: cider apples imported 16 September–15 December carry 7.2% with a minimum of €0.36/100 kg. The specific incidence from €16,912,335 and 70,300,083 kg is 1.4964%; the minimum rule therefore selects 7.2%.
Floor, cap or clamp
The published base B, lower bound L and upper bound U are calculated separately, then applied in legal order.
Released example — Canada 02071191: C$1,446,556 ÷ 430,116 kg = C$3.3632/kg. The 4.74¢ floor is 1.4094%, the 9.48¢ cap is 2.8188%, and the base is 5%; clamping 5% to those bounds gives 2.8188%.
Seasonal or date-conditioned duty
The date condition remains part of the tariff line. Periods are combined only when matching period-specific trade exists.
Released example — Austria 08081010: the 7.2%-minimum formula applies only to bulk cider apples imported from 16 September through 15 December. Its released 7.2% is attached to that dated scope, not silently averaged with another season.
Content or supplementary-quantity duty
Qc must be the quantity named by the law—litres of alcohol, kilograms of sugar, grams of lead or item count—not an unrelated net weight.
Released example — New Zealand 22072023: the tariff includes a charge per gram of lead. The trade source does not provide grams of lead, so no AVE using official tariff-line trade is calculated; the all-method series uses a flagged median estimate from comparable country and product lines of 99.7537%.
Exchange-rate conversion
The duty amount is converted into the customs-value currency before incidence is calculated. The stored orientation is value-currency units per duty-currency unit, so it is never inferred by back-solving the AVE.
Sources: cross-currency AVEs based on official tariff-line trade use the documented official central-bank series (European Central Bank in this release). Estimated AVEs and the unit-value diagnostic use the World Bank period-average official exchange-rate series for the selected trade year, or an identified prior year. A factor of 1 for identical currencies is not treated as a separate exchange-rate observation.
Unit-value validity and plausibility checks. Value and matching legal-unit quantity must be positive, and the customs value must represent at least US$2,500 after currency conversion for this diagnostic. Where a same-unit world HS6 comparison is available, the release computes uTL = VUSD/Q, rho = uTL / world HS6 unit value, and the AVE implied by the world unit value. An observation is excluded from the official-tariff-line-trade series when rho > 1.40 and the world-denominator AVE exceeds the tariff-line AVE by more than 20 percentage points. Ratios below 0.20 or above 5.00 are also excluded as a conservative check. The original calculation remains in the audit data; the all-method series uses a same-unit world HS6 denominator or, if unavailable, a clearly labelled median from comparable lines. Gross weight is never treated as net weight. Lines without a comparable external unit value can remain in the official-trade series, with the absence of a comparison recorded. The sequence used to select an alternative denominator follows the WITS AVE methodology.
Excluded official-trade example — Canada 07061031: the published compound rate gives 4.2230% from only C$801 and 190 kg. The observation is below the US$2,500 information threshold, so 4.2230% is retained in the audit data but excluded from the official-trade series; the all-method series uses the same-unit world HS6 estimate of 5.0230%.
Published-formula verification. Duties stated per 1,000 kg or items are normalized with q0 = 1,000. Additive specific terms sharing the same mass basis are first converted to one per-kg duty and summed. Only extraction or parsing errors supported by source evidence are corrected, and each original value is retained in the row-level audit file; correctly normalized formulas are not changed.
Currency and legal-unit verification. Every selected AVE is checked against the published duty currency, the customs-value currency, the annual conversion-factor direction and the legal duty unit. Same-currency factors cancel and are never rescaled. A trade quantity is accepted only when it measures the legal base: ordinary litres cannot stand for litres of pure alcohol; ordinary net kilograms cannot stand for dry matter, drained weight or total-alcohol mass; and product mass cannot stand for sugar content, metal content, polarization degrees or proof litres. If the required denominator is unavailable, the AVE is excluded from the official-trade series and any value in the all-method series is identified as an estimate. In Fiji this corrects the Fiji-dollar conversion and prevents the wrong quantity from being used for all 12 alcohol-content lines. In the United States, six metal-content duties remain in the official-trade series because the official Census QTY2 field supplies content kilograms. In Turkmenistan, one published per-gram duty is corrected after it had been treated as per kilogram.
Cross-year HS classifications, unit equivalence, combined customs territories and exchange rates are documented in the source and calculation notes. Missing inputs are not inferred from the resulting AVE.
Official-trade AVEs must be reproducible. Each released value must be recomputable from the structured duty, trade and exchange-rate fields. Prose alone is not treated as proof.
Estimates when official tariff-line trade is unavailable
Where an AVE using official tariff-line trade is unavailable, the all-method series selects the first estimate that passes the stated checks in a fixed evidence sequence. Each row below identifies the evidence used, its calculation and the exact formula. Qualifications concerning continuity, exchange rates and data quality are retained in the source and calculation notes. The original candidate, selected estimate and reason for exclusion remain separate so the decision can be reviewed.
Reporter HS6 unit value
Cyprus 22042910: the legal duty is €32/hl. National tariff-line trade was unavailable, so 2025 reporter HS6 evidence—US$1,956,204.195 and 2,219,900 litres—supplies the broader unit value. Converting euros with the recorded period-average factor and normalizing one hectolitre to 100 litres gives 39.3051%. The selected method is Reporter HS6 unit value — current year, not official tariff-line trade.
Partner-reported unit value
Turkmenistan 0202201000: reporter imports were unavailable. Two partners reported 2025 exports to Turkmenistan of US$78,653 and 26,970 kg. The duty and reported value are both in US dollars, so f = 1. Applying the published US$0.15/kg duty gives 5.1435%. The source and calculation notes identify the partner exports, HS6 aggregation, currency identity and use of partner-reported trade.
FOB/CIF sensitivity: partner exports are normally FOB while import values are normally CIF. A lower FOB unit value makes the AVE upward-biased, all else equal. No evidence-based line-specific CIF factor was available, so the selected estimate stays unadjusted. A separate illustrative 10% CIF uplift gives 5.1435 ÷ 1.10 = 4.6759%; it is not the selected AVE. See the UN trade-valuation convention.
Global HS6 unit value
Gibraltar 24011035: neither national tariff-line nor usable reporter HS6 trade was available. The 2025 median unit value from 39 positive country reporters is applied to the published £0.10/kg component, with the recorded GBP/USD conversion, producing 1.8532%. The denominator is global rather than Gibraltar trade.
Median AVE from comparable lines
Åland 01022910: the formula is 10.2% + €93.10/100 kg, but no territory-specific denominator is released. The log-scale median of 94 accepted comparable lines with the same normalized formula and product group gives 29.2778%. This estimates the whole AVE rather than a tariff-line unit value, and is identified as “median, exact formula”.
Large AVEs and the optional 249% sensitivity. The published duty and selected raw estimate are never overwritten. Solomon Islands 24013000, for example, has a selected global-HS6 estimate of 22,966.3375% from its published SBD2,558.823/kg duty. A separate capped sensitivity stores 249% for analyses that would otherwise be dominated by that observation, while log(1 + AVE/100) compresses its scale without changing its rank. These are additional analytical variables, not alternative claims about the legal tariff. The 249% threshold is the 99.5th percentile of clean validated AVEs; results should be compared across raw, capped and logarithmic specifications.
Tariff-rate quotas
A published in-quota or out-of-quota AVE is a tier rate, not an effective blended rate. For the legal MFN series and its default simulation scalar, TINA uses the out-of-quota rate. This follows the WTO tariff-data convention, under which MFN tariffs for tariff quotas are the rates applied to quantities outside the quota. The published in-quota rate is retained as a separate scenario observation and is never averaged into the ordinary MFN rate merely because both tiers are printed in the schedule.
No-quantity fallback. If utilization cannot be measured for the same quota, product, unit and period, TINA does not omit the tariff and does not invent a 50/50 or fill-rate blend. The official out-of-quota MFN rate remains the default simulation treatment; the published in-quota rate remains available as a sensitivity. For an out-of-quota non-ad-valorem formula, the ordinary AVE evidence ladder applies. If its denominator fails the quality gate but the legal formula contains an ad-valorem floor, that floor may be released as an explicitly labelled lower-bound scalar. If neither an exact denominator nor a statutory floor exists, the last-resort simulation scalar is a flagged robust median from the narrowest available official-duty donor pool (same reporter and HS6, then HS4 and duty form, then wider peers). It is labelled modelled/low-confidence and never represented as an exact statutory percentage.
Preferred value-weighted incidence
Swiss customs territory, HS6 010121: the in-quota line has CHF7,904,210 at 0.2702%; the out-of-quota line has CHF17,048 at 44.9789%. Their value-weighted incidence is (7,904,210 × 0.2702 + 17,048 × 44.9789) ÷ 7,921,258 = 0.3665%.
Quantity-weighted diagnostic
For the same basket, 178 in-quota units and 2 out-of-quota units give (178 × 0.2702 + 2 × 44.9789) ÷ 180 = 0.7670%. This differs from 0.3665% because quantity weighting assumes equal unit values across tiers, so it is a sensitivity rather than the primary estimate.
Observed effective incidence is released only as a separate analytical scenario. The hierarchy is: (1) disjoint tier customs values and duties, value-weighted; (2) disjoint tier quantities, shown only as an equal-unit-value proxy; or (3) WTO AGIMS in-quota imports combined with a separately validated total-import denominator for the exact quota, product, unit and period, shown like (2) only as a quantity-share/equal-unit-value proxy. A fill rate is quota use divided by quota capacity—it is not, by itself, the in-quota share of total imports. The official-data calculation criteria currently support 126 independent Swiss-customs-territory HS6 groupings from 370 TN8 observations; they are not asserted to be legal quota-pool utilization groups.
Official references: WTO Tariff Analysis Online help; WTO tariff-rate quota dataset; WTO Consolidated Tariff Schedules, Section 1B; the WTO tariff-quota explanation; and UNCTAD's AVE methodology, which uses progressively wider official import-unit-value evidence when tariff-line denominators are unavailable. TINA's donor median is a disclosed simulation-only extension for formulas whose legal unit cannot be matched to trade quantities.
Preferential methodology
Preferential formulas and released examples
Offer coverage, legal MFN caps, standalone AVEs and complete-scope publication rules are retained separately from the MFN calculation path.
Publication and disclosure
The page includes every canonical 2026 agreement package selected for release. In a model-complete package, every applicable HS6 fact has a selected 2026 scalar: either a rate evaluated directly from the published schedule or a documented model fallback. A partial package leaves unresolved applicable facts explicitly withheld and suppresses any incomplete group average. Prohibited and legally inapplicable facts remain in the package but are excluded from the AVE denominator.
Direct and model-derived values are not conflated. The selected scalar, its method code, any legal-treatment or geometry issue code, the evidence year and the affected national and HS6 codes remain traceable in the package and in each agreement’s collapsed issue ledger.
Payable 2026 legal treatment and rate cross-check
Published duty-free treatment enters as zero. A direct ad-valorem preference uses the published 2026 percentage or the exact scheduled reduction/staging factor; when the importer may instead claim a lower current MFN rate, the selected scalar is the lower legally applicable rate and the MFN bridge is identified. Exclusions, no-concession lines, conditions, TRQs and prohibitions remain distinct legal treatments.
Legally supported MFN cap
This is a legal-rate bridge, not an AVE formula. It is used only where the agreement or implementing law permits the lower current MFN duty.
Matched-product MFN comparison
Mo contains only the offer's applicable released HS6 facts that have the same validated current MFN HS6. Each matched HS6 is weighted once in both means. If the recorded HS revisions differ, a pair enters Mo only through a bidirectional one-to-one official UNSD relation. Unmatched products are omitted and coverage is reported. The offer is never compared with an all-products MFN headline. This diagnostic is not a bilateral preference margin.
Standalone tariff-line AVE calculation
Only after the payable non-ad-valorem formula, unit, currency and legal operator are fixed is it converted to an ad-valorem equivalent. Currency and unit normalization remain visible as separate operands:
d is the published monetary duty in the duty currency per q0 legal units; f is value-currency units per duty-currency unit; d* is the normalized duty per one compatible unit; Q is matching import quantity; and V is customs value. q0 equals one only after that normalization is retained and reproducible.
Compound duty
The published ad-valorem component a is added to the evaluated specific component.
Higher/lower-of duty
Each complete payable branch is evaluated first; then the source's higher-of or lower-of operator is applied. A bare “or” is never guessed.
Strict evidence has an economic validity gate. Value and quantity must be positive and sufficiently informative; the quantity unit must match the legal unit; and the implied unit value must pass a documented unit-value plausibility check against compatible broader reporter and external unit values. A reproducible extreme AVE is not automatically accepted as strict evidence.
Ordered AVE fallbacks and rollback
Legal-rate bridge
First establish the payable 2026 legal treatment. Direct published rates come first. Where the agreement calls for staging, a reduction base, an exclusion/no-concession default, a lower-of-MFN rule, a cap or a conditional treatment, the bridge uses the current official reporter tariff line and only official UNSD mappings for nomenclature changes. The legal treatment, staging operator and MFN/default bridge receive their own codes; they are not AVE evidence methods.
AVE evidence hierarchy
After the legal non-AV formula is resolved, strict AVE evidence requires official value and exact legal-unit quantity for the same offerer, eligible recipient or recipient group, tariff-line scope, period and import flow, with official customs or central-bank exchange rates. If that gate fails, the separate model hierarchy broadens through prior exact official evidence, current/prior reporter HS6, reporter-world data, mirror evidence where appropriate, a global same-unit HS6 unit value, exact-formula donors, and then HS6, HS4, country or global duty-form medians. Reporter-world evidence is a model fallback for a bilateral preference, not recipient-specific strict evidence.
A mirror calculation keeps the FX term and legal quantity basis. Partner exports are ordinarily FOB while importer customs values are ordinarily CIF. An evidence-based FOB-to-CIF adjustment is applied when available; otherwise the result is labelled an unadjusted FOB mirror proxy, with the likely upward AVE bias disclosed. Peer medians are labelled modelled AVE proxies / peer-imputed AVEs, because they do not use the target tariff's own observed price.
Legal treatment remains separate from the numerical method. An exclusion or no-preference line uses the separately coded MFN evidence method. A conditional, end-use or special-goods line retains that condition and uses a separately coded conservative scalar. A prohibited line has no AVE, is marked not applicable and is omitted from averages. Thus the legal issue code and the numerical method code answer different questions and are never reused as aliases.
Every fallback can be rolled back to its source path because its method code, source year, source product level and any UNSD target codes are retained. The agreement ledger below each schedule groups those records without replacing the line-level files.
Actual released non-AV examples
The Maldives offer under the China–Maldives FTA contains two dynamic lower-of caps. The current calculation uses exact Maldives Customs tariff-line value and quantity but a non-official Comtrade exchange-rate factor; it is therefore retained as a model estimate, not a strict official AVE, pending an official customs or central-bank factor.
For cigarettes, the published MFN duty is 8 MVR per stick plus 50%. Official 2025 Maldives Customs imports provide 4,236,031.95 USD of value and 76,214,347 sticks. The 300% WTO bound caps the preferential AVE:
For the two matching soft-drink lines, official tariff-line trade produces MFN AVEs of 44.4134% and 37.1338%; the published 30% WTO bound therefore releases a 30% preferential AVE.
HS6 and summary averages
National tariff lines are evaluated first. The released HS6 value is the tariff-line-weighted mean of the selected 2026 scalars for applicable lines within an offer and HS6. Prohibited or otherwise not-applicable lines are counted separately and do not enter the denominator.
Agreement, offer and MTN-category simple averages reconstruct the applicable tariff-line mean by weighting each released HS6 mean by its applicable tariff-line count:
The group average is shown only when every applicable fact in that group has a finite selected 2026 scalar. Otherwise it is a dash; an evaluable-subset mean is not promoted as the released result.
TRQs, exclusions and non-applicable lines
TRQ is an orthogonal marker: the tariff line is counted once in the primary legal partition and separately in the TRQ column. The selected scalar is the published out-of-quota or default treatment unless observed utilization and compatible import quantities support a blend. In-quota tiers, quota quantities and eligibility details remain in the TRQ notes and are not averaged into a synthetic rate.
Exclusions and no-concession lines remain applicable and use the current legal/MFN treatment identified by their numerical method. Conditional lines remain applicable with the condition attached. Prohibited lines remain visible in counts and code-level notes but are not assigned an AVE and do not enter the denominator.
Comparison
Where MFN and preferential tariffs differ
The calculation discipline is shared, but the legal unit, recipient scope and rate selection are not.
| Question | MFN | Preferential |
|---|---|---|
| What is the record? | One reporter’s national MFN tariff schedule. | One agreement and its directional offers. |
| Who receives the rate? | Trading partners that do not receive a more favourable treatment. | Only the beneficiaries mapped to that agreement offer. |
| Which rate is selected? | The latest confirmed applied MFN treatment for 2026. | The agreement treatment payable in 2026, including the relevant staging step and conditions. |
| What are the primary sources? | Official national customs or tariff schedules. | Agreement schedules and legal instruments, supplemented by official implementation or customs sources. |
| How is the product universe defined? | By the reporter’s tariff schedule and its national nomenclature. | Separately for each offer or genuinely distinct schedule surface. |
| How are lower MFN rates handled? | The MFN rate is the reported treatment itself. | A lower current MFN rate is used only where the agreement legally permits that treatment. |
| What does coverage mean? | Available national tariff lines and their HS6 summaries for one reporter. | Available offers and HS6 values within an agreement; beneficiary mappings do not duplicate the schedule. |
| When is an average shown? | When the schedule’s included tariff lines have usable direct rates or supported AVEs. | When all applicable values in the selected agreement scope support a defensible average. |
Quick reference
Definitions
- MFN
- The general tariff treatment a reporter applies where no more favourable treatment governs.
- Offer
- The tariff treatment applied by one economy to one or more agreement partners.
- Tariff line
- The most detailed product code in a national tariff schedule.
- HS6
- A six-digit Harmonized System product category used for comparison.
- Non-AV
- A duty expressed in a form other than a percentage of value.
- AVE
- An estimated percentage equivalent of a non-AV duty.